Copy package from geth fork
This commit is contained in:
parent
88a2405805
commit
bb00c46735
323
README.md
323
README.md
@ -1,3 +1,324 @@
|
||||
# plugeth-statediff
|
||||
|
||||
cerc/go-ethereum statediff functionality packaged as a plugeth plugin
|
||||
`cerc/go-ethereum` statediff functionality packaged as a plugeth plugin.
|
||||
|
||||
## Package
|
||||
|
||||
This package provides an auxiliary service that asynchronously processes state diff objects from chain events,
|
||||
either relaying the state objects to RPC subscribers or writing them directly to Postgres as IPLD objects.
|
||||
|
||||
It also exposes RPC endpoints for fetching or writing to Postgres the state diff at a specific block height
|
||||
or for a specific block hash, this operates on historical block and state data and so depends on a complete state archive.
|
||||
|
||||
Data is emitted in this differential format in order to make it feasible to IPLD-ize and index the _entire_ Ethereum state
|
||||
(including intermediate state and storage trie nodes). If this state diff process is ran continuously from genesis,
|
||||
the entire state at any block can be materialized from the cumulative differentials up to that point.
|
||||
|
||||
## Statediff object
|
||||
|
||||
A state diff `StateObject` is the collection of all the state and storage trie nodes that have been updated in a given block.
|
||||
For convenience, we also associate these nodes with the block number and hash, and optionally the set of code hashes and code for any
|
||||
contracts deployed in this block.
|
||||
|
||||
A complete state diff `StateObject` will include all state and storage intermediate nodes, which is necessary for generating proofs and for
|
||||
traversing the tries.
|
||||
|
||||
```go
|
||||
// StateObject is a collection of state (and linked storage nodes) as well as the associated block number, block hash,
|
||||
// and a set of code hashes and their code
|
||||
type StateObject struct {
|
||||
BlockNumber *big.Int `json:"blockNumber" gencodec:"required"`
|
||||
BlockHash common.Hash `json:"blockHash" gencodec:"required"`
|
||||
Nodes []StateNode `json:"nodes" gencodec:"required"`
|
||||
CodeAndCodeHashes []CodeAndCodeHash `json:"codeMapping"`
|
||||
}
|
||||
|
||||
// StateNode holds the data for a single state diff node
|
||||
type StateNode struct {
|
||||
NodeType NodeType `json:"nodeType" gencodec:"required"`
|
||||
Path []byte `json:"path" gencodec:"required"`
|
||||
NodeValue []byte `json:"value" gencodec:"required"`
|
||||
StorageNodes []StorageNode `json:"storage"`
|
||||
LeafKey []byte `json:"leafKey"`
|
||||
}
|
||||
|
||||
// StorageNode holds the data for a single storage diff node
|
||||
type StorageNode struct {
|
||||
NodeType NodeType `json:"nodeType" gencodec:"required"`
|
||||
Path []byte `json:"path" gencodec:"required"`
|
||||
NodeValue []byte `json:"value" gencodec:"required"`
|
||||
LeafKey []byte `json:"leafKey"`
|
||||
}
|
||||
|
||||
// CodeAndCodeHash struct for holding codehash => code mappings
|
||||
// we can't use an actual map because they are not rlp serializable
|
||||
type CodeAndCodeHash struct {
|
||||
Hash common.Hash `json:"codeHash"`
|
||||
Code []byte `json:"code"`
|
||||
}
|
||||
```
|
||||
|
||||
These objects are packed into a `Payload` structure which can additionally associate the `StateObject`
|
||||
with the block (header, uncles, and transactions), receipts, and total difficulty.
|
||||
This `Payload` encapsulates all of the differential data at a given block, and allows us to index the entire Ethereum data structure
|
||||
as hash-linked IPLD objects.
|
||||
|
||||
```go
|
||||
// Payload packages the data to send to state diff subscriptions
|
||||
type Payload struct {
|
||||
BlockRlp []byte `json:"blockRlp"`
|
||||
TotalDifficulty *big.Int `json:"totalDifficulty"`
|
||||
ReceiptsRlp []byte `json:"receiptsRlp"`
|
||||
StateObjectRlp []byte `json:"stateObjectRlp" gencodec:"required"`
|
||||
|
||||
encoded []byte
|
||||
err error
|
||||
}
|
||||
```
|
||||
|
||||
## Usage
|
||||
|
||||
This state diffing service runs as an auxiliary service concurrent to the regular syncing process of the geth node.
|
||||
|
||||
### CLI configuration
|
||||
|
||||
This service introduces a CLI flag namespace `statediff`
|
||||
|
||||
`--statediff` flag is used to turn on the service
|
||||
|
||||
`--statediff.writing` is used to tell the service to write state diff objects it produces from synced ChainEvents directly to a configured Postgres database
|
||||
|
||||
`--statediff.workers` is used to set the number of concurrent workers to process state diff objects and write them into the database
|
||||
|
||||
`--statediff.db.type` is the type of database we write out to (current options: postgres, dump, file)
|
||||
|
||||
`--statediff.dump.dst` is the destination to write to when operating in database dump mode (stdout, stderr, discard)
|
||||
|
||||
`--statediff.db.driver` is the specific driver to use for the database (current options for postgres: pgx and sqlx)
|
||||
|
||||
`--statediff.db.host` is the hostname/ip to dial to connect to the database
|
||||
|
||||
`--statediff.db.port` is the port to dial to connect to the database
|
||||
|
||||
`--statediff.db.name` is the name of the database to connect to
|
||||
|
||||
`--statediff.db.user` is the user to connect to the database as
|
||||
|
||||
`--statediff.db.password` is the password to use to connect to the database
|
||||
|
||||
`--statediff.db.conntimeout` is the connection timeout (in seconds)
|
||||
|
||||
`--statediff.db.maxconns` is the maximum number of database connections
|
||||
|
||||
`--statediff.db.minconns` is the minimum number of database connections
|
||||
|
||||
`--statediff.db.maxidleconns` is the maximum number of idle connections
|
||||
|
||||
`--statediff.db.maxconnidletime` is the maximum lifetime for an idle connection (in seconds)
|
||||
|
||||
`--statediff.db.maxconnlifetime` is the maximum lifetime for a connection (in seconds)
|
||||
|
||||
`--statediff.db.nodeid` is the node id to use in the Postgres database
|
||||
|
||||
`--statediff.db.clientname` is the client name to use in the Postgres database
|
||||
|
||||
`--statediff.db.upsert` whether or not the service, when operating in a direct database writing mode, should overwrite any existing conflicting data
|
||||
|
||||
`--statediff.file.path` full path (including filename) to write statediff data out to when operating in file mode
|
||||
|
||||
`--statediff.file.wapath` full path (including filename) to write statediff watched addresses out to when operating in file mode
|
||||
|
||||
The service can only operate in full sync mode (`--syncmode=full`), but only the historical RPC endpoints require an archive node (`--gcmode=archive`)
|
||||
|
||||
e.g.
|
||||
`./build/bin/geth --syncmode=full --gcmode=archive --statediff --statediff.writing --statediff.db.type=postgres --statediff.db.driver=sqlx --statediff.db.host=localhost --statediff.db.port=5432 --statediff.db.name=cerc_testing --statediff.db.user=postgres --statediff.db.nodeid=nodeid --statediff.db.clientname=clientname`
|
||||
|
||||
When operating in `--statediff.db.type=file` mode, the service will write SQL statements out to the file designated by
|
||||
`--statediff.file.path`. Please note that it writes out SQL statements with all `ON CONFLICT` constraint checks dropped.
|
||||
This is done so that we can scale out the production of the SQL statements horizontally, merge the separate SQL files produced,
|
||||
de-duplicate using unix tools (`sort statediff.sql | uniq` or `sort -u statediff.sql`), bulk load using psql
|
||||
(`psql db_name --set ON_ERROR_STOP=on -f statediff.sql`), and then add our primary and foreign key constraints and indexes
|
||||
back afterwards.
|
||||
|
||||
### RPC endpoints
|
||||
|
||||
The state diffing service exposes both a WS subscription endpoint, and a number of HTTP unary endpoints.
|
||||
|
||||
Each of these endpoints requires a set of parameters provided by the caller
|
||||
|
||||
```go
|
||||
// Params is used to carry in parameters from subscribing/requesting clients configuration
|
||||
type Params struct {
|
||||
IntermediateStateNodes bool
|
||||
IntermediateStorageNodes bool
|
||||
IncludeBlock bool
|
||||
IncludeReceipts bool
|
||||
IncludeTD bool
|
||||
IncludeCode bool
|
||||
WatchedAddresses []common.Address
|
||||
}
|
||||
```
|
||||
|
||||
Using these params we can tell the service whether to include state and/or storage intermediate nodes; whether
|
||||
to include the associated block (header, uncles, and transactions); whether to include the associated receipts;
|
||||
whether to include the total difficulty for this block; whether to include the set of code hashes and code for
|
||||
contracts deployed in this block; whether to limit the diffing process to a list of specific addresses.
|
||||
|
||||
#### Subscription endpoint
|
||||
|
||||
A websocket supporting RPC endpoint is exposed for subscribing to state diff `StateObjects` that come off the head of the chain while the geth node syncs.
|
||||
|
||||
```go
|
||||
// Stream is a subscription endpoint that fires off state diff payloads as they are created
|
||||
Stream(ctx context.Context, params Params) (*rpc.Subscription, error)
|
||||
```
|
||||
|
||||
To expose this endpoint the node needs to have the websocket server turned on (`--ws`),
|
||||
and the `statediff` namespace exposed (`--ws.api=statediff`).
|
||||
|
||||
Go code subscriptions to this endpoint can be created using the `rpc.Client.Subscribe()` method,
|
||||
with the "statediff" namespace, a `statediff.Payload` channel, and the name of the statediff api's rpc method: "stream".
|
||||
|
||||
e.g.
|
||||
|
||||
```go
|
||||
|
||||
cli, err := rpc.Dial("ipcPathOrWsURL")
|
||||
if err != nil {
|
||||
// handle error
|
||||
}
|
||||
stateDiffPayloadChan := make(chan statediff.Payload, 20000)
|
||||
methodName := "stream"
|
||||
params := statediff.Params{
|
||||
IncludeBlock: true,
|
||||
IncludeTD: true,
|
||||
IncludeReceipts: true,
|
||||
IntermediateStorageNodes: true,
|
||||
IntermediateStateNodes: true,
|
||||
}
|
||||
rpcSub, err := cli.Subscribe(context.Background(), statediff.APIName, stateDiffPayloadChan, methodName, params)
|
||||
if err != nil {
|
||||
// handle error
|
||||
}
|
||||
for {
|
||||
select {
|
||||
case stateDiffPayload := <- stateDiffPayloadChan:
|
||||
// process the payload
|
||||
case err := <- rpcSub.Err():
|
||||
// handle rpc subscription error
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
#### Unary endpoints
|
||||
|
||||
The service also exposes unary RPC endpoints for retrieving the state diff `StateObject` for a specific block height/hash.
|
||||
|
||||
```go
|
||||
// StateDiffAt returns a state diff payload at the specific blockheight
|
||||
StateDiffAt(ctx context.Context, blockNumber uint64, params Params) (*Payload, error)
|
||||
|
||||
// StateDiffFor returns a state diff payload for the specific blockhash
|
||||
StateDiffFor(ctx context.Context, blockHash common.Hash, params Params) (*Payload, error)
|
||||
```
|
||||
|
||||
To expose this endpoint the node needs to have the HTTP server turned on (`--http`),
|
||||
and the `statediff` namespace exposed (`--http.api=statediff`).
|
||||
|
||||
### Direct indexing into Postgres
|
||||
|
||||
If `--statediff.writing` is set, the service will convert the state diff `StateObject` data into IPLD objects, persist them directly to Postgres,
|
||||
and generate secondary indexes around the IPLD data.
|
||||
|
||||
The schema and migrations for this Postgres database are provided in `statediff/db/`.
|
||||
|
||||
#### Postgres setup
|
||||
|
||||
We use [pressly/goose](https://github.com/pressly/goose) as our Postgres migration manager.
|
||||
You can also load the Postgres schema directly into a database using
|
||||
|
||||
`psql database_name < schema.sql`
|
||||
|
||||
This will only work on a version 12.4 Postgres database.
|
||||
|
||||
#### Schema overview
|
||||
|
||||
Our Postgres schemas are built around a single IPFS backing Postgres IPLD blockstore table (`ipld.blocks`) that conforms with [go-ds-sql](https://github.com/ipfs/go-ds-sql/blob/master/postgres/postgres.go).
|
||||
All IPLD objects are stored in this table, where `key` is the blockstore-prefixed multihash key for the IPLD object and `data` contains
|
||||
the bytes for the IPLD block (in the case of all Ethereum IPLDs, this is the RLP byte encoding of the Ethereum object).
|
||||
|
||||
The IPLD objects in this table can be traversed using an IPLD DAG interface, but since this table only maps multihash to raw IPLD object
|
||||
it is not particularly useful for searching through the data by looking up Ethereum objects by their constituent fields
|
||||
(e.g. by block number, tx source/recipient, state/storage trie node path). To improve the accessibility of these objects
|
||||
we create an Ethereum [advanced data layout](https://github.com/ipld/specs#schemas-and-advanced-data-layouts) (ADL) by generating secondary
|
||||
indexes on top of the raw IPLDs in other Postgres tables.
|
||||
|
||||
These secondary index tables fall under the `eth` schema and follow an `{objectType}_cids` naming convention.
|
||||
These tables provide a view into individual fields of the underlying Ethereum IPLD objects, allowing lookups on these fields, and reference the raw IPLD objects stored in `ipld.blocks`
|
||||
by foreign keys to their multihash keys.
|
||||
Additionally, these tables maintain the hash-linked nature of Ethereum objects to one another. E.g. a storage trie node entry in the `storage_cids`
|
||||
table contains a `state_id` foreign key which references the `id` for the `state_cids` entry that contains the state leaf node for the contract that storage node belongs to,
|
||||
and in turn that `state_cids` entry contains a `header_id` foreign key which references the `id` of the `header_cids` entry that contains the header for the block these state and storage nodes were updated (diffed).
|
||||
|
||||
### Optimization
|
||||
|
||||
On mainnet this process is extremely IO intensive and requires significant resources to allow it to keep up with the head of the chain.
|
||||
The state diff processing time for a specific block is dependent on the number and complexity of the state changes that occur in a block and
|
||||
the number of updated state nodes that are available in the in-memory cache vs must be retrieved from disc.
|
||||
|
||||
If memory permits, one means of improving the efficiency of this process is to increase the in-memory trie cache allocation.
|
||||
This can be done by increasing the overall `--cache` allocation and/or by increasing the % of the cache allocated to trie
|
||||
usage with `--cache.trie`.
|
||||
|
||||
## Versioning, Branches, Rebasing, and Releasing
|
||||
|
||||
Internal tagged releases are maintained for building the latest version of statediffing geth or using it as a go mod dependency.
|
||||
When a new core go-ethereum version is released, statediffing geth is rebased onto and adjusted to work with the new tag.
|
||||
|
||||
We want to maintain a complete record of our git history, but in order to make frequent and timely rebases feasible we also
|
||||
need to be able to squash our work before performing a rebase. To this end we retain multiple branches with partial incremental history that culminate in
|
||||
the full incremental history.
|
||||
|
||||
### Versioning
|
||||
|
||||
Example: `v1.10.16-statediff-3.0.2`
|
||||
|
||||
- The first section, `v1.10.16`, corresponds to the release of the root branch this version is rebased onto (e.g., [](https://github.com/ethereum/go-ethereum/releases/tag/v1.10.16)[https://github.com/ethereum/go-ethereum/releases/tag/v1.10.16](https://github.com/ethereum/go-ethereum/releases/tag/v1.10.16))
|
||||
- The second section, `3.0.2`, corresponds to the version of our statediffing code. The major version here (3) should always correspond with the major version of the `ipld-eth-db` schema version it works with (e.g., [](https://github.com/cerc-io/ipld-eth-db/releases/tag/v3.0.6)[https://github.com/vulcanize/ipld-eth-db/releases/tag/v3.0.6](https://github.com/vulcanize/ipld-eth-db/releases/tag/v3.0.6)); it is only bumped when we bump the major version of the schema.
|
||||
- The major version of the schema is only bumped when a breaking change is made to the schema.
|
||||
- The minor version is bumped when a new feature is added, or a fix is performed that breaks or updates the statediffing API or CLI in some way.
|
||||
- The patch version is bumped whenever minor fixes/patches/features are done that don’t change/break API/CLI compatibility.
|
||||
- We are very strict about the first section and the major version of the statediffing code, but some discretion is required when deciding to bump minor versus patch version of the statediffing code.
|
||||
|
||||
The statediff version is included in the `VersionMeta` in params/version.go
|
||||
|
||||
### Branches
|
||||
|
||||
We maintain two official kinds of branches:
|
||||
|
||||
Major Branch: `{Root Version}-statediff`
|
||||
Major branches retain the cumulative state of all changes made before the latest root version rebase and track the full incremental history of changes made between the latest root version rebase and the next.
|
||||
Aside from creating the branch by performing the rebase described in the section below, these branches are never worked off of or committed to directly.
|
||||
|
||||
Feature Branch: `{Root Version}-statediff-{Statediff Version}`
|
||||
Feature branches are checked out from a major branch in order to work on a new feature or fix for the statediffing code.
|
||||
The statediff version of a feature branch is the new version it affects on the major branch when merged. Internal tagged releases
|
||||
are cut against these branches after they are merged back to the major branch.
|
||||
|
||||
If a developer is unsure what version their patch should affect, they should remain working on an unofficial branch. From there
|
||||
they can open a PR against the targeted root branch and be directed to the appropriate feature version and branch.
|
||||
|
||||
### Rebasing
|
||||
|
||||
When a new root tagged release comes out we rebase our statediffing code on top of the new tag using the following process:
|
||||
|
||||
1. Checkout a new major branch for the tag from the current major branch
|
||||
2. On the new major branch, squash all our commits since the last major rebase
|
||||
3. On the new major branch, perform the rebase against the new tag
|
||||
4. Push the new major branch to the remote
|
||||
5. From the new major branch, checkout a new feature branch based on the new major version and the last statediff version
|
||||
6. On this new feature branch, add the new major branch to the .github/workflows/on-master.yml list of "on push" branches
|
||||
7. On this new feature branch, make any fixes/adjustments required for all statediffing geth tests to pass
|
||||
8. PR this feature branch into the new major branch, this PR will trigger CI tests and builds.
|
||||
9. After merging PR, rebase feature branch onto major branch
|
||||
10. Cut a new release targeting the feature branch, this release should have the new root version but the same statediff version as the last release
|
||||
|
206
api.go
Normal file
206
api.go
Normal file
@ -0,0 +1,206 @@
|
||||
// Copyright 2019 The go-ethereum Authors
|
||||
// This file is part of the go-ethereum library.
|
||||
//
|
||||
// The go-ethereum library is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Lesser General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
//
|
||||
// The go-ethereum library is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Lesser General Public License for more details.
|
||||
//
|
||||
// You should have received a copy of the GNU Lesser General Public License
|
||||
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package statediff
|
||||
|
||||
import (
|
||||
"context"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/types"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
"github.com/ethereum/go-ethereum/rpc"
|
||||
)
|
||||
|
||||
// APIName is the namespace used for the state diffing service API
|
||||
const APIName = "statediff"
|
||||
|
||||
// APIVersion is the version of the state diffing service API
|
||||
const APIVersion = "0.0.1"
|
||||
|
||||
// PublicStateDiffAPI provides an RPC subscription interface
|
||||
// that can be used to stream out state diffs as they
|
||||
// are produced by a full node
|
||||
type PublicStateDiffAPI struct {
|
||||
sds IService
|
||||
}
|
||||
|
||||
// NewPublicStateDiffAPI creates an rpc subscription interface for the underlying statediff service
|
||||
func NewPublicStateDiffAPI(sds IService) *PublicStateDiffAPI {
|
||||
return &PublicStateDiffAPI{
|
||||
sds: sds,
|
||||
}
|
||||
}
|
||||
|
||||
// Stream is the public method to setup a subscription that fires off statediff service payloads as they are created
|
||||
func (api *PublicStateDiffAPI) Stream(ctx context.Context, params Params) (*rpc.Subscription, error) {
|
||||
// ensure that the RPC connection supports subscriptions
|
||||
notifier, supported := rpc.NotifierFromContext(ctx)
|
||||
if !supported {
|
||||
return nil, rpc.ErrNotificationsUnsupported
|
||||
}
|
||||
|
||||
// create subscription and start waiting for events
|
||||
rpcSub := notifier.CreateSubscription()
|
||||
|
||||
go func() {
|
||||
// subscribe to events from the statediff service
|
||||
payloadChannel := make(chan Payload, chainEventChanSize)
|
||||
quitChan := make(chan bool, 1)
|
||||
api.sds.Subscribe(rpcSub.ID, payloadChannel, quitChan, params)
|
||||
// loop and await payloads and relay them to the subscriber with the notifier
|
||||
for {
|
||||
select {
|
||||
case payload := <-payloadChannel:
|
||||
if err := notifier.Notify(rpcSub.ID, payload); err != nil {
|
||||
log.Error("Failed to send state diff packet; error: " + err.Error())
|
||||
if err := api.sds.Unsubscribe(rpcSub.ID); err != nil {
|
||||
log.Error("Failed to unsubscribe from the state diff service; error: " + err.Error())
|
||||
}
|
||||
return
|
||||
}
|
||||
case err := <-rpcSub.Err():
|
||||
if err != nil {
|
||||
log.Error("State diff service rpcSub error: " + err.Error())
|
||||
err = api.sds.Unsubscribe(rpcSub.ID)
|
||||
if err != nil {
|
||||
log.Error("Failed to unsubscribe from the state diff service; error: " + err.Error())
|
||||
}
|
||||
return
|
||||
}
|
||||
case <-quitChan:
|
||||
// don't need to unsubscribe, service does so before sending the quit signal
|
||||
return
|
||||
}
|
||||
}
|
||||
}()
|
||||
|
||||
return rpcSub, nil
|
||||
}
|
||||
|
||||
// StateDiffAt returns a state diff payload at the specific blockheight
|
||||
func (api *PublicStateDiffAPI) StateDiffAt(ctx context.Context, blockNumber uint64, params Params) (*Payload, error) {
|
||||
return api.sds.StateDiffAt(blockNumber, params)
|
||||
}
|
||||
|
||||
// StateDiffFor returns a state diff payload for the specific blockhash
|
||||
func (api *PublicStateDiffAPI) StateDiffFor(ctx context.Context, blockHash common.Hash, params Params) (*Payload, error) {
|
||||
return api.sds.StateDiffFor(blockHash, params)
|
||||
}
|
||||
|
||||
// StreamCodeAndCodeHash writes all of the codehash=>code pairs out to a websocket channel
|
||||
func (api *PublicStateDiffAPI) StreamCodeAndCodeHash(ctx context.Context, blockNumber uint64) (*rpc.Subscription, error) {
|
||||
// ensure that the RPC connection supports subscriptions
|
||||
notifier, supported := rpc.NotifierFromContext(ctx)
|
||||
if !supported {
|
||||
return nil, rpc.ErrNotificationsUnsupported
|
||||
}
|
||||
|
||||
// create subscription and start waiting for events
|
||||
rpcSub := notifier.CreateSubscription()
|
||||
payloadChan := make(chan types.CodeAndCodeHash, chainEventChanSize)
|
||||
quitChan := make(chan bool)
|
||||
api.sds.StreamCodeAndCodeHash(blockNumber, payloadChan, quitChan)
|
||||
go func() {
|
||||
for {
|
||||
select {
|
||||
case payload := <-payloadChan:
|
||||
if err := notifier.Notify(rpcSub.ID, payload); err != nil {
|
||||
log.Error("Failed to send code and codehash packet", "err", err)
|
||||
return
|
||||
}
|
||||
case err := <-rpcSub.Err():
|
||||
log.Error("State diff service rpcSub error", "err", err)
|
||||
return
|
||||
case <-quitChan:
|
||||
return
|
||||
}
|
||||
}
|
||||
}()
|
||||
|
||||
return rpcSub, nil
|
||||
}
|
||||
|
||||
// WriteStateDiffAt writes a state diff object directly to DB at the specific blockheight
|
||||
func (api *PublicStateDiffAPI) WriteStateDiffAt(ctx context.Context, blockNumber uint64, params Params) JobID {
|
||||
var err error
|
||||
start, logger := countApiRequestBegin("writeStateDiffAt", blockNumber)
|
||||
defer countApiRequestEnd(start, logger, err)
|
||||
|
||||
return api.sds.WriteStateDiffAt(blockNumber, params)
|
||||
}
|
||||
|
||||
// WriteStateDiffFor writes a state diff object directly to DB for the specific block hash
|
||||
func (api *PublicStateDiffAPI) WriteStateDiffFor(ctx context.Context, blockHash common.Hash, params Params) error {
|
||||
var err error
|
||||
start, logger := countApiRequestBegin("writeStateDiffFor", blockHash.Hex())
|
||||
defer countApiRequestEnd(start, logger, err)
|
||||
|
||||
err = api.sds.WriteStateDiffFor(blockHash, params)
|
||||
return err
|
||||
}
|
||||
|
||||
// WatchAddress changes the list of watched addresses to which the direct indexing is restricted according to given operation
|
||||
func (api *PublicStateDiffAPI) WatchAddress(operation types.OperationType, args []types.WatchAddressArg) error {
|
||||
return api.sds.WatchAddress(operation, args)
|
||||
}
|
||||
|
||||
// StreamWrites sets up a subscription that streams the status of completed calls to WriteStateDiff*
|
||||
func (api *PublicStateDiffAPI) StreamWrites(ctx context.Context) (*rpc.Subscription, error) {
|
||||
// ensure that the RPC connection supports subscriptions
|
||||
notifier, supported := rpc.NotifierFromContext(ctx)
|
||||
if !supported {
|
||||
return nil, rpc.ErrNotificationsUnsupported
|
||||
}
|
||||
|
||||
// create subscription and start waiting for events
|
||||
rpcSub := notifier.CreateSubscription()
|
||||
|
||||
go func() {
|
||||
// subscribe to events from the statediff service
|
||||
statusChan := make(chan JobStatus, chainEventChanSize)
|
||||
quitChan := make(chan bool, 1)
|
||||
api.sds.SubscribeWriteStatus(rpcSub.ID, statusChan, quitChan)
|
||||
|
||||
var err error
|
||||
defer func() {
|
||||
if err = api.sds.UnsubscribeWriteStatus(rpcSub.ID); err != nil {
|
||||
log.Error("Failed to unsubscribe from job status stream: " + err.Error())
|
||||
}
|
||||
}()
|
||||
// loop and await payloads and relay them to the subscriber with the notifier
|
||||
for {
|
||||
select {
|
||||
case status := <-statusChan:
|
||||
if err = notifier.Notify(rpcSub.ID, status); err != nil {
|
||||
log.Error("Failed to send job status; error: " + err.Error())
|
||||
return
|
||||
}
|
||||
case err = <-rpcSub.Err():
|
||||
if err != nil {
|
||||
log.Error("statediff_StreamWrites RPC subscription error: " + err.Error())
|
||||
return
|
||||
}
|
||||
case <-quitChan:
|
||||
// don't need to unsubscribe, service does so before sending the quit signal
|
||||
return
|
||||
}
|
||||
}
|
||||
}()
|
||||
|
||||
return rpcSub, nil
|
||||
}
|
677
builder.go
Normal file
677
builder.go
Normal file
@ -0,0 +1,677 @@
|
||||
// Copyright 2019 The go-ethereum Authors
|
||||
// This file is part of the go-ethereum library.
|
||||
//
|
||||
// The go-ethereum library is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Lesser General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
//
|
||||
// The go-ethereum library is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Lesser General Public License for more details.
|
||||
//
|
||||
// You should have received a copy of the GNU Lesser General Public License
|
||||
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
// Contains a batch of utility type declarations used by the tests. As the node
|
||||
// operates on unique types, a lot of them are needed to check various features.
|
||||
|
||||
package statediff
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"fmt"
|
||||
"time"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/core/state"
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/crypto"
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
"github.com/ethereum/go-ethereum/rlp"
|
||||
metrics2 "github.com/ethereum/go-ethereum/statediff/indexer/database/metrics"
|
||||
ipld2 "github.com/ethereum/go-ethereum/statediff/indexer/ipld"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/shared"
|
||||
"github.com/ethereum/go-ethereum/statediff/trie_helpers"
|
||||
types2 "github.com/ethereum/go-ethereum/statediff/types"
|
||||
"github.com/ethereum/go-ethereum/trie"
|
||||
)
|
||||
|
||||
var (
|
||||
emptyNode, _ = rlp.EncodeToBytes(&[]byte{})
|
||||
emptyContractRoot = crypto.Keccak256Hash(emptyNode)
|
||||
nullCodeHash = crypto.Keccak256Hash([]byte{}).Bytes()
|
||||
nullNodeHash = common.Hex2Bytes("0000000000000000000000000000000000000000000000000000000000000000")
|
||||
)
|
||||
|
||||
// Builder interface exposes the method for building a state diff between two blocks
|
||||
type Builder interface {
|
||||
BuildStateDiffObject(args Args, params Params) (types2.StateObject, error)
|
||||
WriteStateDiffObject(args Args, params Params, output types2.StateNodeSink, ipldOutput types2.IPLDSink) error
|
||||
}
|
||||
|
||||
type StateDiffBuilder struct {
|
||||
StateCache state.Database
|
||||
}
|
||||
|
||||
type IterPair struct {
|
||||
Older, Newer trie.NodeIterator
|
||||
}
|
||||
|
||||
func StateNodeAppender(nodes *[]types2.StateLeafNode) types2.StateNodeSink {
|
||||
return func(node types2.StateLeafNode) error {
|
||||
*nodes = append(*nodes, node)
|
||||
return nil
|
||||
}
|
||||
}
|
||||
func StorageNodeAppender(nodes *[]types2.StorageLeafNode) types2.StorageNodeSink {
|
||||
return func(node types2.StorageLeafNode) error {
|
||||
*nodes = append(*nodes, node)
|
||||
return nil
|
||||
}
|
||||
}
|
||||
func IPLDMappingAppender(iplds *[]types2.IPLD) types2.IPLDSink {
|
||||
return func(c types2.IPLD) error {
|
||||
*iplds = append(*iplds, c)
|
||||
return nil
|
||||
}
|
||||
}
|
||||
|
||||
// NewBuilder is used to create a statediff builder
|
||||
func NewBuilder(stateCache state.Database) Builder {
|
||||
return &StateDiffBuilder{
|
||||
StateCache: stateCache, // state cache is safe for concurrent reads
|
||||
}
|
||||
}
|
||||
|
||||
// BuildStateDiffObject builds a statediff object from two blocks and the provided parameters
|
||||
func (sdb *StateDiffBuilder) BuildStateDiffObject(args Args, params Params) (types2.StateObject, error) {
|
||||
defer metrics2.UpdateDuration(time.Now(), metrics2.IndexerMetrics.BuildStateDiffObjectTimer)
|
||||
var stateNodes []types2.StateLeafNode
|
||||
var iplds []types2.IPLD
|
||||
err := sdb.WriteStateDiffObject(args, params, StateNodeAppender(&stateNodes), IPLDMappingAppender(&iplds))
|
||||
if err != nil {
|
||||
return types2.StateObject{}, err
|
||||
}
|
||||
return types2.StateObject{
|
||||
BlockHash: args.BlockHash,
|
||||
BlockNumber: args.BlockNumber,
|
||||
Nodes: stateNodes,
|
||||
IPLDs: iplds,
|
||||
}, nil
|
||||
}
|
||||
|
||||
// WriteStateDiffObject writes a statediff object to output sinks
|
||||
func (sdb *StateDiffBuilder) WriteStateDiffObject(args Args, params Params, output types2.StateNodeSink,
|
||||
ipldOutput types2.IPLDSink) error {
|
||||
defer metrics2.UpdateDuration(time.Now(), metrics2.IndexerMetrics.WriteStateDiffObjectTimer)
|
||||
// Load tries for old and new states
|
||||
oldTrie, err := sdb.StateCache.OpenTrie(args.OldStateRoot)
|
||||
if err != nil {
|
||||
return fmt.Errorf("error creating trie for oldStateRoot: %v", err)
|
||||
}
|
||||
newTrie, err := sdb.StateCache.OpenTrie(args.NewStateRoot)
|
||||
if err != nil {
|
||||
return fmt.Errorf("error creating trie for newStateRoot: %v", err)
|
||||
}
|
||||
|
||||
// we do two state trie iterations:
|
||||
// one for new/updated nodes,
|
||||
// one for deleted/updated nodes;
|
||||
// prepare 2 iterator instances for each task
|
||||
iterPairs := []IterPair{
|
||||
{
|
||||
Older: oldTrie.NodeIterator([]byte{}),
|
||||
Newer: newTrie.NodeIterator([]byte{}),
|
||||
},
|
||||
{
|
||||
Older: oldTrie.NodeIterator([]byte{}),
|
||||
Newer: newTrie.NodeIterator([]byte{}),
|
||||
},
|
||||
}
|
||||
|
||||
logger := log.New("hash", args.BlockHash.Hex(), "number", args.BlockNumber)
|
||||
return sdb.BuildStateDiffWithIntermediateStateNodes(iterPairs, params, output, ipldOutput, logger, nil)
|
||||
}
|
||||
|
||||
func (sdb *StateDiffBuilder) BuildStateDiffWithIntermediateStateNodes(iterPairs []IterPair, params Params,
|
||||
output types2.StateNodeSink, ipldOutput types2.IPLDSink, logger log.Logger, prefixPath []byte) error {
|
||||
logger.Debug("statediff BEGIN BuildStateDiffWithIntermediateStateNodes")
|
||||
defer metrics2.ReportAndUpdateDuration("statediff END BuildStateDiffWithIntermediateStateNodes", time.Now(), logger, metrics2.IndexerMetrics.BuildStateDiffWithIntermediateStateNodesTimer)
|
||||
// collect a slice of all the nodes that were touched and exist at B (B-A)
|
||||
// a map of their leafkey to all the accounts that were touched and exist at B
|
||||
// and a slice of all the paths for the nodes in both of the above sets
|
||||
diffAccountsAtB, err := sdb.createdAndUpdatedState(
|
||||
iterPairs[0].Older, iterPairs[0].Newer, params.watchedAddressesLeafPaths, ipldOutput, logger, prefixPath)
|
||||
if err != nil {
|
||||
return fmt.Errorf("error collecting createdAndUpdatedNodes: %v", err)
|
||||
}
|
||||
|
||||
// collect a slice of all the nodes that existed at a path in A that doesn't exist in B
|
||||
// a map of their leafkey to all the accounts that were touched and exist at A
|
||||
diffAccountsAtA, err := sdb.deletedOrUpdatedState(
|
||||
iterPairs[1].Older, iterPairs[1].Newer, diffAccountsAtB,
|
||||
params.watchedAddressesLeafPaths, output, logger, prefixPath)
|
||||
if err != nil {
|
||||
return fmt.Errorf("error collecting deletedOrUpdatedNodes: %v", err)
|
||||
}
|
||||
|
||||
// collect and sort the leafkey keys for both account mappings into a slice
|
||||
t := time.Now()
|
||||
createKeys := trie_helpers.SortKeys(diffAccountsAtB)
|
||||
deleteKeys := trie_helpers.SortKeys(diffAccountsAtA)
|
||||
logger.Debug(fmt.Sprintf("statediff BuildStateDiffWithIntermediateStateNodes sort duration=%dms", time.Since(t).Milliseconds()))
|
||||
|
||||
// and then find the intersection of these keys
|
||||
// these are the leafkeys for the accounts which exist at both A and B but are different
|
||||
// this also mutates the passed in createKeys and deleteKeys, removing the intersection keys
|
||||
// and leaving the truly created or deleted keys in place
|
||||
t = time.Now()
|
||||
updatedKeys := trie_helpers.FindIntersection(createKeys, deleteKeys)
|
||||
logger.Debug(fmt.Sprintf("statediff BuildStateDiffWithIntermediateStateNodes intersection count=%d duration=%dms",
|
||||
len(updatedKeys),
|
||||
time.Since(t).Milliseconds()))
|
||||
|
||||
// build the diff nodes for the updated accounts using the mappings at both A and B as directed by the keys found as the intersection of the two
|
||||
err = sdb.buildAccountUpdates(diffAccountsAtB, diffAccountsAtA, updatedKeys, output, ipldOutput, logger)
|
||||
if err != nil {
|
||||
return fmt.Errorf("error building diff for updated accounts: %v", err)
|
||||
}
|
||||
// build the diff nodes for created accounts
|
||||
err = sdb.buildAccountCreations(diffAccountsAtB, output, ipldOutput, logger)
|
||||
if err != nil {
|
||||
return fmt.Errorf("error building diff for created accounts: %v", err)
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
// createdAndUpdatedState returns
|
||||
// a slice of all the intermediate nodes that exist in a different state at B than A
|
||||
// a mapping of their leafkeys to all the accounts that exist in a different state at B than A
|
||||
// and a slice of the paths for all of the nodes included in both
|
||||
func (sdb *StateDiffBuilder) createdAndUpdatedState(a, b trie.NodeIterator,
|
||||
watchedAddressesLeafPaths [][]byte, output types2.IPLDSink, logger log.Logger, prefixPath []byte) (types2.AccountMap, error) {
|
||||
logger.Debug("statediff BEGIN createdAndUpdatedState")
|
||||
defer metrics2.ReportAndUpdateDuration("statediff END createdAndUpdatedState", time.Now(), logger, metrics2.IndexerMetrics.CreatedAndUpdatedStateTimer)
|
||||
diffAccountsAtB := make(types2.AccountMap)
|
||||
watchingAddresses := len(watchedAddressesLeafPaths) > 0
|
||||
|
||||
it, itCount := trie.NewDifferenceIterator(a, b)
|
||||
for it.Next(true) {
|
||||
// ignore node if it is not along paths of interest
|
||||
if watchingAddresses && !isValidPrefixPath(watchedAddressesLeafPaths, append(prefixPath, it.Path()...)) {
|
||||
continue
|
||||
}
|
||||
// index values by leaf key
|
||||
if it.Leaf() {
|
||||
// if it is a "value" node, we will index the value by leaf key
|
||||
accountW, err := sdb.processStateValueNode(it, watchedAddressesLeafPaths, prefixPath)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
if accountW == nil {
|
||||
continue
|
||||
}
|
||||
// for now, just add it to diffAccountsAtB
|
||||
// we will compare to diffAccountsAtA to determine which diffAccountsAtB
|
||||
// were creations and which were updates and also identify accounts that were removed going A->B
|
||||
diffAccountsAtB[common.Bytes2Hex(accountW.LeafKey)] = *accountW
|
||||
} else { // trie nodes will be written to blockstore only
|
||||
// reminder that this includes leaf nodes, since the geth iterator.Leaf() actually signifies a "value" node
|
||||
if bytes.Equal(it.Hash().Bytes(), nullNodeHash) {
|
||||
continue
|
||||
}
|
||||
nodeVal := make([]byte, len(it.NodeBlob()))
|
||||
copy(nodeVal, it.NodeBlob())
|
||||
if len(watchedAddressesLeafPaths) > 0 {
|
||||
var elements []interface{}
|
||||
if err := rlp.DecodeBytes(nodeVal, &elements); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
ok, err := isLeaf(elements)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
if ok {
|
||||
nodePath := append(prefixPath, it.Path()...)
|
||||
partialPath := trie.CompactToHex(elements[0].([]byte))
|
||||
valueNodePath := append(nodePath, partialPath...)
|
||||
if !isWatchedAddress(watchedAddressesLeafPaths, valueNodePath) {
|
||||
continue
|
||||
}
|
||||
}
|
||||
}
|
||||
nodeHash := make([]byte, len(it.Hash().Bytes()))
|
||||
copy(nodeHash, it.Hash().Bytes())
|
||||
if err := output(types2.IPLD{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, nodeHash).String(),
|
||||
Content: nodeVal,
|
||||
}); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
}
|
||||
}
|
||||
logger.Debug("statediff COUNTS createdAndUpdatedStateWithIntermediateNodes", "it", itCount, "diffAccountsAtB", len(diffAccountsAtB))
|
||||
metrics2.IndexerMetrics.DifferenceIteratorCounter.Inc(int64(*itCount))
|
||||
return diffAccountsAtB, it.Error()
|
||||
}
|
||||
|
||||
// reminder: it.Leaf() == true when the iterator is positioned at a "value node" which is not something that actually exists in an MMPT
|
||||
func (sdb *StateDiffBuilder) processStateValueNode(it trie.NodeIterator, watchedAddressesLeafPaths [][]byte, prefixPath []byte) (*types2.AccountWrapper, error) {
|
||||
// skip if it is not a watched address
|
||||
// If we aren't watching any specific addresses, we are watching everything
|
||||
if len(watchedAddressesLeafPaths) > 0 && !isWatchedAddress(watchedAddressesLeafPaths, append(prefixPath, it.Path()...)) {
|
||||
return nil, nil
|
||||
}
|
||||
|
||||
// since this is a "value node", we need to move up to the "parent" node which is the actual leaf node
|
||||
// it should be in the fastcache since it necessarily was recently accessed to reach the current node
|
||||
parentNodeRLP, err := sdb.StateCache.TrieDB().Node(it.Parent())
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
var nodeElements []interface{}
|
||||
if err = rlp.DecodeBytes(parentNodeRLP, &nodeElements); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
parentSubPath := make([]byte, len(it.ParentPath()))
|
||||
copy(parentSubPath, it.ParentPath())
|
||||
parentPath := append(prefixPath, parentSubPath...)
|
||||
partialPath := trie.CompactToHex(nodeElements[0].([]byte))
|
||||
valueNodePath := append(parentPath, partialPath...)
|
||||
encodedPath := trie.HexToCompact(valueNodePath)
|
||||
leafKey := encodedPath[1:]
|
||||
|
||||
var account types.StateAccount
|
||||
accountRLP := make([]byte, len(it.LeafBlob()))
|
||||
copy(accountRLP, it.LeafBlob())
|
||||
if err := rlp.DecodeBytes(accountRLP, &account); err != nil {
|
||||
return nil, fmt.Errorf("error decoding account for leaf value at leaf key %x\nerror: %v", leafKey, err)
|
||||
}
|
||||
|
||||
return &types2.AccountWrapper{
|
||||
LeafKey: leafKey,
|
||||
Account: &account,
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(parentNodeRLP)).String(),
|
||||
}, nil
|
||||
}
|
||||
|
||||
// deletedOrUpdatedState returns a slice of all the pathes that are emptied at B
|
||||
// and a mapping of their leafkeys to all the accounts that exist in a different state at A than B
|
||||
func (sdb *StateDiffBuilder) deletedOrUpdatedState(a, b trie.NodeIterator, diffAccountsAtB types2.AccountMap,
|
||||
watchedAddressesLeafPaths [][]byte, output types2.StateNodeSink, logger log.Logger, prefixPath []byte) (types2.AccountMap, error) {
|
||||
logger.Debug("statediff BEGIN deletedOrUpdatedState")
|
||||
defer metrics2.ReportAndUpdateDuration("statediff END deletedOrUpdatedState", time.Now(), logger, metrics2.IndexerMetrics.DeletedOrUpdatedStateTimer)
|
||||
diffAccountAtA := make(types2.AccountMap)
|
||||
watchingAddresses := len(watchedAddressesLeafPaths) > 0
|
||||
|
||||
it, _ := trie.NewDifferenceIterator(b, a)
|
||||
for it.Next(true) {
|
||||
// ignore node if it is not along paths of interest
|
||||
if watchingAddresses && !isValidPrefixPath(watchedAddressesLeafPaths, append(prefixPath, it.Path()...)) {
|
||||
continue
|
||||
}
|
||||
|
||||
if it.Leaf() {
|
||||
accountW, err := sdb.processStateValueNode(it, watchedAddressesLeafPaths, prefixPath)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
if accountW == nil {
|
||||
continue
|
||||
}
|
||||
leafKey := common.Bytes2Hex(accountW.LeafKey)
|
||||
diffAccountAtA[leafKey] = *accountW
|
||||
// if this node's leaf key did not show up in diffAccountsAtB
|
||||
// that means the account was deleted
|
||||
// in that case, emit an empty "removed" diff state node
|
||||
// include empty "removed" diff storage nodes for all the storage slots
|
||||
if _, ok := diffAccountsAtB[leafKey]; !ok {
|
||||
diff := types2.StateLeafNode{
|
||||
AccountWrapper: types2.AccountWrapper{
|
||||
Account: nil,
|
||||
LeafKey: accountW.LeafKey,
|
||||
CID: shared.RemovedNodeStateCID,
|
||||
},
|
||||
Removed: true,
|
||||
}
|
||||
|
||||
storageDiff := make([]types2.StorageLeafNode, 0)
|
||||
err := sdb.buildRemovedAccountStorageNodes(accountW.Account.Root, StorageNodeAppender(&storageDiff))
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("failed building storage diffs for removed state account with key %x\r\nerror: %v", leafKey, err)
|
||||
}
|
||||
diff.StorageDiff = storageDiff
|
||||
if err := output(diff); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
return diffAccountAtA, it.Error()
|
||||
}
|
||||
|
||||
// buildAccountUpdates uses the account diffs maps for A => B and B => A and the known intersection of their leafkeys
|
||||
// to generate the statediff node objects for all of the accounts that existed at both A and B but in different states
|
||||
// needs to be called before building account creations and deletions as this mutates
|
||||
// those account maps to remove the accounts which were updated
|
||||
func (sdb *StateDiffBuilder) buildAccountUpdates(creations, deletions types2.AccountMap, updatedKeys []string,
|
||||
output types2.StateNodeSink, ipldOutput types2.IPLDSink, logger log.Logger) error {
|
||||
logger.Debug("statediff BEGIN buildAccountUpdates", "creations", len(creations), "deletions", len(deletions), "updatedKeys", len(updatedKeys))
|
||||
defer metrics2.ReportAndUpdateDuration("statediff END buildAccountUpdates ", time.Now(), logger, metrics2.IndexerMetrics.BuildAccountUpdatesTimer)
|
||||
var err error
|
||||
for _, key := range updatedKeys {
|
||||
createdAcc := creations[key]
|
||||
deletedAcc := deletions[key]
|
||||
storageDiff := make([]types2.StorageLeafNode, 0)
|
||||
if deletedAcc.Account != nil && createdAcc.Account != nil {
|
||||
oldSR := deletedAcc.Account.Root
|
||||
newSR := createdAcc.Account.Root
|
||||
err = sdb.buildStorageNodesIncremental(
|
||||
oldSR, newSR, StorageNodeAppender(&storageDiff), ipldOutput)
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed building incremental storage diffs for account with leafkey %s\r\nerror: %v", key, err)
|
||||
}
|
||||
}
|
||||
if err = output(types2.StateLeafNode{
|
||||
AccountWrapper: createdAcc,
|
||||
Removed: false,
|
||||
StorageDiff: storageDiff,
|
||||
}); err != nil {
|
||||
return err
|
||||
}
|
||||
delete(creations, key)
|
||||
delete(deletions, key)
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// buildAccountCreations returns the statediff node objects for all the accounts that exist at B but not at A
|
||||
// it also returns the code and codehash for created contract accounts
|
||||
func (sdb *StateDiffBuilder) buildAccountCreations(accounts types2.AccountMap, output types2.StateNodeSink,
|
||||
ipldOutput types2.IPLDSink, logger log.Logger) error {
|
||||
logger.Debug("statediff BEGIN buildAccountCreations")
|
||||
defer metrics2.ReportAndUpdateDuration("statediff END buildAccountCreations", time.Now(), logger, metrics2.IndexerMetrics.BuildAccountCreationsTimer)
|
||||
for _, val := range accounts {
|
||||
diff := types2.StateLeafNode{
|
||||
AccountWrapper: val,
|
||||
Removed: false,
|
||||
}
|
||||
if !bytes.Equal(val.Account.CodeHash, nullCodeHash) {
|
||||
// For contract creations, any storage node contained is a diff
|
||||
storageDiff := make([]types2.StorageLeafNode, 0)
|
||||
err := sdb.buildStorageNodesEventual(val.Account.Root, StorageNodeAppender(&storageDiff), ipldOutput)
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed building eventual storage diffs for node with leaf key %x\r\nerror: %v", val.LeafKey, err)
|
||||
}
|
||||
diff.StorageDiff = storageDiff
|
||||
// emit codehash => code mappings for contract
|
||||
codeHash := common.BytesToHash(val.Account.CodeHash)
|
||||
code, err := sdb.StateCache.ContractCode(common.Hash{}, codeHash)
|
||||
if err != nil {
|
||||
return fmt.Errorf("failed to retrieve code for codehash %s\r\n error: %v", codeHash.String(), err)
|
||||
}
|
||||
if err := ipldOutput(types2.IPLD{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.RawBinary, codeHash.Bytes()).String(),
|
||||
Content: code,
|
||||
}); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
if err := output(diff); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// buildStorageNodesEventual builds the storage diff node objects for a created account
|
||||
// i.e. it returns all the storage nodes at this state, since there is no previous state
|
||||
func (sdb *StateDiffBuilder) buildStorageNodesEventual(sr common.Hash, output types2.StorageNodeSink,
|
||||
ipldOutput types2.IPLDSink) error {
|
||||
defer metrics2.UpdateDuration(time.Now(), metrics2.IndexerMetrics.BuildStorageNodesEventualTimer)
|
||||
if bytes.Equal(sr.Bytes(), emptyContractRoot.Bytes()) {
|
||||
return nil
|
||||
}
|
||||
log.Debug("Storage Root For Eventual Diff", "root", sr.Hex())
|
||||
sTrie, err := sdb.StateCache.OpenTrie(sr)
|
||||
if err != nil {
|
||||
log.Info("error in build storage diff eventual", "error", err)
|
||||
return err
|
||||
}
|
||||
it := sTrie.NodeIterator(make([]byte, 0))
|
||||
err = sdb.buildStorageNodesFromTrie(it, output, ipldOutput)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
// buildStorageNodesFromTrie returns all the storage diff node objects in the provided node interator
|
||||
// including intermediate nodes can be turned on or off
|
||||
func (sdb *StateDiffBuilder) buildStorageNodesFromTrie(it trie.NodeIterator, output types2.StorageNodeSink,
|
||||
ipldOutput types2.IPLDSink) error {
|
||||
defer metrics2.UpdateDuration(time.Now(), metrics2.IndexerMetrics.BuildStorageNodesFromTrieTimer)
|
||||
for it.Next(true) {
|
||||
if it.Leaf() {
|
||||
storageLeafNode, err := sdb.processStorageValueNode(it)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
if err := output(storageLeafNode); err != nil {
|
||||
return err
|
||||
}
|
||||
} else {
|
||||
nodeVal := make([]byte, len(it.NodeBlob()))
|
||||
copy(nodeVal, it.NodeBlob())
|
||||
nodeHash := make([]byte, len(it.Hash().Bytes()))
|
||||
copy(nodeHash, it.Hash().Bytes())
|
||||
if err := ipldOutput(types2.IPLD{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStorageTrie, nodeHash).String(),
|
||||
Content: nodeVal,
|
||||
}); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
}
|
||||
return it.Error()
|
||||
}
|
||||
|
||||
// reminder: it.Leaf() == true when the iterator is positioned at a "value node" which is not something that actually exists in an MMPT
|
||||
func (sdb *StateDiffBuilder) processStorageValueNode(it trie.NodeIterator) (types2.StorageLeafNode, error) {
|
||||
// skip if it is not a watched address
|
||||
leafKey := make([]byte, len(it.LeafKey()))
|
||||
copy(leafKey, it.LeafKey())
|
||||
value := make([]byte, len(it.LeafBlob()))
|
||||
copy(value, it.LeafBlob())
|
||||
|
||||
// since this is a "value node", we need to move up to the "parent" node which is the actual leaf node
|
||||
// it should be in the fastcache since it necessarily was recently accessed to reach the current node
|
||||
parentNodeRLP, err := sdb.StateCache.TrieDB().Node(it.Parent())
|
||||
if err != nil {
|
||||
return types2.StorageLeafNode{}, err
|
||||
}
|
||||
|
||||
return types2.StorageLeafNode{
|
||||
LeafKey: leafKey,
|
||||
Value: value,
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStorageTrie, crypto.Keccak256(parentNodeRLP)).String(),
|
||||
}, nil
|
||||
}
|
||||
|
||||
// buildRemovedAccountStorageNodes builds the "removed" diffs for all the storage nodes for a destroyed account
|
||||
func (sdb *StateDiffBuilder) buildRemovedAccountStorageNodes(sr common.Hash, output types2.StorageNodeSink) error {
|
||||
defer metrics2.UpdateDuration(time.Now(), metrics2.IndexerMetrics.BuildRemovedAccountStorageNodesTimer)
|
||||
if bytes.Equal(sr.Bytes(), emptyContractRoot.Bytes()) {
|
||||
return nil
|
||||
}
|
||||
log.Debug("Storage Root For Removed Diffs", "root", sr.Hex())
|
||||
sTrie, err := sdb.StateCache.OpenTrie(sr)
|
||||
if err != nil {
|
||||
log.Info("error in build removed account storage diffs", "error", err)
|
||||
return err
|
||||
}
|
||||
it := sTrie.NodeIterator(make([]byte, 0))
|
||||
err = sdb.buildRemovedStorageNodesFromTrie(it, output)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
// buildRemovedStorageNodesFromTrie returns diffs for all the storage nodes in the provided node interator
|
||||
func (sdb *StateDiffBuilder) buildRemovedStorageNodesFromTrie(it trie.NodeIterator, output types2.StorageNodeSink) error {
|
||||
defer metrics2.UpdateDuration(time.Now(), metrics2.IndexerMetrics.BuildRemovedStorageNodesFromTrieTimer)
|
||||
for it.Next(true) {
|
||||
if it.Leaf() { // only leaf values are indexed, don't need to demarcate removed intermediate nodes
|
||||
leafKey := make([]byte, len(it.LeafKey()))
|
||||
copy(leafKey, it.LeafKey())
|
||||
if err := output(types2.StorageLeafNode{
|
||||
CID: shared.RemovedNodeStorageCID,
|
||||
Removed: true,
|
||||
LeafKey: leafKey,
|
||||
Value: []byte{},
|
||||
}); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
}
|
||||
return it.Error()
|
||||
}
|
||||
|
||||
// buildStorageNodesIncremental builds the storage diff node objects for all nodes that exist in a different state at B than A
|
||||
func (sdb *StateDiffBuilder) buildStorageNodesIncremental(oldSR common.Hash, newSR common.Hash, output types2.StorageNodeSink,
|
||||
ipldOutput types2.IPLDSink) error {
|
||||
defer metrics2.UpdateDuration(time.Now(), metrics2.IndexerMetrics.BuildStorageNodesIncrementalTimer)
|
||||
if bytes.Equal(newSR.Bytes(), oldSR.Bytes()) {
|
||||
return nil
|
||||
}
|
||||
log.Trace("Storage Roots for Incremental Diff", "old", oldSR.Hex(), "new", newSR.Hex())
|
||||
oldTrie, err := sdb.StateCache.OpenTrie(oldSR)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
newTrie, err := sdb.StateCache.OpenTrie(newSR)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
diffSlotsAtB, err := sdb.createdAndUpdatedStorage(
|
||||
oldTrie.NodeIterator([]byte{}), newTrie.NodeIterator([]byte{}), output, ipldOutput)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
err = sdb.deletedOrUpdatedStorage(oldTrie.NodeIterator([]byte{}), newTrie.NodeIterator([]byte{}),
|
||||
diffSlotsAtB, output)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
func (sdb *StateDiffBuilder) createdAndUpdatedStorage(a, b trie.NodeIterator, output types2.StorageNodeSink,
|
||||
ipldOutput types2.IPLDSink) (map[string]bool, error) {
|
||||
defer metrics2.UpdateDuration(time.Now(), metrics2.IndexerMetrics.CreatedAndUpdatedStorageTimer)
|
||||
diffSlotsAtB := make(map[string]bool)
|
||||
it, _ := trie.NewDifferenceIterator(a, b)
|
||||
for it.Next(true) {
|
||||
if it.Leaf() {
|
||||
storageLeafNode, err := sdb.processStorageValueNode(it)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
if err := output(storageLeafNode); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
diffSlotsAtB[common.Bytes2Hex(storageLeafNode.LeafKey)] = true
|
||||
} else {
|
||||
if bytes.Equal(it.Hash().Bytes(), nullNodeHash) {
|
||||
continue
|
||||
}
|
||||
nodeVal := make([]byte, len(it.NodeBlob()))
|
||||
copy(nodeVal, it.NodeBlob())
|
||||
nodeHash := make([]byte, len(it.Hash().Bytes()))
|
||||
copy(nodeHash, it.Hash().Bytes())
|
||||
if err := ipldOutput(types2.IPLD{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStorageTrie, nodeHash).String(),
|
||||
Content: nodeVal,
|
||||
}); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
}
|
||||
}
|
||||
return diffSlotsAtB, it.Error()
|
||||
}
|
||||
|
||||
func (sdb *StateDiffBuilder) deletedOrUpdatedStorage(a, b trie.NodeIterator, diffSlotsAtB map[string]bool, output types2.StorageNodeSink) error {
|
||||
defer metrics2.UpdateDuration(time.Now(), metrics2.IndexerMetrics.DeletedOrUpdatedStorageTimer)
|
||||
it, _ := trie.NewDifferenceIterator(b, a)
|
||||
for it.Next(true) {
|
||||
if it.Leaf() {
|
||||
leafKey := make([]byte, len(it.LeafKey()))
|
||||
copy(leafKey, it.LeafKey())
|
||||
// if this node's leaf key did not show up in diffSlotsAtB
|
||||
// that means the storage slot was vacated
|
||||
// in that case, emit an empty "removed" diff storage node
|
||||
if _, ok := diffSlotsAtB[common.Bytes2Hex(leafKey)]; !ok {
|
||||
if err := output(types2.StorageLeafNode{
|
||||
CID: shared.RemovedNodeStorageCID,
|
||||
Removed: true,
|
||||
LeafKey: leafKey,
|
||||
Value: []byte{},
|
||||
}); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
return it.Error()
|
||||
}
|
||||
|
||||
// isValidPrefixPath is used to check if a node at currentPath is a parent | ancestor to one of the addresses the builder is configured to watch
|
||||
func isValidPrefixPath(watchedAddressesLeafPaths [][]byte, currentPath []byte) bool {
|
||||
for _, watchedAddressPath := range watchedAddressesLeafPaths {
|
||||
if bytes.HasPrefix(watchedAddressPath, currentPath) {
|
||||
return true
|
||||
}
|
||||
}
|
||||
|
||||
return false
|
||||
}
|
||||
|
||||
// isWatchedAddress is used to check if a state account corresponds to one of the addresses the builder is configured to watch
|
||||
func isWatchedAddress(watchedAddressesLeafPaths [][]byte, valueNodePath []byte) bool {
|
||||
defer metrics2.UpdateDuration(time.Now(), metrics2.IndexerMetrics.IsWatchedAddressTimer)
|
||||
for _, watchedAddressPath := range watchedAddressesLeafPaths {
|
||||
if bytes.Equal(watchedAddressPath, valueNodePath) {
|
||||
return true
|
||||
}
|
||||
}
|
||||
|
||||
return false
|
||||
}
|
||||
|
||||
// isLeaf checks if the node we are at is a leaf
|
||||
func isLeaf(elements []interface{}) (bool, error) {
|
||||
if len(elements) > 2 {
|
||||
return false, nil
|
||||
}
|
||||
if len(elements) < 2 {
|
||||
return false, fmt.Errorf("node cannot be less than two elements in length")
|
||||
}
|
||||
switch elements[0].([]byte)[0] / 16 {
|
||||
case '\x00':
|
||||
return false, nil
|
||||
case '\x01':
|
||||
return false, nil
|
||||
case '\x02':
|
||||
return true, nil
|
||||
case '\x03':
|
||||
return true, nil
|
||||
default:
|
||||
return false, fmt.Errorf("unknown hex prefix")
|
||||
}
|
||||
}
|
3108
builder_test.go
Normal file
3108
builder_test.go
Normal file
File diff suppressed because it is too large
Load Diff
87
config.go
Normal file
87
config.go
Normal file
@ -0,0 +1,87 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package statediff
|
||||
|
||||
import (
|
||||
"context"
|
||||
"math/big"
|
||||
"sync"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/crypto"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/interfaces"
|
||||
)
|
||||
|
||||
// Config contains instantiation parameters for the state diffing service
|
||||
type Config struct {
|
||||
// The configuration used for the stateDiff Indexer
|
||||
IndexerConfig interfaces.Config
|
||||
// A unique ID used for this service
|
||||
ID string
|
||||
// Name for the client this service is running
|
||||
ClientName string
|
||||
// Whether to enable writing state diffs directly to track blockchain head
|
||||
EnableWriteLoop bool
|
||||
// Size of the worker pool
|
||||
NumWorkers uint
|
||||
// Should the statediff service wait until geth has synced to the head of the blockchain?
|
||||
WaitForSync bool
|
||||
// Context
|
||||
Context context.Context
|
||||
}
|
||||
|
||||
// Params contains config parameters for the state diff builder
|
||||
type Params struct {
|
||||
IncludeBlock bool
|
||||
IncludeReceipts bool
|
||||
IncludeTD bool
|
||||
IncludeCode bool
|
||||
WatchedAddresses []common.Address
|
||||
watchedAddressesLeafPaths [][]byte
|
||||
}
|
||||
|
||||
// ComputeWatchedAddressesLeafPaths populates a slice with paths (hex_encoding(Keccak256)) of each of the WatchedAddresses
|
||||
func (p *Params) ComputeWatchedAddressesLeafPaths() {
|
||||
p.watchedAddressesLeafPaths = make([][]byte, len(p.WatchedAddresses))
|
||||
for i, address := range p.WatchedAddresses {
|
||||
p.watchedAddressesLeafPaths[i] = keybytesToHex(crypto.Keccak256(address.Bytes()))
|
||||
}
|
||||
}
|
||||
|
||||
// ParamsWithMutex allows to lock the parameters while they are being updated | read from
|
||||
type ParamsWithMutex struct {
|
||||
Params
|
||||
sync.RWMutex
|
||||
}
|
||||
|
||||
// Args bundles the arguments for the state diff builder
|
||||
type Args struct {
|
||||
OldStateRoot, NewStateRoot, BlockHash common.Hash
|
||||
BlockNumber *big.Int
|
||||
}
|
||||
|
||||
// https://github.com/ethereum/go-ethereum/blob/master/trie/encoding.go#L97
|
||||
func keybytesToHex(str []byte) []byte {
|
||||
l := len(str)*2 + 1
|
||||
var nibbles = make([]byte, l)
|
||||
for i, b := range str {
|
||||
nibbles[i*2] = b / 16
|
||||
nibbles[i*2+1] = b % 16
|
||||
}
|
||||
nibbles[l-1] = 16
|
||||
return nibbles
|
||||
}
|
17
docs/KnownGaps.md
Normal file
17
docs/KnownGaps.md
Normal file
@ -0,0 +1,17 @@
|
||||
# Overview
|
||||
|
||||
This document will provide some insight into the `known_gaps` table, their use cases, and implementation. Please refer to the [following PR](https://github.com/vulcanize/go-ethereum/pull/217) and the [following epic](https://github.com/vulcanize/ops/issues/143) to grasp their inception.
|
||||
|
||||
![known gaps](diagrams/KnownGapsProcess.png)
|
||||
|
||||
# Use Cases
|
||||
|
||||
The known gaps table is updated when the following events occur:
|
||||
|
||||
1. At start up we check the latest block from the `eth.headers_cid` table. We compare the first block that we are processing with the latest block from the DB. If they are not one unit of expectedDifference away from each other, add the gap between the two blocks.
|
||||
2. If there is any error in processing a block (db connection, deadlock, etc), add that block to the knownErrorBlocks slice, when the next block is successfully written, write this slice into the DB.
|
||||
|
||||
# Glossary
|
||||
|
||||
1. `expectedDifference (number)` - This number indicates what the difference between two blocks should be. If we are capturing all events on a geth node then this number would be `1`. But once we scale nodes, the `expectedDifference` might be `2` or greater.
|
||||
2. `processingKey (number)` - This number can be used to keep track of different geth nodes and their specific `expectedDifference`.
|
3
docs/README.md
Normal file
3
docs/README.md
Normal file
@ -0,0 +1,3 @@
|
||||
# Overview
|
||||
|
||||
This folder keeps tracks of random documents as they relate to the `statediff` service.
|
21
docs/database.md
Normal file
21
docs/database.md
Normal file
@ -0,0 +1,21 @@
|
||||
# Overview
|
||||
|
||||
This document will go through some notes on the database component of the statediff service.
|
||||
|
||||
# Components
|
||||
|
||||
- Indexer: The indexer creates IPLD and DB models to insert to the Postgres DB. It performs the insert utilizing and atomic function.
|
||||
- Builder: The builder constructs the statediff object that needs to be inserted.
|
||||
- Known Gaps: Captures any gaps that might have occured and either writes them to the DB, local sql file, to prometeus, or a local error.
|
||||
|
||||
# Making Code Changes
|
||||
|
||||
## Adding a New Function to the Indexer
|
||||
|
||||
If you want to implement a new feature for adding data to the database. Keep the following in mind:
|
||||
|
||||
1. You need to handle `sql`, `file`, and `dump`.
|
||||
1. `sql` - Contains the code needed to write directly to the `sql` db.
|
||||
2. `file` - Contains all the code required to write the SQL statements to a file.
|
||||
3. `dump` - Contains all the code for outputting events to the console.
|
||||
2. You will have to add it to the `interfaces.StateDiffIndexer` interface.
|
BIN
docs/diagrams/KnownGapsProcess.png
Normal file
BIN
docs/diagrams/KnownGapsProcess.png
Normal file
Binary file not shown.
After Width: | Height: | Size: 33 KiB |
81
indexer/constructor.go
Normal file
81
indexer/constructor.go
Normal file
@ -0,0 +1,81 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package indexer
|
||||
|
||||
import (
|
||||
"context"
|
||||
"fmt"
|
||||
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
"github.com/ethereum/go-ethereum/params"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/dump"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/file"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql/postgres"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/interfaces"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/node"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/shared"
|
||||
)
|
||||
|
||||
// NewStateDiffIndexer creates and returns an implementation of the StateDiffIndexer interface.
|
||||
func NewStateDiffIndexer(ctx context.Context, chainConfig *params.ChainConfig, nodeInfo node.Info, config interfaces.Config) (sql.Database, interfaces.StateDiffIndexer, error) {
|
||||
switch config.Type() {
|
||||
case shared.FILE:
|
||||
log.Info("Starting statediff service in SQL file writing mode")
|
||||
fc, ok := config.(file.Config)
|
||||
if !ok {
|
||||
return nil, nil, fmt.Errorf("file config is not the correct type: got %T, expected %T", config, file.Config{})
|
||||
}
|
||||
fc.NodeInfo = nodeInfo
|
||||
ind, err := file.NewStateDiffIndexer(ctx, chainConfig, fc)
|
||||
return nil, ind, err
|
||||
case shared.POSTGRES:
|
||||
log.Info("Starting statediff service in Postgres writing mode")
|
||||
pgc, ok := config.(postgres.Config)
|
||||
if !ok {
|
||||
return nil, nil, fmt.Errorf("postgres config is not the correct type: got %T, expected %T", config, postgres.Config{})
|
||||
}
|
||||
var err error
|
||||
var driver sql.Driver
|
||||
switch pgc.Driver {
|
||||
case postgres.PGX:
|
||||
driver, err = postgres.NewPGXDriver(ctx, pgc, nodeInfo)
|
||||
if err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
case postgres.SQLX:
|
||||
driver, err = postgres.NewSQLXDriver(ctx, pgc, nodeInfo)
|
||||
if err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
default:
|
||||
return nil, nil, fmt.Errorf("unrecognized Postgres driver type: %s", pgc.Driver)
|
||||
}
|
||||
db := postgres.NewPostgresDB(driver, pgc.Upsert)
|
||||
ind, err := sql.NewStateDiffIndexer(ctx, chainConfig, db)
|
||||
return db, ind, err
|
||||
case shared.DUMP:
|
||||
log.Info("Starting statediff service in data dump mode")
|
||||
dumpc, ok := config.(dump.Config)
|
||||
if !ok {
|
||||
return nil, nil, fmt.Errorf("dump config is not the correct type: got %T, expected %T", config, dump.Config{})
|
||||
}
|
||||
return nil, dump.NewStateDiffIndexer(chainConfig, dumpc), nil
|
||||
default:
|
||||
return nil, nil, fmt.Errorf("unrecognized database type: %s", config.Type())
|
||||
}
|
||||
}
|
80
indexer/database/dump/batch_tx.go
Normal file
80
indexer/database/dump/batch_tx.go
Normal file
@ -0,0 +1,80 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package dump
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"io"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/ipld"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/models"
|
||||
)
|
||||
|
||||
// BatchTx wraps a void with the state necessary for building the tx concurrently during trie difference iteration
|
||||
type BatchTx struct {
|
||||
BlockNumber string
|
||||
dump io.Writer
|
||||
quit chan struct{}
|
||||
iplds chan models.IPLDModel
|
||||
ipldCache models.IPLDBatch
|
||||
|
||||
submit func(blockTx *BatchTx, err error) error
|
||||
}
|
||||
|
||||
// Submit satisfies indexer.AtomicTx
|
||||
func (tx *BatchTx) Submit(err error) error {
|
||||
return tx.submit(tx, err)
|
||||
}
|
||||
|
||||
func (tx *BatchTx) flush() error {
|
||||
if _, err := fmt.Fprintf(tx.dump, "%+v\r\n", tx.ipldCache); err != nil {
|
||||
return err
|
||||
}
|
||||
tx.ipldCache = models.IPLDBatch{}
|
||||
return nil
|
||||
}
|
||||
|
||||
// run in background goroutine to synchronize concurrent appends to the ipldCache
|
||||
func (tx *BatchTx) cache() {
|
||||
for {
|
||||
select {
|
||||
case i := <-tx.iplds:
|
||||
tx.ipldCache.Keys = append(tx.ipldCache.Keys, i.Key)
|
||||
tx.ipldCache.Values = append(tx.ipldCache.Values, i.Data)
|
||||
case <-tx.quit:
|
||||
tx.ipldCache = models.IPLDBatch{}
|
||||
return
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (tx *BatchTx) cacheDirect(key string, value []byte) {
|
||||
tx.iplds <- models.IPLDModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
Key: key,
|
||||
Data: value,
|
||||
}
|
||||
}
|
||||
|
||||
func (tx *BatchTx) cacheIPLD(i ipld.IPLD) {
|
||||
tx.iplds <- models.IPLDModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
Key: i.Cid().String(),
|
||||
Data: i.RawData(),
|
||||
}
|
||||
}
|
79
indexer/database/dump/config.go
Normal file
79
indexer/database/dump/config.go
Normal file
@ -0,0 +1,79 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package dump
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"io"
|
||||
"strings"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/shared"
|
||||
)
|
||||
|
||||
// DumpType to explicitly type the dump destination
|
||||
type DumpType string
|
||||
|
||||
const (
|
||||
STDOUT = "Stdout"
|
||||
STDERR = "Stderr"
|
||||
DISCARD = "Discard"
|
||||
UNKNOWN = "Unknown"
|
||||
)
|
||||
|
||||
// ResolveDumpType resolves the dump type for the provided string
|
||||
func ResolveDumpType(str string) (DumpType, error) {
|
||||
switch strings.ToLower(str) {
|
||||
case "stdout", "out", "std out":
|
||||
return STDOUT, nil
|
||||
case "stderr", "err", "std err":
|
||||
return STDERR, nil
|
||||
case "discard", "void", "devnull", "dev null":
|
||||
return DISCARD, nil
|
||||
default:
|
||||
return UNKNOWN, fmt.Errorf("unrecognized dump type: %s", str)
|
||||
}
|
||||
}
|
||||
|
||||
// Config for data dump
|
||||
type Config struct {
|
||||
Dump io.WriteCloser
|
||||
}
|
||||
|
||||
// Type satisfies interfaces.Config
|
||||
func (c Config) Type() shared.DBType {
|
||||
return shared.DUMP
|
||||
}
|
||||
|
||||
// NewDiscardWriterCloser returns a discardWrapper wrapping io.Discard
|
||||
func NewDiscardWriterCloser() io.WriteCloser {
|
||||
return discardWrapper{blackhole: io.Discard}
|
||||
}
|
||||
|
||||
// discardWrapper wraps io.Discard with io.Closer
|
||||
type discardWrapper struct {
|
||||
blackhole io.Writer
|
||||
}
|
||||
|
||||
// Write satisfies io.Writer
|
||||
func (dw discardWrapper) Write(b []byte) (int, error) {
|
||||
return dw.blackhole.Write(b)
|
||||
}
|
||||
|
||||
// Close satisfies io.Closer
|
||||
func (dw discardWrapper) Close() error {
|
||||
return nil
|
||||
}
|
443
indexer/database/dump/indexer.go
Normal file
443
indexer/database/dump/indexer.go
Normal file
@ -0,0 +1,443 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package dump
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"fmt"
|
||||
"io"
|
||||
"math/big"
|
||||
"time"
|
||||
|
||||
"github.com/multiformats/go-multihash"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/crypto"
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
"github.com/ethereum/go-ethereum/params"
|
||||
"github.com/ethereum/go-ethereum/rlp"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/metrics"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/interfaces"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/ipld"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/models"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/shared"
|
||||
sdtypes "github.com/ethereum/go-ethereum/statediff/types"
|
||||
)
|
||||
|
||||
var _ interfaces.StateDiffIndexer = &StateDiffIndexer{}
|
||||
|
||||
// StateDiffIndexer satisfies the indexer.StateDiffIndexer interface for ethereum statediff objects on top of a void
|
||||
type StateDiffIndexer struct {
|
||||
dump io.WriteCloser
|
||||
chainConfig *params.ChainConfig
|
||||
}
|
||||
|
||||
// NewStateDiffIndexer creates a void implementation of interfaces.StateDiffIndexer
|
||||
func NewStateDiffIndexer(chainConfig *params.ChainConfig, config Config) *StateDiffIndexer {
|
||||
return &StateDiffIndexer{
|
||||
dump: config.Dump,
|
||||
chainConfig: chainConfig,
|
||||
}
|
||||
}
|
||||
|
||||
// ReportDBMetrics has nothing to report for dump
|
||||
func (sdi *StateDiffIndexer) ReportDBMetrics(time.Duration, <-chan bool) {}
|
||||
|
||||
// PushBlock pushes and indexes block data in sql, except state & storage nodes (includes header, uncles, transactions & receipts)
|
||||
// Returns an initiated DB transaction which must be Closed via defer to commit or rollback
|
||||
func (sdi *StateDiffIndexer) PushBlock(block *types.Block, receipts types.Receipts, totalDifficulty *big.Int) (interfaces.Batch, error) {
|
||||
start, t := time.Now(), time.Now()
|
||||
blockHash := block.Hash()
|
||||
blockHashStr := blockHash.String()
|
||||
height := block.NumberU64()
|
||||
traceMsg := fmt.Sprintf("indexer stats for statediff at %d with hash %s:\r\n", height, blockHashStr)
|
||||
transactions := block.Transactions()
|
||||
// Derive any missing fields
|
||||
if err := receipts.DeriveFields(sdi.chainConfig, blockHash, height, block.BaseFee(), transactions); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
// Generate the block iplds
|
||||
headerNode, txNodes, rctNodes, logNodes, err := ipld.FromBlockAndReceipts(block, receipts)
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("error creating IPLD nodes from block and receipts: %v", err)
|
||||
}
|
||||
|
||||
if len(txNodes) != len(rctNodes) {
|
||||
return nil, fmt.Errorf("expected number of transactions (%d), receipts (%d)", len(txNodes), len(rctNodes))
|
||||
}
|
||||
|
||||
// Calculate reward
|
||||
var reward *big.Int
|
||||
// in PoA networks block reward is 0
|
||||
if sdi.chainConfig.Clique != nil {
|
||||
reward = big.NewInt(0)
|
||||
} else {
|
||||
reward = shared.CalcEthBlockReward(block.Header(), block.Uncles(), block.Transactions(), receipts)
|
||||
}
|
||||
t = time.Now()
|
||||
|
||||
blockTx := &BatchTx{
|
||||
BlockNumber: block.Number().String(),
|
||||
dump: sdi.dump,
|
||||
iplds: make(chan models.IPLDModel),
|
||||
quit: make(chan struct{}),
|
||||
ipldCache: models.IPLDBatch{},
|
||||
submit: func(self *BatchTx, err error) error {
|
||||
close(self.quit)
|
||||
close(self.iplds)
|
||||
tDiff := time.Since(t)
|
||||
metrics.IndexerMetrics.StateStoreCodeProcessingTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("state, storage, and code storage processing time: %s\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
if err := self.flush(); err != nil {
|
||||
traceMsg += fmt.Sprintf(" TOTAL PROCESSING DURATION: %s\r\n", time.Since(start).String())
|
||||
log.Debug(traceMsg)
|
||||
return err
|
||||
}
|
||||
tDiff = time.Since(t)
|
||||
metrics.IndexerMetrics.PostgresCommitTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("postgres transaction commit duration: %s\r\n", tDiff.String())
|
||||
traceMsg += fmt.Sprintf(" TOTAL PROCESSING DURATION: %s\r\n", time.Since(start).String())
|
||||
log.Debug(traceMsg)
|
||||
return err
|
||||
},
|
||||
}
|
||||
go blockTx.cache()
|
||||
|
||||
tDiff := time.Since(t)
|
||||
metrics.IndexerMetrics.FreePostgresTimer.Update(tDiff)
|
||||
|
||||
traceMsg += fmt.Sprintf("time spent waiting for free postgres tx: %s:\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
|
||||
// Publish and index header, collect headerID
|
||||
var headerID string
|
||||
headerID, err = sdi.processHeader(blockTx, block.Header(), headerNode, reward, totalDifficulty)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
tDiff = time.Since(t)
|
||||
metrics.IndexerMetrics.HeaderProcessingTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("header processing time: %s\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
// Publish and index uncles
|
||||
err = sdi.processUncles(blockTx, headerID, block.Number(), block.UncleHash(), block.Uncles())
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
tDiff = time.Since(t)
|
||||
metrics.IndexerMetrics.UncleProcessingTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("uncle processing time: %s\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
// Publish and index receipts and txs
|
||||
err = sdi.processReceiptsAndTxs(blockTx, processArgs{
|
||||
headerID: headerID,
|
||||
blockNumber: block.Number(),
|
||||
receipts: receipts,
|
||||
txs: transactions,
|
||||
rctNodes: rctNodes,
|
||||
txNodes: txNodes,
|
||||
logNodes: logNodes,
|
||||
})
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
tDiff = time.Since(t)
|
||||
metrics.IndexerMetrics.TxAndRecProcessingTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("tx and receipt processing time: %s\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
|
||||
return blockTx, err
|
||||
}
|
||||
|
||||
// processHeader publishes and indexes a header IPLD in Postgres
|
||||
// it returns the headerID
|
||||
func (sdi *StateDiffIndexer) processHeader(tx *BatchTx, header *types.Header, headerNode ipld.IPLD, reward, td *big.Int) (string, error) {
|
||||
tx.cacheIPLD(headerNode)
|
||||
|
||||
headerID := header.Hash().String()
|
||||
mod := models.HeaderModel{
|
||||
CID: headerNode.Cid().String(),
|
||||
ParentHash: header.ParentHash.String(),
|
||||
BlockNumber: header.Number.String(),
|
||||
BlockHash: headerID,
|
||||
TotalDifficulty: td.String(),
|
||||
Reward: reward.String(),
|
||||
Bloom: header.Bloom.Bytes(),
|
||||
StateRoot: header.Root.String(),
|
||||
RctRoot: header.ReceiptHash.String(),
|
||||
TxRoot: header.TxHash.String(),
|
||||
UnclesHash: header.UncleHash.String(),
|
||||
Timestamp: header.Time,
|
||||
Coinbase: header.Coinbase.String(),
|
||||
}
|
||||
_, err := fmt.Fprintf(sdi.dump, "%+v\r\n", mod)
|
||||
return headerID, err
|
||||
}
|
||||
|
||||
// processUncles publishes and indexes uncle IPLDs in Postgres
|
||||
func (sdi *StateDiffIndexer) processUncles(tx *BatchTx, headerID string, blockNumber *big.Int, unclesHash common.Hash, uncles []*types.Header) error {
|
||||
// publish and index uncles
|
||||
uncleEncoding, err := rlp.EncodeToBytes(uncles)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
preparedHash := crypto.Keccak256Hash(uncleEncoding)
|
||||
if !bytes.Equal(preparedHash.Bytes(), unclesHash.Bytes()) {
|
||||
return fmt.Errorf("derived uncles hash (%s) does not match the hash in the header (%s)", preparedHash.Hex(), unclesHash.Hex())
|
||||
}
|
||||
unclesCID, err := ipld.RawdataToCid(ipld.MEthHeaderList, uncleEncoding, multihash.KECCAK_256)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
tx.cacheDirect(unclesCID.String(), uncleEncoding)
|
||||
for i, uncle := range uncles {
|
||||
var uncleReward *big.Int
|
||||
// in PoA networks uncle reward is 0
|
||||
if sdi.chainConfig.Clique != nil {
|
||||
uncleReward = big.NewInt(0)
|
||||
} else {
|
||||
uncleReward = shared.CalcUncleMinerReward(blockNumber.Uint64(), uncle.Number.Uint64())
|
||||
}
|
||||
uncle := models.UncleModel{
|
||||
BlockNumber: blockNumber.String(),
|
||||
HeaderID: headerID,
|
||||
CID: unclesCID.String(),
|
||||
ParentHash: uncle.ParentHash.String(),
|
||||
BlockHash: uncle.Hash().String(),
|
||||
Reward: uncleReward.String(),
|
||||
Index: int64(i),
|
||||
}
|
||||
if _, err := fmt.Fprintf(sdi.dump, "%+v\r\n", uncle); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
// processArgs bundles arguments to processReceiptsAndTxs
|
||||
type processArgs struct {
|
||||
headerID string
|
||||
blockNumber *big.Int
|
||||
receipts types.Receipts
|
||||
txs types.Transactions
|
||||
rctNodes []*ipld.EthReceipt
|
||||
txNodes []*ipld.EthTx
|
||||
logNodes [][]*ipld.EthLog
|
||||
}
|
||||
|
||||
// processReceiptsAndTxs publishes and indexes receipt and transaction IPLDs in Postgres
|
||||
func (sdi *StateDiffIndexer) processReceiptsAndTxs(tx *BatchTx, args processArgs) error {
|
||||
// Process receipts and txs
|
||||
signer := types.MakeSigner(sdi.chainConfig, args.blockNumber)
|
||||
for i, receipt := range args.receipts {
|
||||
txNode := args.txNodes[i]
|
||||
tx.cacheIPLD(txNode)
|
||||
|
||||
// Indexing
|
||||
// index tx
|
||||
trx := args.txs[i]
|
||||
trxID := trx.Hash().String()
|
||||
|
||||
var val string
|
||||
if trx.Value() != nil {
|
||||
val = trx.Value().String()
|
||||
}
|
||||
|
||||
// derive sender for the tx that corresponds with this receipt
|
||||
from, err := types.Sender(signer, trx)
|
||||
if err != nil {
|
||||
return fmt.Errorf("error deriving tx sender: %v", err)
|
||||
}
|
||||
txModel := models.TxModel{
|
||||
BlockNumber: args.blockNumber.String(),
|
||||
HeaderID: args.headerID,
|
||||
Dst: shared.HandleZeroAddrPointer(trx.To()),
|
||||
Src: shared.HandleZeroAddr(from),
|
||||
TxHash: trxID,
|
||||
Index: int64(i),
|
||||
CID: txNode.Cid().String(),
|
||||
Type: trx.Type(),
|
||||
Value: val,
|
||||
}
|
||||
if _, err := fmt.Fprintf(sdi.dump, "%+v\r\n", txModel); err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
// this is the contract address if this receipt is for a contract creation tx
|
||||
contract := shared.HandleZeroAddr(receipt.ContractAddress)
|
||||
|
||||
// index the receipt
|
||||
rctModel := &models.ReceiptModel{
|
||||
BlockNumber: args.blockNumber.String(),
|
||||
HeaderID: args.headerID,
|
||||
TxID: trxID,
|
||||
Contract: contract,
|
||||
CID: args.rctNodes[i].Cid().String(),
|
||||
}
|
||||
if len(receipt.PostState) == 0 {
|
||||
rctModel.PostStatus = receipt.Status
|
||||
} else {
|
||||
rctModel.PostState = common.Bytes2Hex(receipt.PostState)
|
||||
}
|
||||
|
||||
if _, err := fmt.Fprintf(sdi.dump, "%+v\r\n", rctModel); err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
logDataSet := make([]*models.LogsModel, len(receipt.Logs))
|
||||
for idx, l := range receipt.Logs {
|
||||
topicSet := make([]string, 4)
|
||||
for ti, topic := range l.Topics {
|
||||
topicSet[ti] = topic.Hex()
|
||||
}
|
||||
|
||||
logDataSet[idx] = &models.LogsModel{
|
||||
BlockNumber: args.blockNumber.String(),
|
||||
HeaderID: args.headerID,
|
||||
ReceiptID: trxID,
|
||||
Address: l.Address.String(),
|
||||
Index: int64(l.Index),
|
||||
CID: args.logNodes[i][idx].Cid().String(),
|
||||
Topic0: topicSet[0],
|
||||
Topic1: topicSet[1],
|
||||
Topic2: topicSet[2],
|
||||
Topic3: topicSet[3],
|
||||
}
|
||||
}
|
||||
|
||||
if _, err := fmt.Fprintf(sdi.dump, "%+v\r\n", logDataSet); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// PushStateNode publishes and indexes a state diff node object (including any child storage nodes) in the IPLD sql
|
||||
func (sdi *StateDiffIndexer) PushStateNode(batch interfaces.Batch, stateNode sdtypes.StateLeafNode, headerID string) error {
|
||||
tx, ok := batch.(*BatchTx)
|
||||
if !ok {
|
||||
return fmt.Errorf("dump: batch is expected to be of type %T, got %T", &BatchTx{}, batch)
|
||||
}
|
||||
// publish the state node
|
||||
var stateModel models.StateNodeModel
|
||||
if stateNode.Removed {
|
||||
// short circuit if it is a Removed node
|
||||
// this assumes the db has been initialized and a ipld.blocks entry for the Removed node is present
|
||||
stateModel = models.StateNodeModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
HeaderID: headerID,
|
||||
StateKey: common.BytesToHash(stateNode.AccountWrapper.LeafKey).String(),
|
||||
CID: shared.RemovedNodeStateCID,
|
||||
Removed: true,
|
||||
}
|
||||
} else {
|
||||
stateModel = models.StateNodeModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
HeaderID: headerID,
|
||||
StateKey: common.BytesToHash(stateNode.AccountWrapper.LeafKey).String(),
|
||||
CID: stateNode.AccountWrapper.CID,
|
||||
Removed: false,
|
||||
Balance: stateNode.AccountWrapper.Account.Balance.String(),
|
||||
Nonce: stateNode.AccountWrapper.Account.Nonce,
|
||||
CodeHash: common.BytesToHash(stateNode.AccountWrapper.Account.CodeHash).String(),
|
||||
StorageRoot: stateNode.AccountWrapper.Account.Root.String(),
|
||||
}
|
||||
}
|
||||
|
||||
// index the state node, collect the stateID to reference by FK
|
||||
if _, err := fmt.Fprintf(sdi.dump, "%+v\r\n", stateModel); err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
// if there are any storage nodes associated with this node, publish and index them
|
||||
for _, storageNode := range stateNode.StorageDiff {
|
||||
if storageNode.Removed {
|
||||
// short circuit if it is a Removed node
|
||||
// this assumes the db has been initialized and a ipld.blocks entry for the Removed node is present
|
||||
storageModel := models.StorageNodeModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
HeaderID: headerID,
|
||||
StateKey: common.BytesToHash(stateNode.AccountWrapper.LeafKey).String(),
|
||||
StorageKey: common.BytesToHash(storageNode.LeafKey).String(),
|
||||
CID: shared.RemovedNodeStorageCID,
|
||||
Removed: true,
|
||||
}
|
||||
if _, err := fmt.Fprintf(sdi.dump, "%+v\r\n", storageModel); err != nil {
|
||||
return err
|
||||
}
|
||||
continue
|
||||
}
|
||||
storageModel := models.StorageNodeModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
HeaderID: headerID,
|
||||
StateKey: common.BytesToHash(stateNode.AccountWrapper.LeafKey).String(),
|
||||
StorageKey: common.BytesToHash(storageNode.LeafKey).String(),
|
||||
CID: storageNode.CID,
|
||||
Removed: false,
|
||||
Value: storageNode.Value,
|
||||
}
|
||||
if _, err := fmt.Fprintf(sdi.dump, "%+v\r\n", storageModel); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// PushIPLD publishes iplds to ipld.blocks
|
||||
func (sdi *StateDiffIndexer) PushIPLD(batch interfaces.Batch, ipld sdtypes.IPLD) error {
|
||||
tx, ok := batch.(*BatchTx)
|
||||
if !ok {
|
||||
return fmt.Errorf("dump: batch is expected to be of type %T, got %T", &BatchTx{}, batch)
|
||||
}
|
||||
tx.cacheDirect(ipld.CID, ipld.Content)
|
||||
return nil
|
||||
}
|
||||
|
||||
// Close satisfies io.Closer
|
||||
func (sdi *StateDiffIndexer) Close() error {
|
||||
return sdi.dump.Close()
|
||||
}
|
||||
|
||||
// LoadWatchedAddresses satisfies the interfaces.StateDiffIndexer interface
|
||||
func (sdi *StateDiffIndexer) LoadWatchedAddresses() ([]common.Address, error) {
|
||||
return nil, nil
|
||||
}
|
||||
|
||||
// InsertWatchedAddresses satisfies the interfaces.StateDiffIndexer interface
|
||||
func (sdi *StateDiffIndexer) InsertWatchedAddresses(args []sdtypes.WatchAddressArg, currentBlockNumber *big.Int) error {
|
||||
return nil
|
||||
}
|
||||
|
||||
// RemoveWatchedAddresses satisfies the interfaces.StateDiffIndexer interface
|
||||
func (sdi *StateDiffIndexer) RemoveWatchedAddresses(args []sdtypes.WatchAddressArg) error {
|
||||
return nil
|
||||
}
|
||||
|
||||
// SetWatchedAddresses satisfies the interfaces.StateDiffIndexer interface
|
||||
func (sdi *StateDiffIndexer) SetWatchedAddresses(args []sdtypes.WatchAddressArg, currentBlockNumber *big.Int) error {
|
||||
return nil
|
||||
}
|
||||
|
||||
// ClearWatchedAddresses satisfies the interfaces.StateDiffIndexer interface
|
||||
func (sdi *StateDiffIndexer) ClearWatchedAddresses() error {
|
||||
return nil
|
||||
}
|
29
indexer/database/file/batch_tx.go
Normal file
29
indexer/database/file/batch_tx.go
Normal file
@ -0,0 +1,29 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package file
|
||||
|
||||
// BatchTx wraps a void with the state necessary for building the tx concurrently during trie difference iteration
|
||||
type BatchTx struct {
|
||||
BlockNumber string
|
||||
|
||||
submit func(blockTx *BatchTx, err error) error
|
||||
}
|
||||
|
||||
// Submit satisfies indexer.AtomicTx
|
||||
func (tx *BatchTx) Submit(err error) error {
|
||||
return tx.submit(tx, err)
|
||||
}
|
84
indexer/database/file/config.go
Normal file
84
indexer/database/file/config.go
Normal file
@ -0,0 +1,84 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package file
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"strings"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/node"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/shared"
|
||||
)
|
||||
|
||||
// FileMode to explicitly type the mode of file writer we are using
|
||||
type FileMode string
|
||||
|
||||
const (
|
||||
CSV FileMode = "CSV"
|
||||
SQL FileMode = "SQL"
|
||||
Unknown FileMode = "Unknown"
|
||||
)
|
||||
|
||||
// ResolveFileMode resolves a FileMode from a provided string
|
||||
func ResolveFileMode(str string) (FileMode, error) {
|
||||
switch strings.ToLower(str) {
|
||||
case "csv":
|
||||
return CSV, nil
|
||||
case "sql":
|
||||
return SQL, nil
|
||||
default:
|
||||
return Unknown, fmt.Errorf("unrecognized file type string: %s", str)
|
||||
}
|
||||
}
|
||||
|
||||
// Config holds params for writing out CSV or SQL files
|
||||
type Config struct {
|
||||
Mode FileMode
|
||||
OutputDir string
|
||||
FilePath string
|
||||
WatchedAddressesFilePath string
|
||||
NodeInfo node.Info
|
||||
}
|
||||
|
||||
// Type satisfies interfaces.Config
|
||||
func (c Config) Type() shared.DBType {
|
||||
return shared.FILE
|
||||
}
|
||||
|
||||
var nodeInfo = node.Info{
|
||||
GenesisBlock: "0xd4e56740f876aef8c010b86a40d5f56745a118d0906a34e69aec8c0db1cb8fa3",
|
||||
NetworkID: "1",
|
||||
ChainID: 1,
|
||||
ID: "mockNodeID",
|
||||
ClientName: "go-ethereum",
|
||||
}
|
||||
|
||||
// CSVTestConfig config for unit tests
|
||||
var CSVTestConfig = Config{
|
||||
Mode: CSV,
|
||||
OutputDir: "./statediffing_test",
|
||||
WatchedAddressesFilePath: "./statediffing_watched_addresses_test_file.csv",
|
||||
NodeInfo: nodeInfo,
|
||||
}
|
||||
|
||||
// SQLTestConfig config for unit tests
|
||||
var SQLTestConfig = Config{
|
||||
Mode: SQL,
|
||||
FilePath: "./statediffing_test_file.sql",
|
||||
WatchedAddressesFilePath: "./statediffing_watched_addresses_test_file.sql",
|
||||
NodeInfo: nodeInfo,
|
||||
}
|
118
indexer/database/file/csv_indexer_legacy_test.go
Normal file
118
indexer/database/file/csv_indexer_legacy_test.go
Normal file
@ -0,0 +1,118 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2022 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package file_test
|
||||
|
||||
import (
|
||||
"context"
|
||||
"errors"
|
||||
"fmt"
|
||||
"os"
|
||||
"path/filepath"
|
||||
"strings"
|
||||
"testing"
|
||||
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/file"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql/postgres"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/shared/schema"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test_helpers"
|
||||
)
|
||||
|
||||
const dbDirectory = "/file_indexer"
|
||||
const pgCopyStatement = `COPY %s FROM '%s' CSV`
|
||||
|
||||
func setupLegacyCSVIndexer(t *testing.T) {
|
||||
if _, err := os.Stat(file.CSVTestConfig.OutputDir); !errors.Is(err, os.ErrNotExist) {
|
||||
err := os.RemoveAll(file.CSVTestConfig.OutputDir)
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
ind, err = file.NewStateDiffIndexer(context.Background(), test.LegacyConfig, file.CSVTestConfig)
|
||||
require.NoError(t, err)
|
||||
|
||||
db, err = postgres.SetupSQLXDB()
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
}
|
||||
|
||||
func setupLegacyCSV(t *testing.T) {
|
||||
setupLegacyCSVIndexer(t)
|
||||
test.SetupLegacyTestData(t, ind)
|
||||
}
|
||||
|
||||
func dumpCSVFileData(t *testing.T) {
|
||||
outputDir := filepath.Join(dbDirectory, file.CSVTestConfig.OutputDir)
|
||||
workingDir, err := os.Getwd()
|
||||
require.NoError(t, err)
|
||||
|
||||
localOutputDir := filepath.Join(workingDir, file.CSVTestConfig.OutputDir)
|
||||
|
||||
for _, tbl := range file.Tables {
|
||||
err := test_helpers.DedupFile(file.TableFilePath(localOutputDir, tbl.Name))
|
||||
require.NoError(t, err)
|
||||
|
||||
var stmt string
|
||||
varcharColumns := tbl.VarcharColumns()
|
||||
if len(varcharColumns) > 0 {
|
||||
stmt = fmt.Sprintf(
|
||||
pgCopyStatement+" FORCE NOT NULL %s",
|
||||
tbl.Name,
|
||||
file.TableFilePath(outputDir, tbl.Name),
|
||||
strings.Join(varcharColumns, ", "),
|
||||
)
|
||||
} else {
|
||||
stmt = fmt.Sprintf(pgCopyStatement, tbl.Name, file.TableFilePath(outputDir, tbl.Name))
|
||||
}
|
||||
|
||||
_, err = db.Exec(context.Background(), stmt)
|
||||
require.NoError(t, err)
|
||||
}
|
||||
}
|
||||
|
||||
func resetAndDumpWatchedAddressesCSVFileData(t *testing.T) {
|
||||
test_helpers.TearDownDB(t, db)
|
||||
|
||||
outputFilePath := filepath.Join(dbDirectory, file.CSVTestConfig.WatchedAddressesFilePath)
|
||||
stmt := fmt.Sprintf(pgCopyStatement, schema.TableWatchedAddresses.Name, outputFilePath)
|
||||
|
||||
_, err = db.Exec(context.Background(), stmt)
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
func tearDownCSV(t *testing.T) {
|
||||
test_helpers.TearDownDB(t, db)
|
||||
require.NoError(t, db.Close())
|
||||
|
||||
require.NoError(t, os.RemoveAll(file.CSVTestConfig.OutputDir))
|
||||
|
||||
if err := os.Remove(file.CSVTestConfig.WatchedAddressesFilePath); !errors.Is(err, os.ErrNotExist) {
|
||||
require.NoError(t, err)
|
||||
}
|
||||
}
|
||||
|
||||
func TestLegacyCSVFileIndexer(t *testing.T) {
|
||||
t.Run("Publish and index header IPLDs", func(t *testing.T) {
|
||||
setupLegacyCSV(t)
|
||||
dumpCSVFileData(t)
|
||||
defer tearDownCSV(t)
|
||||
|
||||
test.TestLegacyIndexer(t, db)
|
||||
})
|
||||
}
|
255
indexer/database/file/csv_indexer_test.go
Normal file
255
indexer/database/file/csv_indexer_test.go
Normal file
@ -0,0 +1,255 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2022 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package file_test
|
||||
|
||||
import (
|
||||
"context"
|
||||
"errors"
|
||||
"math/big"
|
||||
"os"
|
||||
"testing"
|
||||
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/file"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql/postgres"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/mocks"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test"
|
||||
)
|
||||
|
||||
func setupCSVIndexer(t *testing.T) {
|
||||
file.CSVTestConfig.OutputDir = "./statediffing_test"
|
||||
|
||||
if _, err := os.Stat(file.CSVTestConfig.OutputDir); !errors.Is(err, os.ErrNotExist) {
|
||||
err := os.RemoveAll(file.CSVTestConfig.OutputDir)
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
if _, err := os.Stat(file.CSVTestConfig.WatchedAddressesFilePath); !errors.Is(err, os.ErrNotExist) {
|
||||
err := os.Remove(file.CSVTestConfig.WatchedAddressesFilePath)
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
ind, err = file.NewStateDiffIndexer(context.Background(), mocks.TestConfig, file.CSVTestConfig)
|
||||
require.NoError(t, err)
|
||||
|
||||
db, err = postgres.SetupSQLXDB()
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
}
|
||||
|
||||
func setupCSV(t *testing.T) {
|
||||
setupCSVIndexer(t)
|
||||
test.SetupTestData(t, ind)
|
||||
}
|
||||
|
||||
func setupCSVNonCanonical(t *testing.T) {
|
||||
setupCSVIndexer(t)
|
||||
test.SetupTestDataNonCanonical(t, ind)
|
||||
}
|
||||
|
||||
func TestCSVFileIndexer(t *testing.T) {
|
||||
t.Run("Publish and index header IPLDs in a single tx", func(t *testing.T) {
|
||||
setupCSV(t)
|
||||
dumpCSVFileData(t)
|
||||
defer tearDownCSV(t)
|
||||
|
||||
test.TestPublishAndIndexHeaderIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index transaction IPLDs in a single tx", func(t *testing.T) {
|
||||
setupCSV(t)
|
||||
dumpCSVFileData(t)
|
||||
defer tearDownCSV(t)
|
||||
|
||||
test.TestPublishAndIndexTransactionIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index log IPLDs for multiple receipt of a specific block", func(t *testing.T) {
|
||||
setupCSV(t)
|
||||
dumpCSVFileData(t)
|
||||
defer tearDownCSV(t)
|
||||
|
||||
test.TestPublishAndIndexLogIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index receipt IPLDs in a single tx", func(t *testing.T) {
|
||||
setupCSV(t)
|
||||
dumpCSVFileData(t)
|
||||
defer tearDownCSV(t)
|
||||
|
||||
test.TestPublishAndIndexReceiptIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index state IPLDs in a single tx", func(t *testing.T) {
|
||||
setupCSV(t)
|
||||
dumpCSVFileData(t)
|
||||
defer tearDownCSV(t)
|
||||
|
||||
test.TestPublishAndIndexStateIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index storage IPLDs in a single tx", func(t *testing.T) {
|
||||
setupCSV(t)
|
||||
dumpCSVFileData(t)
|
||||
defer tearDownCSV(t)
|
||||
|
||||
test.TestPublishAndIndexStorageIPLDs(t, db)
|
||||
})
|
||||
}
|
||||
|
||||
func TestCSVFileIndexerNonCanonical(t *testing.T) {
|
||||
t.Run("Publish and index header", func(t *testing.T) {
|
||||
setupCSVNonCanonical(t)
|
||||
dumpCSVFileData(t)
|
||||
defer tearDownCSV(t)
|
||||
|
||||
test.TestPublishAndIndexHeaderNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index transactions", func(t *testing.T) {
|
||||
setupCSVNonCanonical(t)
|
||||
dumpCSVFileData(t)
|
||||
defer tearDownCSV(t)
|
||||
|
||||
test.TestPublishAndIndexTransactionsNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index receipts", func(t *testing.T) {
|
||||
setupCSVNonCanonical(t)
|
||||
dumpCSVFileData(t)
|
||||
defer tearDownCSV(t)
|
||||
|
||||
test.TestPublishAndIndexReceiptsNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index logs", func(t *testing.T) {
|
||||
setupCSVNonCanonical(t)
|
||||
dumpCSVFileData(t)
|
||||
defer tearDownCSV(t)
|
||||
|
||||
test.TestPublishAndIndexLogsNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index state nodes", func(t *testing.T) {
|
||||
setupCSVNonCanonical(t)
|
||||
dumpCSVFileData(t)
|
||||
defer tearDownCSV(t)
|
||||
|
||||
test.TestPublishAndIndexStateNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index storage nodes", func(t *testing.T) {
|
||||
setupCSVNonCanonical(t)
|
||||
dumpCSVFileData(t)
|
||||
defer tearDownCSV(t)
|
||||
|
||||
test.TestPublishAndIndexStorageNonCanonical(t, db)
|
||||
})
|
||||
}
|
||||
|
||||
func TestCSVFileWatchAddressMethods(t *testing.T) {
|
||||
setupCSVIndexer(t)
|
||||
defer tearDownCSV(t)
|
||||
|
||||
t.Run("Load watched addresses (empty table)", func(t *testing.T) {
|
||||
test.TestLoadEmptyWatchedAddresses(t, ind)
|
||||
})
|
||||
|
||||
t.Run("Insert watched addresses", func(t *testing.T) {
|
||||
args := mocks.GetInsertWatchedAddressesArgs()
|
||||
err = ind.InsertWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt1)))
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesCSVFileData(t)
|
||||
|
||||
test.TestInsertWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Insert watched addresses (some already watched)", func(t *testing.T) {
|
||||
args := mocks.GetInsertAlreadyWatchedAddressesArgs()
|
||||
err = ind.InsertWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt2)))
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesCSVFileData(t)
|
||||
|
||||
test.TestInsertAlreadyWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Remove watched addresses", func(t *testing.T) {
|
||||
args := mocks.GetRemoveWatchedAddressesArgs()
|
||||
err = ind.RemoveWatchedAddresses(args)
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesCSVFileData(t)
|
||||
|
||||
test.TestRemoveWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Remove watched addresses (some non-watched)", func(t *testing.T) {
|
||||
args := mocks.GetRemoveNonWatchedAddressesArgs()
|
||||
err = ind.RemoveWatchedAddresses(args)
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesCSVFileData(t)
|
||||
|
||||
test.TestRemoveNonWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Set watched addresses", func(t *testing.T) {
|
||||
args := mocks.GetSetWatchedAddressesArgs()
|
||||
err = ind.SetWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt2)))
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesCSVFileData(t)
|
||||
|
||||
test.TestSetWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Set watched addresses (some already watched)", func(t *testing.T) {
|
||||
args := mocks.GetSetAlreadyWatchedAddressesArgs()
|
||||
err = ind.SetWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt3)))
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesCSVFileData(t)
|
||||
|
||||
test.TestSetAlreadyWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Load watched addresses", func(t *testing.T) {
|
||||
test.TestLoadWatchedAddresses(t, ind)
|
||||
})
|
||||
|
||||
t.Run("Clear watched addresses", func(t *testing.T) {
|
||||
err = ind.ClearWatchedAddresses()
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesCSVFileData(t)
|
||||
|
||||
test.TestClearWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Clear watched addresses (empty table)", func(t *testing.T) {
|
||||
err = ind.ClearWatchedAddresses()
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesCSVFileData(t)
|
||||
|
||||
test.TestClearEmptyWatchedAddresses(t, db)
|
||||
})
|
||||
}
|
418
indexer/database/file/csv_writer.go
Normal file
418
indexer/database/file/csv_writer.go
Normal file
@ -0,0 +1,418 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2022 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package file
|
||||
|
||||
import (
|
||||
"encoding/csv"
|
||||
"errors"
|
||||
"fmt"
|
||||
"math/big"
|
||||
"os"
|
||||
"path/filepath"
|
||||
"strconv"
|
||||
|
||||
"github.com/thoas/go-funk"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/metrics"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/ipld"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/models"
|
||||
nodeinfo "github.com/ethereum/go-ethereum/statediff/indexer/node"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/shared/schema"
|
||||
sdtypes "github.com/ethereum/go-ethereum/statediff/types"
|
||||
)
|
||||
|
||||
var (
|
||||
Tables = []*schema.Table{
|
||||
&schema.TableIPLDBlock,
|
||||
&schema.TableNodeInfo,
|
||||
&schema.TableHeader,
|
||||
&schema.TableStateNode,
|
||||
&schema.TableStorageNode,
|
||||
&schema.TableUncle,
|
||||
&schema.TableTransaction,
|
||||
&schema.TableReceipt,
|
||||
&schema.TableLog,
|
||||
}
|
||||
)
|
||||
|
||||
type tableRow struct {
|
||||
table schema.Table
|
||||
values []interface{}
|
||||
}
|
||||
|
||||
type CSVWriter struct {
|
||||
// dir containing output files
|
||||
dir string
|
||||
|
||||
writers fileWriters
|
||||
watchedAddressesWriter fileWriter
|
||||
|
||||
rows chan tableRow
|
||||
flushChan chan struct{}
|
||||
flushFinished chan struct{}
|
||||
quitChan chan struct{}
|
||||
doneChan chan struct{}
|
||||
}
|
||||
|
||||
type fileWriter struct {
|
||||
*csv.Writer
|
||||
file *os.File
|
||||
}
|
||||
|
||||
// fileWriters wraps the file writers for each output table
|
||||
type fileWriters map[string]fileWriter
|
||||
|
||||
func newFileWriter(path string) (ret fileWriter, err error) {
|
||||
file, err := os.OpenFile(path, os.O_WRONLY|os.O_CREATE|os.O_APPEND, 0644)
|
||||
if err != nil {
|
||||
return
|
||||
}
|
||||
|
||||
ret = fileWriter{
|
||||
Writer: csv.NewWriter(file),
|
||||
file: file,
|
||||
}
|
||||
|
||||
return
|
||||
}
|
||||
|
||||
func makeFileWriters(dir string, tables []*schema.Table) (fileWriters, error) {
|
||||
if err := os.MkdirAll(dir, 0755); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
writers := fileWriters{}
|
||||
for _, tbl := range tables {
|
||||
w, err := newFileWriter(TableFilePath(dir, tbl.Name))
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
writers[tbl.Name] = w
|
||||
}
|
||||
return writers, nil
|
||||
}
|
||||
|
||||
func (tx fileWriters) write(tbl *schema.Table, args ...interface{}) error {
|
||||
row := tbl.ToCsvRow(args...)
|
||||
return tx[tbl.Name].Write(row)
|
||||
}
|
||||
|
||||
func (tx fileWriters) close() error {
|
||||
for _, w := range tx {
|
||||
err := w.file.Close()
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
func (tx fileWriters) flush() error {
|
||||
for _, w := range tx {
|
||||
w.Flush()
|
||||
if err := w.Error(); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
func NewCSVWriter(path string, watchedAddressesFilePath string) (*CSVWriter, error) {
|
||||
if err := os.MkdirAll(path, 0777); err != nil {
|
||||
return nil, fmt.Errorf("unable to make MkdirAll for path: %s err: %s", path, err)
|
||||
}
|
||||
|
||||
writers, err := makeFileWriters(path, Tables)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
watchedAddressesWriter, err := newFileWriter(watchedAddressesFilePath)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
csvWriter := &CSVWriter{
|
||||
writers: writers,
|
||||
watchedAddressesWriter: watchedAddressesWriter,
|
||||
dir: path,
|
||||
rows: make(chan tableRow),
|
||||
flushChan: make(chan struct{}),
|
||||
flushFinished: make(chan struct{}),
|
||||
quitChan: make(chan struct{}),
|
||||
doneChan: make(chan struct{}),
|
||||
}
|
||||
return csvWriter, nil
|
||||
}
|
||||
|
||||
func (csw *CSVWriter) Loop() {
|
||||
go func() {
|
||||
defer close(csw.doneChan)
|
||||
for {
|
||||
select {
|
||||
case row := <-csw.rows:
|
||||
err := csw.writers.write(&row.table, row.values...)
|
||||
if err != nil {
|
||||
panic(fmt.Sprintf("error writing csv buffer: %v", err))
|
||||
}
|
||||
case <-csw.quitChan:
|
||||
if err := csw.writers.flush(); err != nil {
|
||||
panic(fmt.Sprintf("error writing csv buffer to file: %v", err))
|
||||
}
|
||||
return
|
||||
case <-csw.flushChan:
|
||||
if err := csw.writers.flush(); err != nil {
|
||||
panic(fmt.Sprintf("error writing csv buffer to file: %v", err))
|
||||
}
|
||||
csw.flushFinished <- struct{}{}
|
||||
}
|
||||
}
|
||||
}()
|
||||
}
|
||||
|
||||
// Flush sends a flush signal to the looping process
|
||||
func (csw *CSVWriter) Flush() {
|
||||
csw.flushChan <- struct{}{}
|
||||
<-csw.flushFinished
|
||||
}
|
||||
|
||||
func TableFilePath(dir, name string) string { return filepath.Join(dir, name+".csv") }
|
||||
|
||||
// Close satisfies io.Closer
|
||||
func (csw *CSVWriter) Close() error {
|
||||
close(csw.quitChan)
|
||||
<-csw.doneChan
|
||||
close(csw.rows)
|
||||
close(csw.flushChan)
|
||||
close(csw.flushFinished)
|
||||
return csw.writers.close()
|
||||
}
|
||||
|
||||
func (csw *CSVWriter) upsertNode(node nodeinfo.Info) {
|
||||
var values []interface{}
|
||||
values = append(values, node.GenesisBlock, node.NetworkID, node.ID, node.ClientName, node.ChainID)
|
||||
csw.rows <- tableRow{schema.TableNodeInfo, values}
|
||||
}
|
||||
|
||||
func (csw *CSVWriter) upsertIPLD(ipld models.IPLDModel) {
|
||||
var values []interface{}
|
||||
values = append(values, ipld.BlockNumber, ipld.Key, ipld.Data)
|
||||
csw.rows <- tableRow{schema.TableIPLDBlock, values}
|
||||
}
|
||||
|
||||
func (csw *CSVWriter) upsertIPLDDirect(blockNumber, key string, value []byte) {
|
||||
csw.upsertIPLD(models.IPLDModel{
|
||||
BlockNumber: blockNumber,
|
||||
Key: key,
|
||||
Data: value,
|
||||
})
|
||||
}
|
||||
|
||||
func (csw *CSVWriter) upsertIPLDNode(blockNumber string, i ipld.IPLD) {
|
||||
csw.upsertIPLD(models.IPLDModel{
|
||||
BlockNumber: blockNumber,
|
||||
Key: i.Cid().String(),
|
||||
Data: i.RawData(),
|
||||
})
|
||||
}
|
||||
|
||||
func (csw *CSVWriter) upsertHeaderCID(header models.HeaderModel) {
|
||||
var values []interface{}
|
||||
values = append(values, header.BlockNumber, header.BlockHash, header.ParentHash, header.CID,
|
||||
header.TotalDifficulty, header.NodeIDs, header.Reward, header.StateRoot, header.TxRoot,
|
||||
header.RctRoot, header.UnclesHash, header.Bloom, strconv.FormatUint(header.Timestamp, 10), header.Coinbase)
|
||||
csw.rows <- tableRow{schema.TableHeader, values}
|
||||
metrics.IndexerMetrics.BlocksCounter.Inc(1)
|
||||
}
|
||||
|
||||
func (csw *CSVWriter) upsertUncleCID(uncle models.UncleModel) {
|
||||
var values []interface{}
|
||||
values = append(values, uncle.BlockNumber, uncle.BlockHash, uncle.HeaderID, uncle.ParentHash, uncle.CID,
|
||||
uncle.Reward, uncle.Index)
|
||||
csw.rows <- tableRow{schema.TableUncle, values}
|
||||
}
|
||||
|
||||
func (csw *CSVWriter) upsertTransactionCID(transaction models.TxModel) {
|
||||
var values []interface{}
|
||||
values = append(values, transaction.BlockNumber, transaction.HeaderID, transaction.TxHash, transaction.CID, transaction.Dst,
|
||||
transaction.Src, transaction.Index, transaction.Type, transaction.Value)
|
||||
csw.rows <- tableRow{schema.TableTransaction, values}
|
||||
metrics.IndexerMetrics.TransactionsCounter.Inc(1)
|
||||
}
|
||||
|
||||
func (csw *CSVWriter) upsertReceiptCID(rct *models.ReceiptModel) {
|
||||
var values []interface{}
|
||||
values = append(values, rct.BlockNumber, rct.HeaderID, rct.TxID, rct.CID, rct.Contract,
|
||||
rct.PostState, rct.PostStatus)
|
||||
csw.rows <- tableRow{schema.TableReceipt, values}
|
||||
metrics.IndexerMetrics.ReceiptsCounter.Inc(1)
|
||||
}
|
||||
|
||||
func (csw *CSVWriter) upsertLogCID(logs []*models.LogsModel) {
|
||||
for _, l := range logs {
|
||||
var values []interface{}
|
||||
values = append(values, l.BlockNumber, l.HeaderID, l.CID, l.ReceiptID, l.Address, l.Index, l.Topic0,
|
||||
l.Topic1, l.Topic2, l.Topic3)
|
||||
csw.rows <- tableRow{schema.TableLog, values}
|
||||
metrics.IndexerMetrics.LogsCounter.Inc(1)
|
||||
}
|
||||
}
|
||||
|
||||
func (csw *CSVWriter) upsertStateCID(stateNode models.StateNodeModel) {
|
||||
balance := stateNode.Balance
|
||||
if stateNode.Removed {
|
||||
balance = "0"
|
||||
}
|
||||
|
||||
var values []interface{}
|
||||
values = append(values, stateNode.BlockNumber, stateNode.HeaderID, stateNode.StateKey, stateNode.CID,
|
||||
true, balance, strconv.FormatUint(stateNode.Nonce, 10), stateNode.CodeHash, stateNode.StorageRoot, stateNode.Removed)
|
||||
csw.rows <- tableRow{schema.TableStateNode, values}
|
||||
}
|
||||
|
||||
func (csw *CSVWriter) upsertStorageCID(storageCID models.StorageNodeModel) {
|
||||
var values []interface{}
|
||||
values = append(values, storageCID.BlockNumber, storageCID.HeaderID, storageCID.StateKey, storageCID.StorageKey, storageCID.CID,
|
||||
true, storageCID.Value, storageCID.Removed)
|
||||
csw.rows <- tableRow{schema.TableStorageNode, values}
|
||||
}
|
||||
|
||||
// LoadWatchedAddresses loads watched addresses from a file
|
||||
func (csw *CSVWriter) loadWatchedAddresses() ([]common.Address, error) {
|
||||
watchedAddressesFilePath := csw.watchedAddressesWriter.file.Name()
|
||||
// load csv rows from watched addresses file
|
||||
rows, err := loadWatchedAddressesRows(watchedAddressesFilePath)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
// extract addresses from the csv rows
|
||||
watchedAddresses := funk.Map(rows, func(row []string) common.Address {
|
||||
// first column is for address in eth_meta.watched_addresses
|
||||
addressString := row[0]
|
||||
|
||||
return common.HexToAddress(addressString)
|
||||
}).([]common.Address)
|
||||
|
||||
return watchedAddresses, nil
|
||||
}
|
||||
|
||||
// InsertWatchedAddresses inserts the given addresses in a file
|
||||
func (csw *CSVWriter) insertWatchedAddresses(args []sdtypes.WatchAddressArg, currentBlockNumber *big.Int) error {
|
||||
// load csv rows from watched addresses file
|
||||
watchedAddresses, err := csw.loadWatchedAddresses()
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
// append rows for new addresses to existing csv file
|
||||
for _, arg := range args {
|
||||
// ignore if already watched
|
||||
if funk.Contains(watchedAddresses, common.HexToAddress(arg.Address)) {
|
||||
continue
|
||||
}
|
||||
|
||||
var values []interface{}
|
||||
values = append(values, arg.Address, strconv.FormatUint(arg.CreatedAt, 10), currentBlockNumber.String(), "0")
|
||||
row := schema.TableWatchedAddresses.ToCsvRow(values...)
|
||||
|
||||
// writing directly instead of using rows channel as it needs to be flushed immediately
|
||||
err = csw.watchedAddressesWriter.Write(row)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
|
||||
// watched addresses need to be flushed immediately to the file to keep them in sync with in-memory watched addresses
|
||||
csw.watchedAddressesWriter.Flush()
|
||||
err = csw.watchedAddressesWriter.Error()
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// RemoveWatchedAddresses removes the given watched addresses from a file
|
||||
func (csw *CSVWriter) removeWatchedAddresses(args []sdtypes.WatchAddressArg) error {
|
||||
// load csv rows from watched addresses file
|
||||
watchedAddressesFilePath := csw.watchedAddressesWriter.file.Name()
|
||||
rows, err := loadWatchedAddressesRows(watchedAddressesFilePath)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
// get rid of rows having addresses to be removed
|
||||
filteredRows := funk.Filter(rows, func(row []string) bool {
|
||||
return !funk.Contains(args, func(arg sdtypes.WatchAddressArg) bool {
|
||||
// Compare first column in table for address
|
||||
return arg.Address == row[0]
|
||||
})
|
||||
}).([][]string)
|
||||
|
||||
return dumpWatchedAddressesRows(csw.watchedAddressesWriter, filteredRows)
|
||||
}
|
||||
|
||||
// SetWatchedAddresses clears and inserts the given addresses in a file
|
||||
func (csw *CSVWriter) setWatchedAddresses(args []sdtypes.WatchAddressArg, currentBlockNumber *big.Int) error {
|
||||
var rows [][]string
|
||||
for _, arg := range args {
|
||||
row := schema.TableWatchedAddresses.ToCsvRow(arg.Address, strconv.FormatUint(arg.CreatedAt, 10), currentBlockNumber.String(), "0")
|
||||
rows = append(rows, row)
|
||||
}
|
||||
|
||||
return dumpWatchedAddressesRows(csw.watchedAddressesWriter, rows)
|
||||
}
|
||||
|
||||
// loadCSVWatchedAddresses loads csv rows from the given file
|
||||
func loadWatchedAddressesRows(filePath string) ([][]string, error) {
|
||||
file, err := os.Open(filePath)
|
||||
if err != nil {
|
||||
if errors.Is(err, os.ErrNotExist) {
|
||||
return [][]string{}, nil
|
||||
}
|
||||
|
||||
return nil, fmt.Errorf("error opening watched addresses file: %v", err)
|
||||
}
|
||||
|
||||
defer file.Close()
|
||||
reader := csv.NewReader(file)
|
||||
|
||||
return reader.ReadAll()
|
||||
}
|
||||
|
||||
// dumpWatchedAddressesRows dumps csv rows to the given file
|
||||
func dumpWatchedAddressesRows(watchedAddressesWriter fileWriter, filteredRows [][]string) error {
|
||||
file := watchedAddressesWriter.file
|
||||
file.Close()
|
||||
|
||||
file, err := os.Create(file.Name())
|
||||
if err != nil {
|
||||
return fmt.Errorf("error creating watched addresses file: %v", err)
|
||||
}
|
||||
|
||||
watchedAddressesWriter.Writer = csv.NewWriter(file)
|
||||
watchedAddressesWriter.file = file
|
||||
|
||||
for _, row := range filteredRows {
|
||||
watchedAddressesWriter.Write(row)
|
||||
}
|
||||
|
||||
watchedAddressesWriter.Flush()
|
||||
|
||||
return nil
|
||||
}
|
60
indexer/database/file/helpers.go
Normal file
60
indexer/database/file/helpers.go
Normal file
@ -0,0 +1,60 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package file
|
||||
|
||||
import "bytes"
|
||||
|
||||
// formatPostgresStringArray parses an array of strings into the proper Postgres string representation of that array
|
||||
func formatPostgresStringArray(a []string) string {
|
||||
if a == nil {
|
||||
return ""
|
||||
}
|
||||
|
||||
if n := len(a); n > 0 {
|
||||
// There will be at least two curly brackets, 2*N bytes of quotes,
|
||||
// and N-1 bytes of delimiters.
|
||||
b := make([]byte, 1, 1+3*n)
|
||||
b[0] = '{'
|
||||
|
||||
b = appendArrayQuotedBytes(b, []byte(a[0]))
|
||||
for i := 1; i < n; i++ {
|
||||
b = append(b, ',')
|
||||
b = appendArrayQuotedBytes(b, []byte(a[i]))
|
||||
}
|
||||
|
||||
return string(append(b, '}'))
|
||||
}
|
||||
|
||||
return "{}"
|
||||
}
|
||||
|
||||
func appendArrayQuotedBytes(b, v []byte) []byte {
|
||||
b = append(b, '"')
|
||||
for {
|
||||
i := bytes.IndexAny(v, `"\`)
|
||||
if i < 0 {
|
||||
b = append(b, v...)
|
||||
break
|
||||
}
|
||||
if i > 0 {
|
||||
b = append(b, v[:i]...)
|
||||
}
|
||||
b = append(b, '\\', v[i])
|
||||
v = v[i+1:]
|
||||
}
|
||||
return append(b, '"')
|
||||
}
|
492
indexer/database/file/indexer.go
Normal file
492
indexer/database/file/indexer.go
Normal file
@ -0,0 +1,492 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package file
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"context"
|
||||
"errors"
|
||||
"fmt"
|
||||
"math/big"
|
||||
"os"
|
||||
"sync"
|
||||
"sync/atomic"
|
||||
"time"
|
||||
|
||||
"github.com/lib/pq"
|
||||
"github.com/multiformats/go-multihash"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/crypto"
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
"github.com/ethereum/go-ethereum/params"
|
||||
"github.com/ethereum/go-ethereum/rlp"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/metrics"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/interfaces"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/ipld"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/models"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/shared"
|
||||
sdtypes "github.com/ethereum/go-ethereum/statediff/types"
|
||||
)
|
||||
|
||||
const defaultCSVOutputDir = "./statediff_output"
|
||||
const defaultSQLFilePath = "./statediff.sql"
|
||||
const defaultWatchedAddressesCSVFilePath = "./statediff-watched-addresses.csv"
|
||||
const defaultWatchedAddressesSQLFilePath = "./statediff-watched-addresses.sql"
|
||||
|
||||
const watchedAddressesInsert = "INSERT INTO eth_meta.watched_addresses (address, created_at, watched_at) VALUES ('%s', '%d', '%d') ON CONFLICT (address) DO NOTHING;"
|
||||
|
||||
var _ interfaces.StateDiffIndexer = &StateDiffIndexer{}
|
||||
|
||||
// StateDiffIndexer satisfies the indexer.StateDiffIndexer interface for ethereum statediff objects on top of a void
|
||||
type StateDiffIndexer struct {
|
||||
fileWriter FileWriter
|
||||
chainConfig *params.ChainConfig
|
||||
nodeID string
|
||||
wg *sync.WaitGroup
|
||||
removedCacheFlag *uint32
|
||||
}
|
||||
|
||||
// NewStateDiffIndexer creates a void implementation of interfaces.StateDiffIndexer
|
||||
func NewStateDiffIndexer(ctx context.Context, chainConfig *params.ChainConfig, config Config) (*StateDiffIndexer, error) {
|
||||
var err error
|
||||
var writer FileWriter
|
||||
|
||||
watchedAddressesFilePath := config.WatchedAddressesFilePath
|
||||
|
||||
switch config.Mode {
|
||||
case CSV:
|
||||
outputDir := config.OutputDir
|
||||
if outputDir == "" {
|
||||
outputDir = defaultCSVOutputDir
|
||||
}
|
||||
|
||||
if _, err := os.Stat(outputDir); !errors.Is(err, os.ErrNotExist) {
|
||||
return nil, fmt.Errorf("cannot create output directory, directory (%s) already exists", outputDir)
|
||||
}
|
||||
log.Info("Writing statediff CSV files to directory", "file", outputDir)
|
||||
|
||||
if watchedAddressesFilePath == "" {
|
||||
watchedAddressesFilePath = defaultWatchedAddressesCSVFilePath
|
||||
}
|
||||
log.Info("Writing watched addresses to file", "file", watchedAddressesFilePath)
|
||||
|
||||
writer, err = NewCSVWriter(outputDir, watchedAddressesFilePath)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
case SQL:
|
||||
filePath := config.FilePath
|
||||
if filePath == "" {
|
||||
filePath = defaultSQLFilePath
|
||||
}
|
||||
if _, err := os.Stat(filePath); !errors.Is(err, os.ErrNotExist) {
|
||||
return nil, fmt.Errorf("cannot create file, file (%s) already exists", filePath)
|
||||
}
|
||||
file, err := os.Create(filePath)
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("unable to create file (%s), err: %v", filePath, err)
|
||||
}
|
||||
log.Info("Writing statediff SQL statements to file", "file", filePath)
|
||||
|
||||
if watchedAddressesFilePath == "" {
|
||||
watchedAddressesFilePath = defaultWatchedAddressesSQLFilePath
|
||||
}
|
||||
log.Info("Writing watched addresses to file", "file", watchedAddressesFilePath)
|
||||
|
||||
writer = NewSQLWriter(file, watchedAddressesFilePath)
|
||||
default:
|
||||
return nil, fmt.Errorf("unrecognized file mode: %s", config.Mode)
|
||||
}
|
||||
|
||||
wg := new(sync.WaitGroup)
|
||||
writer.Loop()
|
||||
writer.upsertNode(config.NodeInfo)
|
||||
|
||||
return &StateDiffIndexer{
|
||||
fileWriter: writer,
|
||||
chainConfig: chainConfig,
|
||||
nodeID: config.NodeInfo.ID,
|
||||
wg: wg,
|
||||
}, nil
|
||||
}
|
||||
|
||||
// ReportDBMetrics has nothing to report for dump
|
||||
func (sdi *StateDiffIndexer) ReportDBMetrics(time.Duration, <-chan bool) {}
|
||||
|
||||
// PushBlock pushes and indexes block data in sql, except state & storage nodes (includes header, uncles, transactions & receipts)
|
||||
// Returns an initiated DB transaction which must be Closed via defer to commit or rollback
|
||||
func (sdi *StateDiffIndexer) PushBlock(block *types.Block, receipts types.Receipts, totalDifficulty *big.Int) (interfaces.Batch, error) {
|
||||
sdi.removedCacheFlag = new(uint32)
|
||||
start, t := time.Now(), time.Now()
|
||||
blockHash := block.Hash()
|
||||
blockHashStr := blockHash.String()
|
||||
height := block.NumberU64()
|
||||
traceMsg := fmt.Sprintf("indexer stats for statediff at %d with hash %s:\r\n", height, blockHashStr)
|
||||
transactions := block.Transactions()
|
||||
// Derive any missing fields
|
||||
if err := receipts.DeriveFields(sdi.chainConfig, blockHash, height, block.BaseFee(), transactions); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
// Generate the block iplds
|
||||
headerNode, txNodes, rctNodes, logNodes, err := ipld.FromBlockAndReceipts(block, receipts)
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("error creating IPLD nodes from block and receipts: %v", err)
|
||||
}
|
||||
|
||||
if len(txNodes) != len(rctNodes) {
|
||||
return nil, fmt.Errorf("expected number of transactions (%d), receipts (%d)", len(txNodes), len(rctNodes))
|
||||
}
|
||||
|
||||
// Calculate reward
|
||||
var reward *big.Int
|
||||
// in PoA networks block reward is 0
|
||||
if sdi.chainConfig.Clique != nil {
|
||||
reward = big.NewInt(0)
|
||||
} else {
|
||||
reward = shared.CalcEthBlockReward(block.Header(), block.Uncles(), block.Transactions(), receipts)
|
||||
}
|
||||
t = time.Now()
|
||||
|
||||
blockTx := &BatchTx{
|
||||
BlockNumber: block.Number().String(),
|
||||
submit: func(self *BatchTx, err error) error {
|
||||
tDiff := time.Since(t)
|
||||
metrics.IndexerMetrics.StateStoreCodeProcessingTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("state, storage, and code storage processing time: %s\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
sdi.fileWriter.Flush()
|
||||
tDiff = time.Since(t)
|
||||
metrics.IndexerMetrics.PostgresCommitTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("postgres transaction commit duration: %s\r\n", tDiff.String())
|
||||
traceMsg += fmt.Sprintf(" TOTAL PROCESSING DURATION: %s\r\n", time.Since(start).String())
|
||||
log.Debug(traceMsg)
|
||||
return err
|
||||
},
|
||||
}
|
||||
tDiff := time.Since(t)
|
||||
metrics.IndexerMetrics.FreePostgresTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("time spent waiting for free postgres tx: %s:\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
|
||||
// write header, collect headerID
|
||||
headerID := sdi.processHeader(block.Header(), headerNode, reward, totalDifficulty)
|
||||
tDiff = time.Since(t)
|
||||
metrics.IndexerMetrics.HeaderProcessingTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("header processing time: %s\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
|
||||
// write uncles
|
||||
sdi.processUncles(headerID, block.Number(), block.UncleHash(), block.Uncles())
|
||||
tDiff = time.Since(t)
|
||||
metrics.IndexerMetrics.UncleProcessingTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("uncle processing time: %s\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
|
||||
// write receipts and txs
|
||||
err = sdi.processReceiptsAndTxs(processArgs{
|
||||
headerID: headerID,
|
||||
blockNumber: block.Number(),
|
||||
receipts: receipts,
|
||||
txs: transactions,
|
||||
rctNodes: rctNodes,
|
||||
txNodes: txNodes,
|
||||
logNodes: logNodes,
|
||||
})
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
tDiff = time.Since(t)
|
||||
metrics.IndexerMetrics.TxAndRecProcessingTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("tx and receipt processing time: %s\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
|
||||
return blockTx, err
|
||||
}
|
||||
|
||||
// processHeader write a header IPLD insert SQL stmt to a file
|
||||
// it returns the headerID
|
||||
func (sdi *StateDiffIndexer) processHeader(header *types.Header, headerNode ipld.IPLD, reward, td *big.Int) string {
|
||||
sdi.fileWriter.upsertIPLDNode(header.Number.String(), headerNode)
|
||||
|
||||
var baseFee *string
|
||||
if header.BaseFee != nil {
|
||||
baseFee = new(string)
|
||||
*baseFee = header.BaseFee.String()
|
||||
}
|
||||
headerID := header.Hash().String()
|
||||
sdi.fileWriter.upsertHeaderCID(models.HeaderModel{
|
||||
NodeIDs: pq.StringArray([]string{sdi.nodeID}),
|
||||
CID: headerNode.Cid().String(),
|
||||
ParentHash: header.ParentHash.String(),
|
||||
BlockNumber: header.Number.String(),
|
||||
BlockHash: headerID,
|
||||
TotalDifficulty: td.String(),
|
||||
Reward: reward.String(),
|
||||
Bloom: header.Bloom.Bytes(),
|
||||
StateRoot: header.Root.String(),
|
||||
RctRoot: header.ReceiptHash.String(),
|
||||
TxRoot: header.TxHash.String(),
|
||||
UnclesHash: header.UncleHash.String(),
|
||||
Timestamp: header.Time,
|
||||
Coinbase: header.Coinbase.String(),
|
||||
})
|
||||
return headerID
|
||||
}
|
||||
|
||||
// processUncles publishes and indexes uncle IPLDs in Postgres
|
||||
func (sdi *StateDiffIndexer) processUncles(headerID string, blockNumber *big.Int, unclesHash common.Hash, uncles []*types.Header) error {
|
||||
// publish and index uncles
|
||||
uncleEncoding, err := rlp.EncodeToBytes(uncles)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
preparedHash := crypto.Keccak256Hash(uncleEncoding)
|
||||
if !bytes.Equal(preparedHash.Bytes(), unclesHash.Bytes()) {
|
||||
return fmt.Errorf("derived uncles hash (%s) does not match the hash in the header (%s)", preparedHash.Hex(), unclesHash.Hex())
|
||||
}
|
||||
unclesCID, err := ipld.RawdataToCid(ipld.MEthHeaderList, uncleEncoding, multihash.KECCAK_256)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
sdi.fileWriter.upsertIPLDDirect(blockNumber.String(), unclesCID.String(), uncleEncoding)
|
||||
for i, uncle := range uncles {
|
||||
var uncleReward *big.Int
|
||||
// in PoA networks uncle reward is 0
|
||||
if sdi.chainConfig.Clique != nil {
|
||||
uncleReward = big.NewInt(0)
|
||||
} else {
|
||||
uncleReward = shared.CalcUncleMinerReward(blockNumber.Uint64(), uncle.Number.Uint64())
|
||||
}
|
||||
sdi.fileWriter.upsertUncleCID(models.UncleModel{
|
||||
BlockNumber: blockNumber.String(),
|
||||
HeaderID: headerID,
|
||||
CID: unclesCID.String(),
|
||||
ParentHash: uncle.ParentHash.String(),
|
||||
BlockHash: uncle.Hash().String(),
|
||||
Reward: uncleReward.String(),
|
||||
Index: int64(i),
|
||||
})
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
// processArgs bundles arguments to processReceiptsAndTxs
|
||||
type processArgs struct {
|
||||
headerID string
|
||||
blockNumber *big.Int
|
||||
receipts types.Receipts
|
||||
txs types.Transactions
|
||||
rctNodes []*ipld.EthReceipt
|
||||
txNodes []*ipld.EthTx
|
||||
logNodes [][]*ipld.EthLog
|
||||
}
|
||||
|
||||
// processReceiptsAndTxs writes receipt and tx IPLD insert SQL stmts to a file
|
||||
func (sdi *StateDiffIndexer) processReceiptsAndTxs(args processArgs) error {
|
||||
// Process receipts and txs
|
||||
signer := types.MakeSigner(sdi.chainConfig, args.blockNumber)
|
||||
for i, receipt := range args.receipts {
|
||||
txNode := args.txNodes[i]
|
||||
sdi.fileWriter.upsertIPLDNode(args.blockNumber.String(), txNode)
|
||||
sdi.fileWriter.upsertIPLDNode(args.blockNumber.String(), args.rctNodes[i])
|
||||
|
||||
// index tx
|
||||
trx := args.txs[i]
|
||||
txID := trx.Hash().String()
|
||||
|
||||
var val string
|
||||
if trx.Value() != nil {
|
||||
val = trx.Value().String()
|
||||
}
|
||||
|
||||
// derive sender for the tx that corresponds with this receipt
|
||||
from, err := types.Sender(signer, trx)
|
||||
if err != nil {
|
||||
return fmt.Errorf("error deriving tx sender: %v", err)
|
||||
}
|
||||
txModel := models.TxModel{
|
||||
BlockNumber: args.blockNumber.String(),
|
||||
HeaderID: args.headerID,
|
||||
Dst: shared.HandleZeroAddrPointer(trx.To()),
|
||||
Src: shared.HandleZeroAddr(from),
|
||||
TxHash: txID,
|
||||
Index: int64(i),
|
||||
CID: txNode.Cid().String(),
|
||||
Type: trx.Type(),
|
||||
Value: val,
|
||||
}
|
||||
sdi.fileWriter.upsertTransactionCID(txModel)
|
||||
|
||||
// this is the contract address if this receipt is for a contract creation tx
|
||||
contract := shared.HandleZeroAddr(receipt.ContractAddress)
|
||||
|
||||
// index receipt
|
||||
rctModel := &models.ReceiptModel{
|
||||
BlockNumber: args.blockNumber.String(),
|
||||
HeaderID: args.headerID,
|
||||
TxID: txID,
|
||||
Contract: contract,
|
||||
CID: args.rctNodes[i].Cid().String(),
|
||||
}
|
||||
if len(receipt.PostState) == 0 {
|
||||
rctModel.PostStatus = receipt.Status
|
||||
} else {
|
||||
rctModel.PostState = common.BytesToHash(receipt.PostState).String()
|
||||
}
|
||||
sdi.fileWriter.upsertReceiptCID(rctModel)
|
||||
|
||||
// index logs
|
||||
logDataSet := make([]*models.LogsModel, len(receipt.Logs))
|
||||
for idx, l := range receipt.Logs {
|
||||
sdi.fileWriter.upsertIPLDNode(args.blockNumber.String(), args.logNodes[i][idx])
|
||||
topicSet := make([]string, 4)
|
||||
for ti, topic := range l.Topics {
|
||||
topicSet[ti] = topic.Hex()
|
||||
}
|
||||
|
||||
logDataSet[idx] = &models.LogsModel{
|
||||
BlockNumber: args.blockNumber.String(),
|
||||
HeaderID: args.headerID,
|
||||
ReceiptID: txID,
|
||||
Address: l.Address.String(),
|
||||
Index: int64(l.Index),
|
||||
CID: args.logNodes[i][idx].Cid().String(),
|
||||
Topic0: topicSet[0],
|
||||
Topic1: topicSet[1],
|
||||
Topic2: topicSet[2],
|
||||
Topic3: topicSet[3],
|
||||
}
|
||||
}
|
||||
sdi.fileWriter.upsertLogCID(logDataSet)
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// PushStateNode writes a state diff node object (including any child storage nodes) IPLD insert SQL stmt to a file
|
||||
func (sdi *StateDiffIndexer) PushStateNode(batch interfaces.Batch, stateNode sdtypes.StateLeafNode, headerID string) error {
|
||||
tx, ok := batch.(*BatchTx)
|
||||
if !ok {
|
||||
return fmt.Errorf("file: batch is expected to be of type %T, got %T", &BatchTx{}, batch)
|
||||
}
|
||||
// publish the state node
|
||||
var stateModel models.StateNodeModel
|
||||
if stateNode.Removed {
|
||||
if atomic.LoadUint32(sdi.removedCacheFlag) == 0 {
|
||||
atomic.StoreUint32(sdi.removedCacheFlag, 1)
|
||||
sdi.fileWriter.upsertIPLDDirect(tx.BlockNumber, shared.RemovedNodeStateCID, []byte{})
|
||||
}
|
||||
stateModel = models.StateNodeModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
HeaderID: headerID,
|
||||
StateKey: common.BytesToHash(stateNode.AccountWrapper.LeafKey).String(),
|
||||
CID: shared.RemovedNodeStateCID,
|
||||
Removed: true,
|
||||
}
|
||||
} else {
|
||||
stateModel = models.StateNodeModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
HeaderID: headerID,
|
||||
StateKey: common.BytesToHash(stateNode.AccountWrapper.LeafKey).String(),
|
||||
CID: stateNode.AccountWrapper.CID,
|
||||
Removed: false,
|
||||
Balance: stateNode.AccountWrapper.Account.Balance.String(),
|
||||
Nonce: stateNode.AccountWrapper.Account.Nonce,
|
||||
CodeHash: common.BytesToHash(stateNode.AccountWrapper.Account.CodeHash).String(),
|
||||
StorageRoot: stateNode.AccountWrapper.Account.Root.String(),
|
||||
}
|
||||
}
|
||||
|
||||
// index the state node
|
||||
sdi.fileWriter.upsertStateCID(stateModel)
|
||||
|
||||
// if there are any storage nodes associated with this node, publish and index them
|
||||
for _, storageNode := range stateNode.StorageDiff {
|
||||
if storageNode.Removed {
|
||||
if atomic.LoadUint32(sdi.removedCacheFlag) == 0 {
|
||||
atomic.StoreUint32(sdi.removedCacheFlag, 1)
|
||||
sdi.fileWriter.upsertIPLDDirect(tx.BlockNumber, shared.RemovedNodeStorageCID, []byte{})
|
||||
}
|
||||
storageModel := models.StorageNodeModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
HeaderID: headerID,
|
||||
StateKey: common.BytesToHash(stateNode.AccountWrapper.LeafKey).String(),
|
||||
StorageKey: common.BytesToHash(storageNode.LeafKey).String(),
|
||||
CID: shared.RemovedNodeStorageCID,
|
||||
Removed: true,
|
||||
Value: []byte{},
|
||||
}
|
||||
sdi.fileWriter.upsertStorageCID(storageModel)
|
||||
continue
|
||||
}
|
||||
storageModel := models.StorageNodeModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
HeaderID: headerID,
|
||||
StateKey: common.BytesToHash(stateNode.AccountWrapper.LeafKey).String(),
|
||||
StorageKey: common.BytesToHash(storageNode.LeafKey).String(),
|
||||
CID: storageNode.CID,
|
||||
Removed: false,
|
||||
Value: storageNode.Value,
|
||||
}
|
||||
sdi.fileWriter.upsertStorageCID(storageModel)
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// PushIPLD writes iplds to ipld.blocks
|
||||
func (sdi *StateDiffIndexer) PushIPLD(batch interfaces.Batch, ipld sdtypes.IPLD) error {
|
||||
tx, ok := batch.(*BatchTx)
|
||||
if !ok {
|
||||
return fmt.Errorf("file: batch is expected to be of type %T, got %T", &BatchTx{}, batch)
|
||||
}
|
||||
sdi.fileWriter.upsertIPLDDirect(tx.BlockNumber, ipld.CID, ipld.Content)
|
||||
return nil
|
||||
}
|
||||
|
||||
// Close satisfies io.Closer
|
||||
func (sdi *StateDiffIndexer) Close() error {
|
||||
return sdi.fileWriter.Close()
|
||||
}
|
||||
|
||||
// LoadWatchedAddresses loads watched addresses from a file
|
||||
func (sdi *StateDiffIndexer) LoadWatchedAddresses() ([]common.Address, error) {
|
||||
return sdi.fileWriter.loadWatchedAddresses()
|
||||
}
|
||||
|
||||
// InsertWatchedAddresses inserts the given addresses in a file
|
||||
func (sdi *StateDiffIndexer) InsertWatchedAddresses(args []sdtypes.WatchAddressArg, currentBlockNumber *big.Int) error {
|
||||
return sdi.fileWriter.insertWatchedAddresses(args, currentBlockNumber)
|
||||
}
|
||||
|
||||
// RemoveWatchedAddresses removes the given watched addresses from a file
|
||||
func (sdi *StateDiffIndexer) RemoveWatchedAddresses(args []sdtypes.WatchAddressArg) error {
|
||||
return sdi.fileWriter.removeWatchedAddresses(args)
|
||||
}
|
||||
|
||||
// SetWatchedAddresses clears and inserts the given addresses in a file
|
||||
func (sdi *StateDiffIndexer) SetWatchedAddresses(args []sdtypes.WatchAddressArg, currentBlockNumber *big.Int) error {
|
||||
return sdi.fileWriter.setWatchedAddresses(args, currentBlockNumber)
|
||||
}
|
||||
|
||||
// ClearWatchedAddresses clears all the watched addresses from a file
|
||||
func (sdi *StateDiffIndexer) ClearWatchedAddresses() error {
|
||||
return sdi.SetWatchedAddresses([]sdtypes.WatchAddressArg{}, big.NewInt(0))
|
||||
}
|
57
indexer/database/file/interfaces.go
Normal file
57
indexer/database/file/interfaces.go
Normal file
@ -0,0 +1,57 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2022 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package file
|
||||
|
||||
import (
|
||||
"math/big"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/ipld"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/models"
|
||||
nodeinfo "github.com/ethereum/go-ethereum/statediff/indexer/node"
|
||||
"github.com/ethereum/go-ethereum/statediff/types"
|
||||
)
|
||||
|
||||
// Writer interface required by the file indexer
|
||||
type FileWriter interface {
|
||||
// Methods used to control the writer
|
||||
Loop()
|
||||
Close() error
|
||||
Flush()
|
||||
|
||||
// Methods to upsert ethereum data model objects
|
||||
upsertNode(node nodeinfo.Info)
|
||||
upsertHeaderCID(header models.HeaderModel)
|
||||
upsertUncleCID(uncle models.UncleModel)
|
||||
upsertTransactionCID(transaction models.TxModel)
|
||||
upsertReceiptCID(rct *models.ReceiptModel)
|
||||
upsertLogCID(logs []*models.LogsModel)
|
||||
upsertStateCID(stateNode models.StateNodeModel)
|
||||
upsertStorageCID(storageCID models.StorageNodeModel)
|
||||
upsertIPLD(ipld models.IPLDModel)
|
||||
|
||||
// Methods to upsert IPLD in different ways
|
||||
upsertIPLDDirect(blockNumber, key string, value []byte)
|
||||
upsertIPLDNode(blockNumber string, i ipld.IPLD)
|
||||
|
||||
// Methods to read and write watched addresses
|
||||
loadWatchedAddresses() ([]common.Address, error)
|
||||
insertWatchedAddresses(args []types.WatchAddressArg, currentBlockNumber *big.Int) error
|
||||
removeWatchedAddresses(args []types.WatchAddressArg) error
|
||||
setWatchedAddresses(args []types.WatchAddressArg, currentBlockNumber *big.Int) error
|
||||
}
|
112
indexer/database/file/mainnet_tests/indexer_test.go
Normal file
112
indexer/database/file/mainnet_tests/indexer_test.go
Normal file
@ -0,0 +1,112 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package mainnet_tests
|
||||
|
||||
import (
|
||||
"context"
|
||||
"errors"
|
||||
"fmt"
|
||||
"math/big"
|
||||
"os"
|
||||
"testing"
|
||||
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/params"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/file"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql/postgres"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/interfaces"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test_helpers"
|
||||
)
|
||||
|
||||
var (
|
||||
err error
|
||||
db sql.Database
|
||||
ind interfaces.StateDiffIndexer
|
||||
chainConf = params.MainnetChainConfig
|
||||
)
|
||||
|
||||
func init() {
|
||||
if os.Getenv("MODE") != "statediff" {
|
||||
fmt.Println("Skipping statediff test")
|
||||
os.Exit(0)
|
||||
}
|
||||
if os.Getenv("STATEDIFF_DB") != "file" {
|
||||
fmt.Println("Skipping statediff .sql file writing mode test")
|
||||
os.Exit(0)
|
||||
}
|
||||
}
|
||||
|
||||
func TestPushBlockAndState(t *testing.T) {
|
||||
conf := test_helpers.GetTestConfig()
|
||||
|
||||
for _, blockNumber := range test_helpers.ProblemBlocks {
|
||||
conf.BlockNumber = big.NewInt(blockNumber)
|
||||
tb, trs, err := test_helpers.TestBlockAndReceipts(conf)
|
||||
require.NoError(t, err)
|
||||
|
||||
testPushBlockAndState(t, tb, trs)
|
||||
}
|
||||
|
||||
testBlock, testReceipts, err := test_helpers.TestBlockAndReceiptsFromEnv(conf)
|
||||
require.NoError(t, err)
|
||||
|
||||
testPushBlockAndState(t, testBlock, testReceipts)
|
||||
}
|
||||
|
||||
func testPushBlockAndState(t *testing.T, block *types.Block, receipts types.Receipts) {
|
||||
t.Run("Test PushBlock and PushStateNode", func(t *testing.T) {
|
||||
setupMainnetIndexer(t)
|
||||
defer dumpData(t)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestBlock(t, ind, block, receipts)
|
||||
})
|
||||
}
|
||||
|
||||
func setupMainnetIndexer(t *testing.T) {
|
||||
if _, err := os.Stat(file.CSVTestConfig.FilePath); !errors.Is(err, os.ErrNotExist) {
|
||||
err := os.Remove(file.CSVTestConfig.FilePath)
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
ind, err = file.NewStateDiffIndexer(context.Background(), chainConf, file.CSVTestConfig)
|
||||
require.NoError(t, err)
|
||||
|
||||
db, err = postgres.SetupSQLXDB()
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
}
|
||||
|
||||
func dumpData(t *testing.T) {
|
||||
sqlFileBytes, err := os.ReadFile(file.CSVTestConfig.FilePath)
|
||||
require.NoError(t, err)
|
||||
|
||||
_, err = db.Exec(context.Background(), string(sqlFileBytes))
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
func tearDown(t *testing.T) {
|
||||
test_helpers.TearDownDB(t, db)
|
||||
require.NoError(t, db.Close())
|
||||
|
||||
require.NoError(t, os.Remove(file.CSVTestConfig.FilePath))
|
||||
}
|
101
indexer/database/file/sql_indexer_legacy_test.go
Normal file
101
indexer/database/file/sql_indexer_legacy_test.go
Normal file
@ -0,0 +1,101 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package file_test
|
||||
|
||||
import (
|
||||
"context"
|
||||
"errors"
|
||||
"os"
|
||||
"testing"
|
||||
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/file"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql/postgres"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/interfaces"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test_helpers"
|
||||
)
|
||||
|
||||
var (
|
||||
db sql.Database
|
||||
err error
|
||||
ind interfaces.StateDiffIndexer
|
||||
)
|
||||
|
||||
func setupLegacySQLIndexer(t *testing.T) {
|
||||
if _, err := os.Stat(file.SQLTestConfig.FilePath); !errors.Is(err, os.ErrNotExist) {
|
||||
err := os.Remove(file.SQLTestConfig.FilePath)
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
ind, err = file.NewStateDiffIndexer(context.Background(), test.LegacyConfig, file.SQLTestConfig)
|
||||
require.NoError(t, err)
|
||||
|
||||
db, err = postgres.SetupSQLXDB()
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
}
|
||||
|
||||
func setupLegacySQL(t *testing.T) {
|
||||
setupLegacySQLIndexer(t)
|
||||
test.SetupLegacyTestData(t, ind)
|
||||
}
|
||||
|
||||
func dumpFileData(t *testing.T) {
|
||||
err := test_helpers.DedupFile(file.SQLTestConfig.FilePath)
|
||||
require.NoError(t, err)
|
||||
|
||||
sqlFileBytes, err := os.ReadFile(file.SQLTestConfig.FilePath)
|
||||
require.NoError(t, err)
|
||||
|
||||
_, err = db.Exec(context.Background(), string(sqlFileBytes))
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
func resetAndDumpWatchedAddressesFileData(t *testing.T) {
|
||||
test_helpers.TearDownDB(t, db)
|
||||
|
||||
sqlFileBytes, err := os.ReadFile(file.SQLTestConfig.WatchedAddressesFilePath)
|
||||
require.NoError(t, err)
|
||||
|
||||
_, err = db.Exec(context.Background(), string(sqlFileBytes))
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
func tearDown(t *testing.T) {
|
||||
test_helpers.TearDownDB(t, db)
|
||||
require.NoError(t, db.Close())
|
||||
|
||||
require.NoError(t, os.Remove(file.SQLTestConfig.FilePath))
|
||||
|
||||
if err := os.Remove(file.SQLTestConfig.WatchedAddressesFilePath); !errors.Is(err, os.ErrNotExist) {
|
||||
require.NoError(t, err)
|
||||
}
|
||||
}
|
||||
|
||||
func TestLegacySQLFileIndexer(t *testing.T) {
|
||||
t.Run("Publish and index header IPLDs", func(t *testing.T) {
|
||||
setupLegacySQL(t)
|
||||
dumpFileData(t)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestLegacyIndexer(t, db)
|
||||
})
|
||||
}
|
253
indexer/database/file/sql_indexer_test.go
Normal file
253
indexer/database/file/sql_indexer_test.go
Normal file
@ -0,0 +1,253 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package file_test
|
||||
|
||||
import (
|
||||
"context"
|
||||
"errors"
|
||||
"math/big"
|
||||
"os"
|
||||
"testing"
|
||||
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/file"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql/postgres"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/mocks"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test"
|
||||
)
|
||||
|
||||
func setupIndexer(t *testing.T) {
|
||||
if _, err := os.Stat(file.SQLTestConfig.FilePath); !errors.Is(err, os.ErrNotExist) {
|
||||
err := os.Remove(file.SQLTestConfig.FilePath)
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
if _, err := os.Stat(file.SQLTestConfig.WatchedAddressesFilePath); !errors.Is(err, os.ErrNotExist) {
|
||||
err := os.Remove(file.SQLTestConfig.WatchedAddressesFilePath)
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
ind, err = file.NewStateDiffIndexer(context.Background(), mocks.TestConfig, file.SQLTestConfig)
|
||||
require.NoError(t, err)
|
||||
|
||||
db, err = postgres.SetupSQLXDB()
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
}
|
||||
|
||||
func setup(t *testing.T) {
|
||||
setupIndexer(t)
|
||||
test.SetupTestData(t, ind)
|
||||
}
|
||||
|
||||
func setupSQLNonCanonical(t *testing.T) {
|
||||
setupIndexer(t)
|
||||
test.SetupTestDataNonCanonical(t, ind)
|
||||
}
|
||||
|
||||
func TestSQLFileIndexer(t *testing.T) {
|
||||
t.Run("Publish and index header IPLDs in a single tx", func(t *testing.T) {
|
||||
setup(t)
|
||||
dumpFileData(t)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestPublishAndIndexHeaderIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index transaction IPLDs in a single tx", func(t *testing.T) {
|
||||
setup(t)
|
||||
dumpFileData(t)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestPublishAndIndexTransactionIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index log IPLDs for multiple receipt of a specific block", func(t *testing.T) {
|
||||
setup(t)
|
||||
dumpFileData(t)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestPublishAndIndexLogIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index receipt IPLDs in a single tx", func(t *testing.T) {
|
||||
setup(t)
|
||||
dumpFileData(t)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestPublishAndIndexReceiptIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index state IPLDs in a single tx", func(t *testing.T) {
|
||||
setup(t)
|
||||
dumpFileData(t)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestPublishAndIndexStateIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index storage IPLDs in a single tx", func(t *testing.T) {
|
||||
setup(t)
|
||||
dumpFileData(t)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestPublishAndIndexStorageIPLDs(t, db)
|
||||
})
|
||||
}
|
||||
|
||||
func TestSQLFileIndexerNonCanonical(t *testing.T) {
|
||||
t.Run("Publish and index header", func(t *testing.T) {
|
||||
setupSQLNonCanonical(t)
|
||||
dumpFileData(t)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestPublishAndIndexHeaderNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index transactions", func(t *testing.T) {
|
||||
setupSQLNonCanonical(t)
|
||||
dumpFileData(t)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestPublishAndIndexTransactionsNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index receipts", func(t *testing.T) {
|
||||
setupSQLNonCanonical(t)
|
||||
dumpFileData(t)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestPublishAndIndexReceiptsNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index logs", func(t *testing.T) {
|
||||
setupSQLNonCanonical(t)
|
||||
dumpFileData(t)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestPublishAndIndexLogsNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index state nodes", func(t *testing.T) {
|
||||
setupSQLNonCanonical(t)
|
||||
dumpFileData(t)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestPublishAndIndexStateNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index storage nodes", func(t *testing.T) {
|
||||
setupSQLNonCanonical(t)
|
||||
dumpFileData(t)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestPublishAndIndexStorageNonCanonical(t, db)
|
||||
})
|
||||
}
|
||||
|
||||
func TestSQLFileWatchAddressMethods(t *testing.T) {
|
||||
setupIndexer(t)
|
||||
defer tearDown(t)
|
||||
|
||||
t.Run("Load watched addresses (empty table)", func(t *testing.T) {
|
||||
test.TestLoadEmptyWatchedAddresses(t, ind)
|
||||
})
|
||||
|
||||
t.Run("Insert watched addresses", func(t *testing.T) {
|
||||
args := mocks.GetInsertWatchedAddressesArgs()
|
||||
err = ind.InsertWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt1)))
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesFileData(t)
|
||||
|
||||
test.TestInsertWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Insert watched addresses (some already watched)", func(t *testing.T) {
|
||||
args := mocks.GetInsertAlreadyWatchedAddressesArgs()
|
||||
err = ind.InsertWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt2)))
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesFileData(t)
|
||||
|
||||
test.TestInsertAlreadyWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Remove watched addresses", func(t *testing.T) {
|
||||
args := mocks.GetRemoveWatchedAddressesArgs()
|
||||
err = ind.RemoveWatchedAddresses(args)
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesFileData(t)
|
||||
|
||||
test.TestRemoveWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Remove watched addresses (some non-watched)", func(t *testing.T) {
|
||||
args := mocks.GetRemoveNonWatchedAddressesArgs()
|
||||
err = ind.RemoveWatchedAddresses(args)
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesFileData(t)
|
||||
|
||||
test.TestRemoveNonWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Set watched addresses", func(t *testing.T) {
|
||||
args := mocks.GetSetWatchedAddressesArgs()
|
||||
err = ind.SetWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt2)))
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesFileData(t)
|
||||
|
||||
test.TestSetWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Set watched addresses (some already watched)", func(t *testing.T) {
|
||||
args := mocks.GetSetAlreadyWatchedAddressesArgs()
|
||||
err = ind.SetWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt3)))
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesFileData(t)
|
||||
|
||||
test.TestSetAlreadyWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Load watched addresses", func(t *testing.T) {
|
||||
test.TestLoadWatchedAddresses(t, ind)
|
||||
})
|
||||
|
||||
t.Run("Clear watched addresses", func(t *testing.T) {
|
||||
err = ind.ClearWatchedAddresses()
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesFileData(t)
|
||||
|
||||
test.TestClearWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Clear watched addresses (empty table)", func(t *testing.T) {
|
||||
err = ind.ClearWatchedAddresses()
|
||||
require.NoError(t, err)
|
||||
|
||||
resetAndDumpWatchedAddressesFileData(t)
|
||||
|
||||
test.TestClearEmptyWatchedAddresses(t, db)
|
||||
})
|
||||
}
|
391
indexer/database/file/sql_writer.go
Normal file
391
indexer/database/file/sql_writer.go
Normal file
@ -0,0 +1,391 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package file
|
||||
|
||||
import (
|
||||
"bufio"
|
||||
"errors"
|
||||
"fmt"
|
||||
"io"
|
||||
"math/big"
|
||||
"os"
|
||||
|
||||
pg_query "github.com/pganalyze/pg_query_go/v2"
|
||||
"github.com/thoas/go-funk"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/metrics"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/ipld"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/models"
|
||||
nodeinfo "github.com/ethereum/go-ethereum/statediff/indexer/node"
|
||||
"github.com/ethereum/go-ethereum/statediff/types"
|
||||
)
|
||||
|
||||
var (
|
||||
pipeSize = 65336 // min(linuxPipeSize, macOSPipeSize)
|
||||
writeBufferSize = pipeSize * 16 * 96
|
||||
)
|
||||
|
||||
// SQLWriter writes sql statements to a file
|
||||
type SQLWriter struct {
|
||||
wc io.WriteCloser
|
||||
stmts chan []byte
|
||||
collatedStmt []byte
|
||||
collationIndex int
|
||||
|
||||
flushChan chan struct{}
|
||||
flushFinished chan struct{}
|
||||
quitChan chan struct{}
|
||||
doneChan chan struct{}
|
||||
|
||||
watchedAddressesFilePath string
|
||||
}
|
||||
|
||||
// NewSQLWriter creates a new pointer to a Writer
|
||||
func NewSQLWriter(wc io.WriteCloser, watchedAddressesFilePath string) *SQLWriter {
|
||||
return &SQLWriter{
|
||||
wc: wc,
|
||||
stmts: make(chan []byte),
|
||||
collatedStmt: make([]byte, writeBufferSize),
|
||||
flushChan: make(chan struct{}),
|
||||
flushFinished: make(chan struct{}),
|
||||
quitChan: make(chan struct{}),
|
||||
doneChan: make(chan struct{}),
|
||||
watchedAddressesFilePath: watchedAddressesFilePath,
|
||||
}
|
||||
}
|
||||
|
||||
// Loop enables concurrent writes to the underlying os.File
|
||||
// since os.File does not buffer, it utilizes an internal buffer that is the size of a unix pipe
|
||||
// by using copy() and tracking the index/size of the buffer, we require only the initial memory allocation
|
||||
func (sqw *SQLWriter) Loop() {
|
||||
sqw.collationIndex = 0
|
||||
go func() {
|
||||
defer close(sqw.doneChan)
|
||||
var l int
|
||||
for {
|
||||
select {
|
||||
case stmt := <-sqw.stmts:
|
||||
l = len(stmt)
|
||||
if sqw.collationIndex+l > writeBufferSize {
|
||||
if err := sqw.flush(); err != nil {
|
||||
panic(fmt.Sprintf("error writing sql stmts buffer to file: %v", err))
|
||||
}
|
||||
if l > writeBufferSize {
|
||||
if _, err := sqw.wc.Write(stmt); err != nil {
|
||||
panic(fmt.Sprintf("error writing large sql stmt to file: %v", err))
|
||||
}
|
||||
continue
|
||||
}
|
||||
}
|
||||
copy(sqw.collatedStmt[sqw.collationIndex:sqw.collationIndex+l], stmt)
|
||||
sqw.collationIndex += l
|
||||
case <-sqw.quitChan:
|
||||
if err := sqw.flush(); err != nil {
|
||||
panic(fmt.Sprintf("error writing sql stmts buffer to file: %v", err))
|
||||
}
|
||||
return
|
||||
case <-sqw.flushChan:
|
||||
if err := sqw.flush(); err != nil {
|
||||
panic(fmt.Sprintf("error writing sql stmts buffer to file: %v", err))
|
||||
}
|
||||
sqw.flushFinished <- struct{}{}
|
||||
}
|
||||
}
|
||||
}()
|
||||
}
|
||||
|
||||
// Close satisfies io.Closer
|
||||
func (sqw *SQLWriter) Close() error {
|
||||
close(sqw.quitChan)
|
||||
<-sqw.doneChan
|
||||
close(sqw.stmts)
|
||||
close(sqw.flushChan)
|
||||
close(sqw.flushFinished)
|
||||
return sqw.wc.Close()
|
||||
}
|
||||
|
||||
// Flush sends a flush signal to the looping process
|
||||
func (sqw *SQLWriter) Flush() {
|
||||
sqw.flushChan <- struct{}{}
|
||||
<-sqw.flushFinished
|
||||
}
|
||||
|
||||
func (sqw *SQLWriter) flush() error {
|
||||
if _, err := sqw.wc.Write(sqw.collatedStmt[0:sqw.collationIndex]); err != nil {
|
||||
return err
|
||||
}
|
||||
sqw.collationIndex = 0
|
||||
return nil
|
||||
}
|
||||
|
||||
const (
|
||||
nodeInsert = "INSERT INTO nodes (genesis_block, network_id, node_id, client_name, chain_id) VALUES " +
|
||||
"('%s', '%s', '%s', '%s', %d);\n"
|
||||
|
||||
ipldInsert = "INSERT INTO ipld.blocks (block_number, key, data) VALUES ('%s', '%s', '\\x%x');\n"
|
||||
|
||||
headerInsert = "INSERT INTO eth.header_cids (block_number, block_hash, parent_hash, cid, td, node_ids, reward, " +
|
||||
"state_root, tx_root, receipt_root, uncles_hash, bloom, timestamp, coinbase) VALUES " +
|
||||
"('%s', '%s', '%s', '%s', '%s', '%s', '%s', '%s', '%s', '%s', '%s', '\\x%x', %d, '%s');\n"
|
||||
|
||||
uncleInsert = "INSERT INTO eth.uncle_cids (block_number, block_hash, header_id, parent_hash, cid, reward, index) VALUES " +
|
||||
"('%s', '%s', '%s', '%s', '%s', '%s', %d);\n"
|
||||
|
||||
txInsert = "INSERT INTO eth.transaction_cids (block_number, header_id, tx_hash, cid, dst, src, index, tx_type, " +
|
||||
"value) VALUES ('%s', '%s', '%s', '%s', '%s', '%s', %d, %d, '%s');\n"
|
||||
|
||||
rctInsert = "INSERT INTO eth.receipt_cids (block_number, header_id, tx_id, cid, contract, post_state, " +
|
||||
"post_status) VALUES ('%s', '%s', '%s', '%s', '%s', '%s', %d);\n"
|
||||
|
||||
logInsert = "INSERT INTO eth.log_cids (block_number, header_id, cid, rct_id, address, index, topic0, topic1, topic2, " +
|
||||
"topic3) VALUES ('%s', '%s', '%s', '%s', '%s', %d, '%s', '%s', '%s', '%s');\n"
|
||||
|
||||
stateInsert = "INSERT INTO eth.state_cids (block_number, header_id, state_leaf_key, cid, removed, diff, " +
|
||||
"balance, nonce, code_hash, storage_root) VALUES ('%s', '%s', '%s', '%s', %t, %t, '%s', %d, '%s', '%s');\n"
|
||||
|
||||
storageInsert = "INSERT INTO eth.storage_cids (block_number, header_id, state_leaf_key, storage_leaf_key, cid, " +
|
||||
"removed, diff, val) VALUES ('%s', '%s', '%s', '%s', '%s', %t, %t, '\\x%x');\n"
|
||||
)
|
||||
|
||||
func (sqw *SQLWriter) upsertNode(node nodeinfo.Info) {
|
||||
sqw.stmts <- []byte(fmt.Sprintf(nodeInsert, node.GenesisBlock, node.NetworkID, node.ID, node.ClientName, node.ChainID))
|
||||
}
|
||||
|
||||
func (sqw *SQLWriter) upsertIPLD(ipld models.IPLDModel) {
|
||||
sqw.stmts <- []byte(fmt.Sprintf(ipldInsert, ipld.BlockNumber, ipld.Key, ipld.Data))
|
||||
}
|
||||
|
||||
func (sqw *SQLWriter) upsertIPLDDirect(blockNumber, key string, value []byte) {
|
||||
sqw.upsertIPLD(models.IPLDModel{
|
||||
BlockNumber: blockNumber,
|
||||
Key: key,
|
||||
Data: value,
|
||||
})
|
||||
}
|
||||
|
||||
func (sqw *SQLWriter) upsertIPLDNode(blockNumber string, i ipld.IPLD) {
|
||||
sqw.upsertIPLD(models.IPLDModel{
|
||||
BlockNumber: blockNumber,
|
||||
Key: i.Cid().String(),
|
||||
Data: i.RawData(),
|
||||
})
|
||||
}
|
||||
|
||||
func (sqw *SQLWriter) upsertHeaderCID(header models.HeaderModel) {
|
||||
stmt := fmt.Sprintf(headerInsert, header.BlockNumber, header.BlockHash, header.ParentHash, header.CID,
|
||||
header.TotalDifficulty, formatPostgresStringArray(header.NodeIDs), header.Reward, header.StateRoot, header.TxRoot,
|
||||
header.RctRoot, header.UnclesHash, header.Bloom, header.Timestamp, header.Coinbase)
|
||||
sqw.stmts <- []byte(stmt)
|
||||
metrics.IndexerMetrics.BlocksCounter.Inc(1)
|
||||
}
|
||||
|
||||
func (sqw *SQLWriter) upsertUncleCID(uncle models.UncleModel) {
|
||||
sqw.stmts <- []byte(fmt.Sprintf(uncleInsert, uncle.BlockNumber, uncle.BlockHash, uncle.HeaderID, uncle.ParentHash, uncle.CID,
|
||||
uncle.Reward, uncle.Index))
|
||||
}
|
||||
|
||||
func (sqw *SQLWriter) upsertTransactionCID(transaction models.TxModel) {
|
||||
sqw.stmts <- []byte(fmt.Sprintf(txInsert, transaction.BlockNumber, transaction.HeaderID, transaction.TxHash, transaction.CID, transaction.Dst,
|
||||
transaction.Src, transaction.Index, transaction.Type, transaction.Value))
|
||||
metrics.IndexerMetrics.TransactionsCounter.Inc(1)
|
||||
}
|
||||
|
||||
func (sqw *SQLWriter) upsertReceiptCID(rct *models.ReceiptModel) {
|
||||
sqw.stmts <- []byte(fmt.Sprintf(rctInsert, rct.BlockNumber, rct.HeaderID, rct.TxID, rct.CID, rct.Contract,
|
||||
rct.PostState, rct.PostStatus))
|
||||
metrics.IndexerMetrics.ReceiptsCounter.Inc(1)
|
||||
}
|
||||
|
||||
func (sqw *SQLWriter) upsertLogCID(logs []*models.LogsModel) {
|
||||
for _, l := range logs {
|
||||
sqw.stmts <- []byte(fmt.Sprintf(logInsert, l.BlockNumber, l.HeaderID, l.CID, l.ReceiptID, l.Address, l.Index, l.Topic0,
|
||||
l.Topic1, l.Topic2, l.Topic3))
|
||||
metrics.IndexerMetrics.LogsCounter.Inc(1)
|
||||
}
|
||||
}
|
||||
|
||||
func (sqw *SQLWriter) upsertStateCID(stateNode models.StateNodeModel) {
|
||||
balance := stateNode.Balance
|
||||
if stateNode.Removed {
|
||||
balance = "0"
|
||||
}
|
||||
sqw.stmts <- []byte(fmt.Sprintf(stateInsert, stateNode.BlockNumber, stateNode.HeaderID, stateNode.StateKey, stateNode.CID,
|
||||
stateNode.Removed, true, balance, stateNode.Nonce, stateNode.CodeHash, stateNode.StorageRoot))
|
||||
}
|
||||
|
||||
func (sqw *SQLWriter) upsertStorageCID(storageCID models.StorageNodeModel) {
|
||||
sqw.stmts <- []byte(fmt.Sprintf(storageInsert, storageCID.BlockNumber, storageCID.HeaderID, storageCID.StateKey, storageCID.StorageKey, storageCID.CID,
|
||||
storageCID.Removed, true, storageCID.Value))
|
||||
}
|
||||
|
||||
// LoadWatchedAddresses loads watched addresses from a file
|
||||
func (sqw *SQLWriter) loadWatchedAddresses() ([]common.Address, error) {
|
||||
// load sql statements from watched addresses file
|
||||
stmts, err := loadWatchedAddressesStatements(sqw.watchedAddressesFilePath)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
// extract addresses from the sql statements
|
||||
watchedAddresses := []common.Address{}
|
||||
for _, stmt := range stmts {
|
||||
addressString, err := parseWatchedAddressStatement(stmt)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
watchedAddresses = append(watchedAddresses, common.HexToAddress(addressString))
|
||||
}
|
||||
|
||||
return watchedAddresses, nil
|
||||
}
|
||||
|
||||
// InsertWatchedAddresses inserts the given addresses in a file
|
||||
func (sqw *SQLWriter) insertWatchedAddresses(args []types.WatchAddressArg, currentBlockNumber *big.Int) error {
|
||||
// load sql statements from watched addresses file
|
||||
stmts, err := loadWatchedAddressesStatements(sqw.watchedAddressesFilePath)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
// get already watched addresses
|
||||
var watchedAddresses []string
|
||||
for _, stmt := range stmts {
|
||||
addressString, err := parseWatchedAddressStatement(stmt)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
watchedAddresses = append(watchedAddresses, addressString)
|
||||
}
|
||||
|
||||
// append statements for new addresses to existing statements
|
||||
for _, arg := range args {
|
||||
// ignore if already watched
|
||||
if funk.Contains(watchedAddresses, arg.Address) {
|
||||
continue
|
||||
}
|
||||
|
||||
stmt := fmt.Sprintf(watchedAddressesInsert, arg.Address, arg.CreatedAt, currentBlockNumber.Uint64())
|
||||
stmts = append(stmts, stmt)
|
||||
}
|
||||
|
||||
return dumpWatchedAddressesStatements(sqw.watchedAddressesFilePath, stmts)
|
||||
}
|
||||
|
||||
// RemoveWatchedAddresses removes the given watched addresses from a file
|
||||
func (sqw *SQLWriter) removeWatchedAddresses(args []types.WatchAddressArg) error {
|
||||
// load sql statements from watched addresses file
|
||||
stmts, err := loadWatchedAddressesStatements(sqw.watchedAddressesFilePath)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
// get rid of statements having addresses to be removed
|
||||
var filteredStmts []string
|
||||
for _, stmt := range stmts {
|
||||
addressString, err := parseWatchedAddressStatement(stmt)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
toRemove := funk.Contains(args, func(arg types.WatchAddressArg) bool {
|
||||
return arg.Address == addressString
|
||||
})
|
||||
|
||||
if !toRemove {
|
||||
filteredStmts = append(filteredStmts, stmt)
|
||||
}
|
||||
}
|
||||
|
||||
return dumpWatchedAddressesStatements(sqw.watchedAddressesFilePath, filteredStmts)
|
||||
}
|
||||
|
||||
// SetWatchedAddresses clears and inserts the given addresses in a file
|
||||
func (sqw *SQLWriter) setWatchedAddresses(args []types.WatchAddressArg, currentBlockNumber *big.Int) error {
|
||||
var stmts []string
|
||||
for _, arg := range args {
|
||||
stmt := fmt.Sprintf(watchedAddressesInsert, arg.Address, arg.CreatedAt, currentBlockNumber.Uint64())
|
||||
stmts = append(stmts, stmt)
|
||||
}
|
||||
|
||||
return dumpWatchedAddressesStatements(sqw.watchedAddressesFilePath, stmts)
|
||||
}
|
||||
|
||||
// loadWatchedAddressesStatements loads sql statements from the given file in a string slice
|
||||
func loadWatchedAddressesStatements(filePath string) ([]string, error) {
|
||||
file, err := os.Open(filePath)
|
||||
if err != nil {
|
||||
if errors.Is(err, os.ErrNotExist) {
|
||||
return []string{}, nil
|
||||
}
|
||||
|
||||
return nil, fmt.Errorf("error opening watched addresses file: %v", err)
|
||||
}
|
||||
defer file.Close()
|
||||
|
||||
stmts := []string{}
|
||||
scanner := bufio.NewScanner(file)
|
||||
for scanner.Scan() {
|
||||
stmts = append(stmts, scanner.Text())
|
||||
}
|
||||
|
||||
if err := scanner.Err(); err != nil {
|
||||
return nil, fmt.Errorf("error loading watched addresses: %v", err)
|
||||
}
|
||||
|
||||
return stmts, nil
|
||||
}
|
||||
|
||||
// dumpWatchedAddressesStatements dumps sql statements to the given file
|
||||
func dumpWatchedAddressesStatements(filePath string, stmts []string) error {
|
||||
file, err := os.Create(filePath)
|
||||
if err != nil {
|
||||
return fmt.Errorf("error creating watched addresses file: %v", err)
|
||||
}
|
||||
defer file.Close()
|
||||
|
||||
for _, stmt := range stmts {
|
||||
_, err := file.Write([]byte(stmt + "\n"))
|
||||
if err != nil {
|
||||
return fmt.Errorf("error inserting watched_addresses entry: %v", err)
|
||||
}
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// parseWatchedAddressStatement parses given sql insert statement to extract the address argument
|
||||
func parseWatchedAddressStatement(stmt string) (string, error) {
|
||||
parseResult, err := pg_query.Parse(stmt)
|
||||
if err != nil {
|
||||
return "", fmt.Errorf("error parsing sql stmt: %v", err)
|
||||
}
|
||||
|
||||
// extract address argument from parse output for a SQL statement of form
|
||||
// "INSERT INTO eth_meta.watched_addresses (address, created_at, watched_at)
|
||||
// VALUES ('0xabc', '123', '130') ON CONFLICT (address) DO NOTHING;"
|
||||
addressString := parseResult.Stmts[0].Stmt.GetInsertStmt().
|
||||
SelectStmt.GetSelectStmt().
|
||||
ValuesLists[0].GetList().
|
||||
Items[0].GetAConst().
|
||||
GetVal().
|
||||
GetString_().
|
||||
Str
|
||||
|
||||
return addressString, nil
|
||||
}
|
263
indexer/database/metrics/metrics.go
Normal file
263
indexer/database/metrics/metrics.go
Normal file
@ -0,0 +1,263 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package metrics
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"strings"
|
||||
"time"
|
||||
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
|
||||
"github.com/ethereum/go-ethereum/metrics"
|
||||
)
|
||||
|
||||
const (
|
||||
namespace = "statediff"
|
||||
)
|
||||
|
||||
var (
|
||||
IndexerMetrics = RegisterIndexerMetrics(metrics.DefaultRegistry)
|
||||
DBMetrics = RegisterDBMetrics(metrics.DefaultRegistry)
|
||||
)
|
||||
|
||||
// Build a fully qualified metric name
|
||||
func metricName(subsystem, name string) string {
|
||||
if name == "" {
|
||||
return ""
|
||||
}
|
||||
parts := []string{namespace, name}
|
||||
if subsystem != "" {
|
||||
parts = []string{namespace, subsystem, name}
|
||||
}
|
||||
// Prometheus uses _ but geth metrics uses / and replaces
|
||||
return strings.Join(parts, "/")
|
||||
}
|
||||
|
||||
type IndexerMetricsHandles struct {
|
||||
// The total number of processed BlocksCounter
|
||||
BlocksCounter metrics.Counter
|
||||
// The total number of processed transactions
|
||||
TransactionsCounter metrics.Counter
|
||||
// The total number of processed receipts
|
||||
ReceiptsCounter metrics.Counter
|
||||
// The total number of processed logs
|
||||
LogsCounter metrics.Counter
|
||||
// The total number of access list entries processed
|
||||
AccessListEntriesCounter metrics.Counter
|
||||
// Time spent waiting for free postgres tx
|
||||
FreePostgresTimer metrics.Timer
|
||||
// Postgres transaction commit duration
|
||||
PostgresCommitTimer metrics.Timer
|
||||
// Header processing time
|
||||
HeaderProcessingTimer metrics.Timer
|
||||
// Uncle processing time
|
||||
UncleProcessingTimer metrics.Timer
|
||||
// Tx and receipt processing time
|
||||
TxAndRecProcessingTimer metrics.Timer
|
||||
// State, storage, and code combined processing time
|
||||
StateStoreCodeProcessingTimer metrics.Timer
|
||||
|
||||
// Fine-grained code timers
|
||||
BuildStateDiffWithIntermediateStateNodesTimer metrics.Timer
|
||||
BuildStateDiffWithoutIntermediateStateNodesTimer metrics.Timer
|
||||
CreatedAndUpdatedStateWithIntermediateNodesTimer metrics.Timer
|
||||
DeletedOrUpdatedStateTimer metrics.Timer
|
||||
BuildAccountUpdatesTimer metrics.Timer
|
||||
BuildAccountCreationsTimer metrics.Timer
|
||||
ResolveNodeTimer metrics.Timer
|
||||
SortKeysTimer metrics.Timer
|
||||
FindIntersectionTimer metrics.Timer
|
||||
OutputTimer metrics.Timer
|
||||
IPLDOutputTimer metrics.Timer
|
||||
DifferenceIteratorNextTimer metrics.Timer
|
||||
DifferenceIteratorCounter metrics.Counter
|
||||
DeletedOrUpdatedStorageTimer metrics.Timer
|
||||
CreatedAndUpdatedStorageTimer metrics.Timer
|
||||
BuildStorageNodesIncrementalTimer metrics.Timer
|
||||
BuildStateTrieObjectTimer metrics.Timer
|
||||
BuildStateTrieTimer metrics.Timer
|
||||
BuildStateDiffObjectTimer metrics.Timer
|
||||
WriteStateDiffObjectTimer metrics.Timer
|
||||
CreatedAndUpdatedStateTimer metrics.Timer
|
||||
BuildStorageNodesEventualTimer metrics.Timer
|
||||
BuildStorageNodesFromTrieTimer metrics.Timer
|
||||
BuildRemovedAccountStorageNodesTimer metrics.Timer
|
||||
BuildRemovedStorageNodesFromTrieTimer metrics.Timer
|
||||
IsWatchedAddressTimer metrics.Timer
|
||||
}
|
||||
|
||||
func RegisterIndexerMetrics(reg metrics.Registry) IndexerMetricsHandles {
|
||||
ctx := IndexerMetricsHandles{
|
||||
BlocksCounter: metrics.NewCounter(),
|
||||
TransactionsCounter: metrics.NewCounter(),
|
||||
ReceiptsCounter: metrics.NewCounter(),
|
||||
LogsCounter: metrics.NewCounter(),
|
||||
AccessListEntriesCounter: metrics.NewCounter(),
|
||||
FreePostgresTimer: metrics.NewTimer(),
|
||||
PostgresCommitTimer: metrics.NewTimer(),
|
||||
HeaderProcessingTimer: metrics.NewTimer(),
|
||||
UncleProcessingTimer: metrics.NewTimer(),
|
||||
TxAndRecProcessingTimer: metrics.NewTimer(),
|
||||
StateStoreCodeProcessingTimer: metrics.NewTimer(),
|
||||
BuildStateDiffWithIntermediateStateNodesTimer: metrics.NewTimer(),
|
||||
BuildStateDiffWithoutIntermediateStateNodesTimer: metrics.NewTimer(),
|
||||
CreatedAndUpdatedStateWithIntermediateNodesTimer: metrics.NewTimer(),
|
||||
DeletedOrUpdatedStateTimer: metrics.NewTimer(),
|
||||
BuildAccountUpdatesTimer: metrics.NewTimer(),
|
||||
BuildAccountCreationsTimer: metrics.NewTimer(),
|
||||
ResolveNodeTimer: metrics.NewTimer(),
|
||||
SortKeysTimer: metrics.NewTimer(),
|
||||
FindIntersectionTimer: metrics.NewTimer(),
|
||||
OutputTimer: metrics.NewTimer(),
|
||||
IPLDOutputTimer: metrics.NewTimer(),
|
||||
DifferenceIteratorNextTimer: metrics.NewTimer(),
|
||||
DifferenceIteratorCounter: metrics.NewCounter(),
|
||||
DeletedOrUpdatedStorageTimer: metrics.NewTimer(),
|
||||
CreatedAndUpdatedStorageTimer: metrics.NewTimer(),
|
||||
BuildStorageNodesIncrementalTimer: metrics.NewTimer(),
|
||||
BuildStateTrieObjectTimer: metrics.NewTimer(),
|
||||
BuildStateTrieTimer: metrics.NewTimer(),
|
||||
BuildStateDiffObjectTimer: metrics.NewTimer(),
|
||||
WriteStateDiffObjectTimer: metrics.NewTimer(),
|
||||
CreatedAndUpdatedStateTimer: metrics.NewTimer(),
|
||||
BuildStorageNodesEventualTimer: metrics.NewTimer(),
|
||||
BuildStorageNodesFromTrieTimer: metrics.NewTimer(),
|
||||
BuildRemovedAccountStorageNodesTimer: metrics.NewTimer(),
|
||||
BuildRemovedStorageNodesFromTrieTimer: metrics.NewTimer(),
|
||||
IsWatchedAddressTimer: metrics.NewTimer(),
|
||||
}
|
||||
subsys := "indexer"
|
||||
reg.Register(metricName(subsys, "blocks"), ctx.BlocksCounter)
|
||||
reg.Register(metricName(subsys, "transactions"), ctx.TransactionsCounter)
|
||||
reg.Register(metricName(subsys, "receipts"), ctx.ReceiptsCounter)
|
||||
reg.Register(metricName(subsys, "logs"), ctx.LogsCounter)
|
||||
reg.Register(metricName(subsys, "access_list_entries"), ctx.AccessListEntriesCounter)
|
||||
reg.Register(metricName(subsys, "t_free_postgres"), ctx.FreePostgresTimer)
|
||||
reg.Register(metricName(subsys, "t_postgres_commit"), ctx.PostgresCommitTimer)
|
||||
reg.Register(metricName(subsys, "t_header_processing"), ctx.HeaderProcessingTimer)
|
||||
reg.Register(metricName(subsys, "t_uncle_processing"), ctx.UncleProcessingTimer)
|
||||
reg.Register(metricName(subsys, "t_tx_receipt_processing"), ctx.TxAndRecProcessingTimer)
|
||||
reg.Register(metricName(subsys, "t_state_store_code_processing"), ctx.StateStoreCodeProcessingTimer)
|
||||
reg.Register(metricName(subsys, "t_build_statediff_with_intermediate_state_nodes"), ctx.BuildStateDiffWithIntermediateStateNodesTimer)
|
||||
reg.Register(metricName(subsys, "t_build_statediff_without_intermediate_state_nodes"), ctx.BuildStateDiffWithoutIntermediateStateNodesTimer)
|
||||
reg.Register(metricName(subsys, "t_created_and_update_state_with_intermediate_nodes"), ctx.CreatedAndUpdatedStateWithIntermediateNodesTimer)
|
||||
reg.Register(metricName(subsys, "t_deleted_or_updated_state"), ctx.DeletedOrUpdatedStateTimer)
|
||||
reg.Register(metricName(subsys, "t_build_account_updates"), ctx.BuildAccountUpdatesTimer)
|
||||
reg.Register(metricName(subsys, "t_build_account_creations"), ctx.BuildAccountCreationsTimer)
|
||||
reg.Register(metricName(subsys, "t_resolve_node"), ctx.ResolveNodeTimer)
|
||||
reg.Register(metricName(subsys, "t_sort_keys"), ctx.SortKeysTimer)
|
||||
reg.Register(metricName(subsys, "t_find_intersection"), ctx.FindIntersectionTimer)
|
||||
reg.Register(metricName(subsys, "t_output_fn"), ctx.OutputTimer)
|
||||
reg.Register(metricName(subsys, "t_ipld_output_fn"), ctx.IPLDOutputTimer)
|
||||
reg.Register(metricName(subsys, "t_difference_iterator_next"), ctx.DifferenceIteratorNextTimer)
|
||||
reg.Register(metricName(subsys, "difference_iterator_counter"), ctx.DifferenceIteratorCounter)
|
||||
reg.Register(metricName(subsys, "t_created_and_updated_storage"), ctx.CreatedAndUpdatedStorageTimer)
|
||||
reg.Register(metricName(subsys, "t_deleted_or_updated_storage"), ctx.DeletedOrUpdatedStorageTimer)
|
||||
reg.Register(metricName(subsys, "t_build_storage_nodes_incremental"), ctx.BuildStorageNodesIncrementalTimer)
|
||||
reg.Register(metricName(subsys, "t_build_state_trie_object"), ctx.BuildStateTrieObjectTimer)
|
||||
reg.Register(metricName(subsys, "t_build_state_trie"), ctx.BuildStateTrieTimer)
|
||||
reg.Register(metricName(subsys, "t_build_statediff_object"), ctx.BuildStateDiffObjectTimer)
|
||||
reg.Register(metricName(subsys, "t_write_statediff_object"), ctx.WriteStateDiffObjectTimer)
|
||||
reg.Register(metricName(subsys, "t_created_and_updated_state"), ctx.CreatedAndUpdatedStateTimer)
|
||||
reg.Register(metricName(subsys, "t_build_storage_nodes_eventual"), ctx.BuildStorageNodesEventualTimer)
|
||||
reg.Register(metricName(subsys, "t_build_storage_nodes_from_trie"), ctx.BuildStorageNodesFromTrieTimer)
|
||||
reg.Register(metricName(subsys, "t_build_removed_accounts_storage_nodes"), ctx.BuildRemovedAccountStorageNodesTimer)
|
||||
reg.Register(metricName(subsys, "t_build_removed_storage_nodes_from_trie"), ctx.BuildRemovedStorageNodesFromTrieTimer)
|
||||
reg.Register(metricName(subsys, "t_is_watched_address"), ctx.IsWatchedAddressTimer)
|
||||
|
||||
log.Debug("Registering statediff indexer metrics.")
|
||||
return ctx
|
||||
}
|
||||
|
||||
type dbMetricsHandles struct {
|
||||
// Maximum number of open connections to the sql
|
||||
maxOpen metrics.Gauge
|
||||
// The number of established connections both in use and idle
|
||||
open metrics.Gauge
|
||||
// The number of connections currently in use
|
||||
inUse metrics.Gauge
|
||||
// The number of idle connections
|
||||
idle metrics.Gauge
|
||||
// The total number of connections waited for
|
||||
waitedFor metrics.Counter
|
||||
// The total time blocked waiting for a new connection
|
||||
blockedMilliseconds metrics.Counter
|
||||
// The total number of connections closed due to SetMaxIdleConns
|
||||
closedMaxIdle metrics.Counter
|
||||
// The total number of connections closed due to SetConnMaxLifetime
|
||||
closedMaxLifetime metrics.Counter
|
||||
}
|
||||
|
||||
func RegisterDBMetrics(reg metrics.Registry) dbMetricsHandles {
|
||||
ctx := dbMetricsHandles{
|
||||
maxOpen: metrics.NewGauge(),
|
||||
open: metrics.NewGauge(),
|
||||
inUse: metrics.NewGauge(),
|
||||
idle: metrics.NewGauge(),
|
||||
waitedFor: metrics.NewCounter(),
|
||||
blockedMilliseconds: metrics.NewCounter(),
|
||||
closedMaxIdle: metrics.NewCounter(),
|
||||
closedMaxLifetime: metrics.NewCounter(),
|
||||
}
|
||||
subsys := "connections"
|
||||
reg.Register(metricName(subsys, "max_open"), ctx.maxOpen)
|
||||
reg.Register(metricName(subsys, "open"), ctx.open)
|
||||
reg.Register(metricName(subsys, "in_use"), ctx.inUse)
|
||||
reg.Register(metricName(subsys, "idle"), ctx.idle)
|
||||
reg.Register(metricName(subsys, "waited_for"), ctx.waitedFor)
|
||||
reg.Register(metricName(subsys, "blocked_milliseconds"), ctx.blockedMilliseconds)
|
||||
reg.Register(metricName(subsys, "closed_max_idle"), ctx.closedMaxIdle)
|
||||
reg.Register(metricName(subsys, "closed_max_lifetime"), ctx.closedMaxLifetime)
|
||||
|
||||
log.Debug("Registering statediff DB metrics.")
|
||||
return ctx
|
||||
}
|
||||
|
||||
// DbStats interface to accommodate different concrete sql stats types
|
||||
type DbStats interface {
|
||||
MaxOpen() int64
|
||||
Open() int64
|
||||
InUse() int64
|
||||
Idle() int64
|
||||
WaitCount() int64
|
||||
WaitDuration() time.Duration
|
||||
MaxIdleClosed() int64
|
||||
MaxLifetimeClosed() int64
|
||||
}
|
||||
|
||||
func (met *dbMetricsHandles) Update(stats DbStats) {
|
||||
met.maxOpen.Update(stats.MaxOpen())
|
||||
met.open.Update(stats.Open())
|
||||
met.inUse.Update(stats.InUse())
|
||||
met.idle.Update(stats.Idle())
|
||||
met.waitedFor.Inc(stats.WaitCount())
|
||||
met.blockedMilliseconds.Inc(stats.WaitDuration().Milliseconds())
|
||||
met.closedMaxIdle.Inc(stats.MaxIdleClosed())
|
||||
met.closedMaxLifetime.Inc(stats.MaxLifetimeClosed())
|
||||
}
|
||||
|
||||
func ReportAndUpdateDuration(msg string, start time.Time, logger log.Logger, timer metrics.Timer) {
|
||||
since := UpdateDuration(start, timer)
|
||||
logger.Trace(fmt.Sprintf("%s duration=%dms", msg, since.Milliseconds()))
|
||||
}
|
||||
|
||||
func UpdateDuration(start time.Time, timer metrics.Timer) time.Duration {
|
||||
since := time.Since(start)
|
||||
timer.Update(since)
|
||||
return since
|
||||
}
|
126
indexer/database/sql/batch_tx.go
Normal file
126
indexer/database/sql/batch_tx.go
Normal file
@ -0,0 +1,126 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package sql
|
||||
|
||||
import (
|
||||
"context"
|
||||
"sync"
|
||||
"sync/atomic"
|
||||
|
||||
"github.com/lib/pq"
|
||||
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/ipld"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/models"
|
||||
)
|
||||
|
||||
const startingCacheCapacity = 1024 * 24
|
||||
|
||||
// BatchTx wraps a sql tx with the state necessary for building the tx concurrently during trie difference iteration
|
||||
type BatchTx struct {
|
||||
BlockNumber string
|
||||
ctx context.Context
|
||||
dbtx Tx
|
||||
stm string
|
||||
quit chan struct{}
|
||||
iplds chan models.IPLDModel
|
||||
ipldCache models.IPLDBatch
|
||||
removedCacheFlag *uint32
|
||||
// Tracks expected cache size and ensures cache is caught up before flush
|
||||
cacheWg sync.WaitGroup
|
||||
|
||||
submit func(blockTx *BatchTx, err error) error
|
||||
}
|
||||
|
||||
// Submit satisfies indexer.AtomicTx
|
||||
func (tx *BatchTx) Submit(err error) error {
|
||||
return tx.submit(tx, err)
|
||||
}
|
||||
|
||||
func (tx *BatchTx) flush() error {
|
||||
tx.cacheWg.Wait()
|
||||
_, err := tx.dbtx.Exec(tx.ctx, tx.stm, pq.Array(tx.ipldCache.BlockNumbers), pq.Array(tx.ipldCache.Keys),
|
||||
pq.Array(tx.ipldCache.Values))
|
||||
if err != nil {
|
||||
log.Debug(insertError{"ipld.blocks", err, tx.stm,
|
||||
struct {
|
||||
blockNumbers []string
|
||||
keys []string
|
||||
values [][]byte
|
||||
}{
|
||||
tx.ipldCache.BlockNumbers,
|
||||
tx.ipldCache.Keys,
|
||||
tx.ipldCache.Values,
|
||||
}}.Error())
|
||||
return insertError{"ipld.blocks", err, tx.stm, "too many arguments; use debug mode for full list"}
|
||||
}
|
||||
tx.ipldCache = models.IPLDBatch{}
|
||||
return nil
|
||||
}
|
||||
|
||||
// run in background goroutine to synchronize concurrent appends to the ipldCache
|
||||
func (tx *BatchTx) cache() {
|
||||
for {
|
||||
select {
|
||||
case i := <-tx.iplds:
|
||||
tx.ipldCache.BlockNumbers = append(tx.ipldCache.BlockNumbers, i.BlockNumber)
|
||||
tx.ipldCache.Keys = append(tx.ipldCache.Keys, i.Key)
|
||||
tx.ipldCache.Values = append(tx.ipldCache.Values, i.Data)
|
||||
tx.cacheWg.Done()
|
||||
case <-tx.quit:
|
||||
tx.ipldCache = models.IPLDBatch{}
|
||||
return
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func (tx *BatchTx) cacheDirect(key string, value []byte) {
|
||||
tx.cacheWg.Add(1)
|
||||
tx.iplds <- models.IPLDModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
Key: key,
|
||||
Data: value,
|
||||
}
|
||||
}
|
||||
|
||||
func (tx *BatchTx) cacheIPLD(i ipld.IPLD) {
|
||||
tx.cacheWg.Add(1)
|
||||
tx.iplds <- models.IPLDModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
Key: i.Cid().String(),
|
||||
Data: i.RawData(),
|
||||
}
|
||||
}
|
||||
|
||||
func (tx *BatchTx) cacheRemoved(key string, value []byte) {
|
||||
if atomic.LoadUint32(tx.removedCacheFlag) == 0 {
|
||||
atomic.StoreUint32(tx.removedCacheFlag, 1)
|
||||
tx.cacheWg.Add(1)
|
||||
tx.iplds <- models.IPLDModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
Key: key,
|
||||
Data: value,
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// rollback sql transaction and log any error
|
||||
func rollback(ctx context.Context, tx Tx) {
|
||||
if err := tx.Rollback(ctx); err != nil {
|
||||
log.Error(err.Error())
|
||||
}
|
||||
}
|
583
indexer/database/sql/indexer.go
Normal file
583
indexer/database/sql/indexer.go
Normal file
@ -0,0 +1,583 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
// Package sql provides an interface for pushing and indexing IPLD objects into a sql database
|
||||
// Metrics for reporting processing and connection stats are defined in ./metrics.go
|
||||
|
||||
package sql
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"context"
|
||||
"fmt"
|
||||
"math/big"
|
||||
"time"
|
||||
|
||||
"github.com/multiformats/go-multihash"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/crypto"
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
"github.com/ethereum/go-ethereum/metrics"
|
||||
"github.com/ethereum/go-ethereum/params"
|
||||
"github.com/ethereum/go-ethereum/rlp"
|
||||
metrics2 "github.com/ethereum/go-ethereum/statediff/indexer/database/metrics"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/interfaces"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/ipld"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/models"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/shared"
|
||||
sdtypes "github.com/ethereum/go-ethereum/statediff/types"
|
||||
)
|
||||
|
||||
var _ interfaces.StateDiffIndexer = &StateDiffIndexer{}
|
||||
|
||||
// StateDiffIndexer satisfies the indexer.StateDiffIndexer interface for ethereum statediff objects on top of an SQL sql
|
||||
type StateDiffIndexer struct {
|
||||
ctx context.Context
|
||||
chainConfig *params.ChainConfig
|
||||
dbWriter *Writer
|
||||
}
|
||||
|
||||
// NewStateDiffIndexer creates a sql implementation of interfaces.StateDiffIndexer
|
||||
func NewStateDiffIndexer(ctx context.Context, chainConfig *params.ChainConfig, db Database) (*StateDiffIndexer, error) {
|
||||
return &StateDiffIndexer{
|
||||
ctx: ctx,
|
||||
chainConfig: chainConfig,
|
||||
dbWriter: NewWriter(db),
|
||||
}, nil
|
||||
}
|
||||
|
||||
// ReportDBMetrics is a reporting function to run as goroutine
|
||||
func (sdi *StateDiffIndexer) ReportDBMetrics(delay time.Duration, quit <-chan bool) {
|
||||
if !metrics.Enabled {
|
||||
return
|
||||
}
|
||||
ticker := time.NewTicker(delay)
|
||||
go func() {
|
||||
for {
|
||||
select {
|
||||
case <-ticker.C:
|
||||
metrics2.DBMetrics.Update(sdi.dbWriter.db.Stats())
|
||||
case <-quit:
|
||||
ticker.Stop()
|
||||
return
|
||||
}
|
||||
}
|
||||
}()
|
||||
}
|
||||
|
||||
// PushBlock pushes and indexes block data in sql, except state & storage nodes (includes header, uncles, transactions & receipts)
|
||||
// Returns an initiated DB transaction which must be Closed via defer to commit or rollback
|
||||
func (sdi *StateDiffIndexer) PushBlock(block *types.Block, receipts types.Receipts, totalDifficulty *big.Int) (interfaces.Batch, error) {
|
||||
start, t := time.Now(), time.Now()
|
||||
blockHash := block.Hash()
|
||||
blockHashStr := blockHash.String()
|
||||
height := block.NumberU64()
|
||||
traceMsg := fmt.Sprintf("indexer stats for statediff at %d with hash %s:\r\n", height, blockHashStr)
|
||||
transactions := block.Transactions()
|
||||
// Derive any missing fields
|
||||
if err := receipts.DeriveFields(sdi.chainConfig, blockHash, height, block.BaseFee(), transactions); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
// Generate the block iplds
|
||||
headerNode, txNodes, rctNodes, logNodes, err := ipld.FromBlockAndReceipts(block, receipts)
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("error creating IPLD nodes from block and receipts: %v", err)
|
||||
}
|
||||
|
||||
if len(txNodes) != len(rctNodes) {
|
||||
return nil, fmt.Errorf("expected number of transactions (%d), receipts (%d)", len(txNodes), len(rctNodes))
|
||||
}
|
||||
|
||||
// Calculate reward
|
||||
var reward *big.Int
|
||||
// in PoA networks block reward is 0
|
||||
if sdi.chainConfig.Clique != nil {
|
||||
reward = big.NewInt(0)
|
||||
} else {
|
||||
reward = shared.CalcEthBlockReward(block.Header(), block.Uncles(), block.Transactions(), receipts)
|
||||
}
|
||||
t = time.Now()
|
||||
|
||||
// Begin new DB tx for everything
|
||||
tx := NewDelayedTx(sdi.dbWriter.db)
|
||||
defer func() {
|
||||
if p := recover(); p != nil {
|
||||
rollback(sdi.ctx, tx)
|
||||
panic(p)
|
||||
} else if err != nil {
|
||||
rollback(sdi.ctx, tx)
|
||||
}
|
||||
}()
|
||||
blockTx := &BatchTx{
|
||||
removedCacheFlag: new(uint32),
|
||||
ctx: sdi.ctx,
|
||||
BlockNumber: block.Number().String(),
|
||||
stm: sdi.dbWriter.db.InsertIPLDsStm(),
|
||||
iplds: make(chan models.IPLDModel),
|
||||
quit: make(chan struct{}),
|
||||
ipldCache: models.IPLDBatch{
|
||||
BlockNumbers: make([]string, 0, startingCacheCapacity),
|
||||
Keys: make([]string, 0, startingCacheCapacity),
|
||||
Values: make([][]byte, 0, startingCacheCapacity),
|
||||
},
|
||||
dbtx: tx,
|
||||
// handle transaction commit or rollback for any return case
|
||||
submit: func(self *BatchTx, err error) error {
|
||||
defer func() {
|
||||
close(self.quit)
|
||||
close(self.iplds)
|
||||
}()
|
||||
if p := recover(); p != nil {
|
||||
log.Info("panic detected before tx submission, rolling back the tx", "panic", p)
|
||||
rollback(sdi.ctx, tx)
|
||||
panic(p)
|
||||
} else if err != nil {
|
||||
log.Info("error detected before tx submission, rolling back the tx", "error", err)
|
||||
rollback(sdi.ctx, tx)
|
||||
} else {
|
||||
tDiff := time.Since(t)
|
||||
metrics2.IndexerMetrics.StateStoreCodeProcessingTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("state, storage, and code storage processing time: %s\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
if err := self.flush(); err != nil {
|
||||
rollback(sdi.ctx, tx)
|
||||
traceMsg += fmt.Sprintf(" TOTAL PROCESSING DURATION: %s\r\n", time.Since(start).String())
|
||||
log.Debug(traceMsg)
|
||||
return err
|
||||
}
|
||||
err = tx.Commit(sdi.ctx)
|
||||
tDiff = time.Since(t)
|
||||
metrics2.IndexerMetrics.PostgresCommitTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("postgres transaction commit duration: %s\r\n", tDiff.String())
|
||||
}
|
||||
traceMsg += fmt.Sprintf(" TOTAL PROCESSING DURATION: %s\r\n", time.Since(start).String())
|
||||
log.Debug(traceMsg)
|
||||
return err
|
||||
},
|
||||
}
|
||||
go blockTx.cache()
|
||||
|
||||
tDiff := time.Since(t)
|
||||
metrics2.IndexerMetrics.FreePostgresTimer.Update(tDiff)
|
||||
|
||||
traceMsg += fmt.Sprintf("time spent waiting for free postgres tx: %s:\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
|
||||
// Publish and index header, collect headerID
|
||||
var headerID string
|
||||
headerID, err = sdi.processHeader(blockTx, block.Header(), headerNode, reward, totalDifficulty)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
tDiff = time.Since(t)
|
||||
metrics2.IndexerMetrics.HeaderProcessingTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("header processing time: %s\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
// Publish and index uncles
|
||||
err = sdi.processUncles(blockTx, headerID, block.Number(), block.UncleHash(), block.Uncles())
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
tDiff = time.Since(t)
|
||||
metrics2.IndexerMetrics.UncleProcessingTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("uncle processing time: %s\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
// Publish and index receipts and txs
|
||||
err = sdi.processReceiptsAndTxs(blockTx, processArgs{
|
||||
headerID: headerID,
|
||||
blockNumber: block.Number(),
|
||||
receipts: receipts,
|
||||
txs: transactions,
|
||||
rctNodes: rctNodes,
|
||||
txNodes: txNodes,
|
||||
logNodes: logNodes,
|
||||
})
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
tDiff = time.Since(t)
|
||||
metrics2.IndexerMetrics.TxAndRecProcessingTimer.Update(tDiff)
|
||||
traceMsg += fmt.Sprintf("tx and receipt processing time: %s\r\n", tDiff.String())
|
||||
t = time.Now()
|
||||
|
||||
return blockTx, err
|
||||
}
|
||||
|
||||
// processHeader publishes and indexes a header IPLD in Postgres
|
||||
// it returns the headerID
|
||||
func (sdi *StateDiffIndexer) processHeader(tx *BatchTx, header *types.Header, headerNode ipld.IPLD, reward, td *big.Int) (string, error) {
|
||||
tx.cacheIPLD(headerNode)
|
||||
|
||||
var baseFee *string
|
||||
if header.BaseFee != nil {
|
||||
baseFee = new(string)
|
||||
*baseFee = header.BaseFee.String()
|
||||
}
|
||||
headerID := header.Hash().String()
|
||||
// index header
|
||||
return headerID, sdi.dbWriter.upsertHeaderCID(tx.dbtx, models.HeaderModel{
|
||||
CID: headerNode.Cid().String(),
|
||||
ParentHash: header.ParentHash.String(),
|
||||
BlockNumber: header.Number.String(),
|
||||
BlockHash: headerID,
|
||||
TotalDifficulty: td.String(),
|
||||
Reward: reward.String(),
|
||||
Bloom: header.Bloom.Bytes(),
|
||||
StateRoot: header.Root.String(),
|
||||
RctRoot: header.ReceiptHash.String(),
|
||||
TxRoot: header.TxHash.String(),
|
||||
UnclesHash: header.UncleHash.String(),
|
||||
Timestamp: header.Time,
|
||||
Coinbase: header.Coinbase.String(),
|
||||
})
|
||||
}
|
||||
|
||||
// processUncles publishes and indexes uncle IPLDs in Postgres
|
||||
func (sdi *StateDiffIndexer) processUncles(tx *BatchTx, headerID string, blockNumber *big.Int, unclesHash common.Hash, uncles []*types.Header) error {
|
||||
// publish and index uncles
|
||||
uncleEncoding, err := rlp.EncodeToBytes(uncles)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
preparedHash := crypto.Keccak256Hash(uncleEncoding)
|
||||
if !bytes.Equal(preparedHash.Bytes(), unclesHash.Bytes()) {
|
||||
return fmt.Errorf("derived uncles hash (%s) does not match the hash in the header (%s)", preparedHash.Hex(), unclesHash.Hex())
|
||||
}
|
||||
unclesCID, err := ipld.RawdataToCid(ipld.MEthHeaderList, uncleEncoding, multihash.KECCAK_256)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
tx.cacheDirect(unclesCID.String(), uncleEncoding)
|
||||
for i, uncle := range uncles {
|
||||
var uncleReward *big.Int
|
||||
// in PoA networks uncle reward is 0
|
||||
if sdi.chainConfig.Clique != nil {
|
||||
uncleReward = big.NewInt(0)
|
||||
} else {
|
||||
uncleReward = shared.CalcUncleMinerReward(blockNumber.Uint64(), uncle.Number.Uint64())
|
||||
}
|
||||
uncle := models.UncleModel{
|
||||
BlockNumber: blockNumber.String(),
|
||||
HeaderID: headerID,
|
||||
CID: unclesCID.String(),
|
||||
ParentHash: uncle.ParentHash.String(),
|
||||
BlockHash: uncle.Hash().String(),
|
||||
Reward: uncleReward.String(),
|
||||
Index: int64(i),
|
||||
}
|
||||
if err := sdi.dbWriter.upsertUncleCID(tx.dbtx, uncle); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
// processArgs bundles arguments to processReceiptsAndTxs
|
||||
type processArgs struct {
|
||||
headerID string
|
||||
blockNumber *big.Int
|
||||
receipts types.Receipts
|
||||
txs types.Transactions
|
||||
rctNodes []*ipld.EthReceipt
|
||||
txNodes []*ipld.EthTx
|
||||
logNodes [][]*ipld.EthLog
|
||||
}
|
||||
|
||||
// processReceiptsAndTxs publishes and indexes receipt and transaction IPLDs in Postgres
|
||||
func (sdi *StateDiffIndexer) processReceiptsAndTxs(tx *BatchTx, args processArgs) error {
|
||||
// Process receipts and txs
|
||||
signer := types.MakeSigner(sdi.chainConfig, args.blockNumber)
|
||||
for i, receipt := range args.receipts {
|
||||
txNode := args.txNodes[i]
|
||||
tx.cacheIPLD(txNode)
|
||||
tx.cacheIPLD(args.rctNodes[i])
|
||||
|
||||
// index tx
|
||||
trx := args.txs[i]
|
||||
txID := trx.Hash().String()
|
||||
|
||||
var val string
|
||||
if trx.Value() != nil {
|
||||
val = trx.Value().String()
|
||||
}
|
||||
|
||||
// derive sender for the tx that corresponds with this receipt
|
||||
from, err := types.Sender(signer, trx)
|
||||
if err != nil {
|
||||
return fmt.Errorf("error deriving tx sender: %v", err)
|
||||
}
|
||||
txModel := models.TxModel{
|
||||
BlockNumber: args.blockNumber.String(),
|
||||
HeaderID: args.headerID,
|
||||
Dst: shared.HandleZeroAddrPointer(trx.To()),
|
||||
Src: shared.HandleZeroAddr(from),
|
||||
TxHash: txID,
|
||||
Index: int64(i),
|
||||
CID: txNode.Cid().String(),
|
||||
Type: trx.Type(),
|
||||
Value: val,
|
||||
}
|
||||
if err := sdi.dbWriter.upsertTransactionCID(tx.dbtx, txModel); err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
// this is the contract address if this receipt is for a contract creation tx
|
||||
contract := shared.HandleZeroAddr(receipt.ContractAddress)
|
||||
|
||||
rctModel := &models.ReceiptModel{
|
||||
BlockNumber: args.blockNumber.String(),
|
||||
HeaderID: args.headerID,
|
||||
TxID: txID,
|
||||
Contract: contract,
|
||||
CID: args.rctNodes[i].Cid().String(),
|
||||
}
|
||||
if len(receipt.PostState) == 0 {
|
||||
rctModel.PostStatus = receipt.Status
|
||||
} else {
|
||||
rctModel.PostState = common.BytesToHash(receipt.PostState).String()
|
||||
}
|
||||
|
||||
if err := sdi.dbWriter.upsertReceiptCID(tx.dbtx, rctModel); err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
// index logs
|
||||
logDataSet := make([]*models.LogsModel, len(receipt.Logs))
|
||||
for idx, l := range receipt.Logs {
|
||||
tx.cacheIPLD(args.logNodes[i][idx])
|
||||
topicSet := make([]string, 4)
|
||||
for ti, topic := range l.Topics {
|
||||
topicSet[ti] = topic.Hex()
|
||||
}
|
||||
|
||||
logDataSet[idx] = &models.LogsModel{
|
||||
BlockNumber: args.blockNumber.String(),
|
||||
HeaderID: args.headerID,
|
||||
ReceiptID: txID,
|
||||
Address: l.Address.String(),
|
||||
Index: int64(l.Index),
|
||||
CID: args.logNodes[i][idx].Cid().String(),
|
||||
Topic0: topicSet[0],
|
||||
Topic1: topicSet[1],
|
||||
Topic2: topicSet[2],
|
||||
Topic3: topicSet[3],
|
||||
}
|
||||
}
|
||||
|
||||
if err := sdi.dbWriter.upsertLogCID(tx.dbtx, logDataSet); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// PushStateNode publishes and indexes a state diff node object (including any child storage nodes) in the IPLD sql
|
||||
func (sdi *StateDiffIndexer) PushStateNode(batch interfaces.Batch, stateNode sdtypes.StateLeafNode, headerID string) error {
|
||||
tx, ok := batch.(*BatchTx)
|
||||
if !ok {
|
||||
return fmt.Errorf("sql: batch is expected to be of type %T, got %T", &BatchTx{}, batch)
|
||||
}
|
||||
// publish the state node
|
||||
var stateModel models.StateNodeModel
|
||||
if stateNode.Removed {
|
||||
tx.cacheRemoved(shared.RemovedNodeStateCID, []byte{})
|
||||
stateModel = models.StateNodeModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
HeaderID: headerID,
|
||||
StateKey: common.BytesToHash(stateNode.AccountWrapper.LeafKey).String(),
|
||||
CID: shared.RemovedNodeStateCID,
|
||||
Removed: true,
|
||||
}
|
||||
} else {
|
||||
stateModel = models.StateNodeModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
HeaderID: headerID,
|
||||
StateKey: common.BytesToHash(stateNode.AccountWrapper.LeafKey).String(),
|
||||
CID: stateNode.AccountWrapper.CID,
|
||||
Removed: false,
|
||||
Balance: stateNode.AccountWrapper.Account.Balance.String(),
|
||||
Nonce: stateNode.AccountWrapper.Account.Nonce,
|
||||
CodeHash: common.BytesToHash(stateNode.AccountWrapper.Account.CodeHash).String(),
|
||||
StorageRoot: stateNode.AccountWrapper.Account.Root.String(),
|
||||
}
|
||||
}
|
||||
|
||||
// index the state node
|
||||
if err := sdi.dbWriter.upsertStateCID(tx.dbtx, stateModel); err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
// if there are any storage nodes associated with this node, publish and index them
|
||||
for _, storageNode := range stateNode.StorageDiff {
|
||||
if storageNode.Removed {
|
||||
tx.cacheRemoved(shared.RemovedNodeStorageCID, []byte{})
|
||||
storageModel := models.StorageNodeModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
HeaderID: headerID,
|
||||
StateKey: common.BytesToHash(stateNode.AccountWrapper.LeafKey).String(),
|
||||
StorageKey: common.BytesToHash(storageNode.LeafKey).String(),
|
||||
CID: shared.RemovedNodeStorageCID,
|
||||
Removed: true,
|
||||
Value: []byte{},
|
||||
}
|
||||
if err := sdi.dbWriter.upsertStorageCID(tx.dbtx, storageModel); err != nil {
|
||||
return err
|
||||
}
|
||||
continue
|
||||
}
|
||||
storageModel := models.StorageNodeModel{
|
||||
BlockNumber: tx.BlockNumber,
|
||||
HeaderID: headerID,
|
||||
StateKey: common.BytesToHash(stateNode.AccountWrapper.LeafKey).String(),
|
||||
StorageKey: common.BytesToHash(storageNode.LeafKey).String(),
|
||||
CID: storageNode.CID,
|
||||
Removed: false,
|
||||
Value: storageNode.Value,
|
||||
}
|
||||
if err := sdi.dbWriter.upsertStorageCID(tx.dbtx, storageModel); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// PushIPLD publishes iplds to ipld.blocks
|
||||
func (sdi *StateDiffIndexer) PushIPLD(batch interfaces.Batch, ipld sdtypes.IPLD) error {
|
||||
tx, ok := batch.(*BatchTx)
|
||||
if !ok {
|
||||
return fmt.Errorf("sql: batch is expected to be of type %T, got %T", &BatchTx{}, batch)
|
||||
}
|
||||
tx.cacheDirect(ipld.CID, ipld.Content)
|
||||
return nil
|
||||
}
|
||||
|
||||
// Close satisfies io.Closer
|
||||
func (sdi *StateDiffIndexer) Close() error {
|
||||
return sdi.dbWriter.Close()
|
||||
}
|
||||
|
||||
// Update the known gaps table with the gap information.
|
||||
|
||||
// LoadWatchedAddresses reads watched addresses from the database
|
||||
func (sdi *StateDiffIndexer) LoadWatchedAddresses() ([]common.Address, error) {
|
||||
addressStrings := make([]string, 0)
|
||||
pgStr := "SELECT address FROM eth_meta.watched_addresses"
|
||||
err := sdi.dbWriter.db.Select(sdi.ctx, &addressStrings, pgStr)
|
||||
if err != nil {
|
||||
return nil, fmt.Errorf("error loading watched addresses: %v", err)
|
||||
}
|
||||
|
||||
watchedAddresses := []common.Address{}
|
||||
for _, addressString := range addressStrings {
|
||||
watchedAddresses = append(watchedAddresses, common.HexToAddress(addressString))
|
||||
}
|
||||
|
||||
return watchedAddresses, nil
|
||||
}
|
||||
|
||||
// InsertWatchedAddresses inserts the given addresses in the database
|
||||
func (sdi *StateDiffIndexer) InsertWatchedAddresses(args []sdtypes.WatchAddressArg, currentBlockNumber *big.Int) (err error) {
|
||||
tx := NewDelayedTx(sdi.dbWriter.db)
|
||||
defer func() {
|
||||
if p := recover(); p != nil {
|
||||
rollback(sdi.ctx, tx)
|
||||
panic(p)
|
||||
} else if err != nil {
|
||||
rollback(sdi.ctx, tx)
|
||||
} else {
|
||||
err = tx.Commit(sdi.ctx)
|
||||
}
|
||||
}()
|
||||
|
||||
for _, arg := range args {
|
||||
_, err = tx.Exec(sdi.ctx, `INSERT INTO eth_meta.watched_addresses (address, created_at, watched_at) VALUES ($1, $2, $3) ON CONFLICT (address) DO NOTHING`,
|
||||
arg.Address, arg.CreatedAt, currentBlockNumber.Uint64())
|
||||
if err != nil {
|
||||
return fmt.Errorf("error inserting watched_addresses entry: %v", err)
|
||||
}
|
||||
}
|
||||
|
||||
return err
|
||||
}
|
||||
|
||||
// RemoveWatchedAddresses removes the given watched addresses from the database
|
||||
func (sdi *StateDiffIndexer) RemoveWatchedAddresses(args []sdtypes.WatchAddressArg) (err error) {
|
||||
tx := NewDelayedTx(sdi.dbWriter.db)
|
||||
defer func() {
|
||||
if p := recover(); p != nil {
|
||||
rollback(sdi.ctx, tx)
|
||||
panic(p)
|
||||
} else if err != nil {
|
||||
rollback(sdi.ctx, tx)
|
||||
} else {
|
||||
err = tx.Commit(sdi.ctx)
|
||||
}
|
||||
}()
|
||||
|
||||
for _, arg := range args {
|
||||
_, err = tx.Exec(sdi.ctx, `DELETE FROM eth_meta.watched_addresses WHERE address = $1`, arg.Address)
|
||||
if err != nil {
|
||||
return fmt.Errorf("error removing watched_addresses entry: %v", err)
|
||||
}
|
||||
}
|
||||
|
||||
return err
|
||||
}
|
||||
|
||||
// SetWatchedAddresses clears and inserts the given addresses in the database
|
||||
func (sdi *StateDiffIndexer) SetWatchedAddresses(args []sdtypes.WatchAddressArg, currentBlockNumber *big.Int) (err error) {
|
||||
tx := NewDelayedTx(sdi.dbWriter.db)
|
||||
defer func() {
|
||||
if p := recover(); p != nil {
|
||||
rollback(sdi.ctx, tx)
|
||||
panic(p)
|
||||
} else if err != nil {
|
||||
rollback(sdi.ctx, tx)
|
||||
} else {
|
||||
err = tx.Commit(sdi.ctx)
|
||||
}
|
||||
}()
|
||||
|
||||
_, err = tx.Exec(sdi.ctx, `DELETE FROM eth_meta.watched_addresses`)
|
||||
if err != nil {
|
||||
return fmt.Errorf("error setting watched_addresses table: %v", err)
|
||||
}
|
||||
|
||||
for _, arg := range args {
|
||||
_, err = tx.Exec(sdi.ctx, `INSERT INTO eth_meta.watched_addresses (address, created_at, watched_at) VALUES ($1, $2, $3) ON CONFLICT (address) DO NOTHING`,
|
||||
arg.Address, arg.CreatedAt, currentBlockNumber.Uint64())
|
||||
if err != nil {
|
||||
return fmt.Errorf("error setting watched_addresses table: %v", err)
|
||||
}
|
||||
}
|
||||
|
||||
return err
|
||||
}
|
||||
|
||||
// ClearWatchedAddresses clears all the watched addresses from the database
|
||||
func (sdi *StateDiffIndexer) ClearWatchedAddresses() error {
|
||||
_, err := sdi.dbWriter.db.Exec(sdi.ctx, `DELETE FROM eth_meta.watched_addresses`)
|
||||
if err != nil {
|
||||
return fmt.Errorf("error clearing watched_addresses table: %v", err)
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
28
indexer/database/sql/indexer_shared_test.go
Normal file
28
indexer/database/sql/indexer_shared_test.go
Normal file
@ -0,0 +1,28 @@
|
||||
package sql_test
|
||||
|
||||
import (
|
||||
"testing"
|
||||
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/interfaces"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test_helpers"
|
||||
)
|
||||
|
||||
var (
|
||||
db sql.Database
|
||||
err error
|
||||
ind interfaces.StateDiffIndexer
|
||||
)
|
||||
|
||||
func checkTxClosure(t *testing.T, idle, inUse, open int64) {
|
||||
require.Equal(t, idle, db.Stats().Idle())
|
||||
require.Equal(t, inUse, db.Stats().InUse())
|
||||
require.Equal(t, open, db.Stats().Open())
|
||||
}
|
||||
|
||||
func tearDown(t *testing.T) {
|
||||
test_helpers.TearDownDB(t, db)
|
||||
require.NoError(t, ind.Close())
|
||||
}
|
88
indexer/database/sql/interfaces.go
Normal file
88
indexer/database/sql/interfaces.go
Normal file
@ -0,0 +1,88 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package sql
|
||||
|
||||
import (
|
||||
"context"
|
||||
"io"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/metrics"
|
||||
)
|
||||
|
||||
// Database interfaces required by the sql indexer
|
||||
type Database interface {
|
||||
Driver
|
||||
Statements
|
||||
}
|
||||
|
||||
// Driver interface has all the methods required by a driver implementation to support the sql indexer
|
||||
type Driver interface {
|
||||
UseCopyFrom() bool
|
||||
QueryRow(ctx context.Context, sql string, args ...interface{}) ScannableRow
|
||||
Exec(ctx context.Context, sql string, args ...interface{}) (Result, error)
|
||||
Select(ctx context.Context, dest interface{}, query string, args ...interface{}) error
|
||||
Get(ctx context.Context, dest interface{}, query string, args ...interface{}) error
|
||||
Begin(ctx context.Context) (Tx, error)
|
||||
Stats() metrics.DbStats
|
||||
NodeID() string
|
||||
Context() context.Context
|
||||
io.Closer
|
||||
}
|
||||
|
||||
// Statements interface to accommodate different SQL query syntax
|
||||
type Statements interface {
|
||||
InsertHeaderStm() string
|
||||
InsertUncleStm() string
|
||||
InsertTxStm() string
|
||||
InsertRctStm() string
|
||||
InsertLogStm() string
|
||||
InsertStateStm() string
|
||||
InsertStorageStm() string
|
||||
InsertIPLDStm() string
|
||||
InsertIPLDsStm() string
|
||||
|
||||
// Table/column descriptions for use with CopyFrom and similar commands.
|
||||
LogTableName() []string
|
||||
LogColumnNames() []string
|
||||
RctTableName() []string
|
||||
RctColumnNames() []string
|
||||
StateTableName() []string
|
||||
StateColumnNames() []string
|
||||
StorageTableName() []string
|
||||
StorageColumnNames() []string
|
||||
TxTableName() []string
|
||||
TxColumnNames() []string
|
||||
}
|
||||
|
||||
// Tx interface to accommodate different concrete SQL transaction types
|
||||
type Tx interface {
|
||||
QueryRow(ctx context.Context, sql string, args ...interface{}) ScannableRow
|
||||
Exec(ctx context.Context, sql string, args ...interface{}) (Result, error)
|
||||
CopyFrom(ctx context.Context, tableName []string, columnNames []string, rows [][]interface{}) (int64, error)
|
||||
Commit(ctx context.Context) error
|
||||
Rollback(ctx context.Context) error
|
||||
}
|
||||
|
||||
// ScannableRow interface to accommodate different concrete row types
|
||||
type ScannableRow interface {
|
||||
Scan(dest ...interface{}) error
|
||||
}
|
||||
|
||||
// Result interface to accommodate different concrete result types
|
||||
type Result interface {
|
||||
RowsAffected() (int64, error)
|
||||
}
|
106
indexer/database/sql/lazy_tx.go
Normal file
106
indexer/database/sql/lazy_tx.go
Normal file
@ -0,0 +1,106 @@
|
||||
package sql
|
||||
|
||||
import (
|
||||
"context"
|
||||
"reflect"
|
||||
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
)
|
||||
|
||||
// Changing this to 1 would make sure only sequential COPYs were combined.
|
||||
const copyFromCheckLimit = 100
|
||||
|
||||
type DelayedTx struct {
|
||||
cache []interface{}
|
||||
db Database
|
||||
}
|
||||
type cachedStmt struct {
|
||||
sql string
|
||||
args []interface{}
|
||||
}
|
||||
|
||||
type copyFrom struct {
|
||||
tableName []string
|
||||
columnNames []string
|
||||
rows [][]interface{}
|
||||
}
|
||||
|
||||
func (cf *copyFrom) appendRows(rows [][]interface{}) {
|
||||
cf.rows = append(cf.rows, rows...)
|
||||
}
|
||||
|
||||
func (cf *copyFrom) matches(tableName []string, columnNames []string) bool {
|
||||
return reflect.DeepEqual(cf.tableName, tableName) && reflect.DeepEqual(cf.columnNames, columnNames)
|
||||
}
|
||||
|
||||
func NewDelayedTx(db Database) *DelayedTx {
|
||||
return &DelayedTx{db: db}
|
||||
}
|
||||
|
||||
func (tx *DelayedTx) QueryRow(ctx context.Context, sql string, args ...interface{}) ScannableRow {
|
||||
return tx.db.QueryRow(ctx, sql, args...)
|
||||
}
|
||||
|
||||
func (tx *DelayedTx) findPrevCopyFrom(tableName []string, columnNames []string, limit int) (*copyFrom, int) {
|
||||
for pos, count := len(tx.cache)-1, 0; pos >= 0 && count < limit; pos, count = pos-1, count+1 {
|
||||
prevCopy, ok := tx.cache[pos].(*copyFrom)
|
||||
if ok && prevCopy.matches(tableName, columnNames) {
|
||||
return prevCopy, count
|
||||
}
|
||||
}
|
||||
return nil, -1
|
||||
}
|
||||
|
||||
func (tx *DelayedTx) CopyFrom(ctx context.Context, tableName []string, columnNames []string, rows [][]interface{}) (int64, error) {
|
||||
if prevCopy, distance := tx.findPrevCopyFrom(tableName, columnNames, copyFromCheckLimit); nil != prevCopy {
|
||||
log.Trace("statediff lazy_tx : Appending to COPY", "table", tableName,
|
||||
"current", len(prevCopy.rows), "new", len(rows), "distance", distance)
|
||||
prevCopy.appendRows(rows)
|
||||
} else {
|
||||
tx.cache = append(tx.cache, ©From{tableName, columnNames, rows})
|
||||
}
|
||||
|
||||
return 0, nil
|
||||
}
|
||||
|
||||
func (tx *DelayedTx) Exec(ctx context.Context, sql string, args ...interface{}) (Result, error) {
|
||||
tx.cache = append(tx.cache, cachedStmt{sql, args})
|
||||
return nil, nil
|
||||
}
|
||||
|
||||
func (tx *DelayedTx) Commit(ctx context.Context) error {
|
||||
base, err := tx.db.Begin(ctx)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
defer func() {
|
||||
if p := recover(); p != nil {
|
||||
rollback(ctx, base)
|
||||
panic(p)
|
||||
} else if err != nil {
|
||||
rollback(ctx, base)
|
||||
}
|
||||
}()
|
||||
for _, item := range tx.cache {
|
||||
switch item := item.(type) {
|
||||
case *copyFrom:
|
||||
_, err := base.CopyFrom(ctx, item.tableName, item.columnNames, item.rows)
|
||||
if err != nil {
|
||||
log.Error("COPY error", "table", item.tableName, "err", err)
|
||||
return err
|
||||
}
|
||||
case cachedStmt:
|
||||
_, err := base.Exec(ctx, item.sql, item.args...)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
}
|
||||
tx.cache = nil
|
||||
return base.Commit(ctx)
|
||||
}
|
||||
|
||||
func (tx *DelayedTx) Rollback(ctx context.Context) error {
|
||||
tx.cache = nil
|
||||
return nil
|
||||
}
|
95
indexer/database/sql/mainnet_tests/indexer_test.go
Normal file
95
indexer/database/sql/mainnet_tests/indexer_test.go
Normal file
@ -0,0 +1,95 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package mainnet_tests
|
||||
|
||||
import (
|
||||
"context"
|
||||
"fmt"
|
||||
"math/big"
|
||||
"os"
|
||||
"testing"
|
||||
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/params"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql/postgres"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/interfaces"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test_helpers"
|
||||
)
|
||||
|
||||
var (
|
||||
err error
|
||||
db sql.Database
|
||||
ind interfaces.StateDiffIndexer
|
||||
chainConf = params.MainnetChainConfig
|
||||
)
|
||||
|
||||
func init() {
|
||||
if os.Getenv("MODE") != "statediff" {
|
||||
fmt.Println("Skipping statediff test")
|
||||
os.Exit(0)
|
||||
}
|
||||
}
|
||||
|
||||
func TestMainnetIndexer(t *testing.T) {
|
||||
conf := test_helpers.GetTestConfig()
|
||||
|
||||
for _, blockNumber := range test_helpers.ProblemBlocks {
|
||||
conf.BlockNumber = big.NewInt(blockNumber)
|
||||
tb, trs, err := test_helpers.TestBlockAndReceipts(conf)
|
||||
require.NoError(t, err)
|
||||
|
||||
testPushBlockAndState(t, tb, trs)
|
||||
}
|
||||
|
||||
testBlock, testReceipts, err := test_helpers.TestBlockAndReceiptsFromEnv(conf)
|
||||
require.NoError(t, err)
|
||||
|
||||
testPushBlockAndState(t, testBlock, testReceipts)
|
||||
}
|
||||
|
||||
func testPushBlockAndState(t *testing.T, block *types.Block, receipts types.Receipts) {
|
||||
t.Run("Test PushBlock and PushStateNode", func(t *testing.T) {
|
||||
setupMainnetIndexer(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
defer tearDown(t)
|
||||
|
||||
test.TestBlock(t, ind, block, receipts)
|
||||
})
|
||||
}
|
||||
|
||||
func setupMainnetIndexer(t *testing.T) {
|
||||
db, err = postgres.SetupSQLXDB()
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
ind, err = sql.NewStateDiffIndexer(context.Background(), chainConf, db)
|
||||
}
|
||||
|
||||
func checkTxClosure(t *testing.T, idle, inUse, open int64) {
|
||||
require.Equal(t, idle, db.Stats().Idle())
|
||||
require.Equal(t, inUse, db.Stats().InUse())
|
||||
require.Equal(t, open, db.Stats().Open())
|
||||
}
|
||||
|
||||
func tearDown(t *testing.T) {
|
||||
test_helpers.TearDownDB(t, db)
|
||||
require.NoError(t, ind.Close())
|
||||
}
|
52
indexer/database/sql/pgx_indexer_legacy_test.go
Normal file
52
indexer/database/sql/pgx_indexer_legacy_test.go
Normal file
@ -0,0 +1,52 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package sql_test
|
||||
|
||||
import (
|
||||
"context"
|
||||
"testing"
|
||||
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql/postgres"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test"
|
||||
)
|
||||
|
||||
func setupLegacyPGXIndexer(t *testing.T) {
|
||||
db, err = postgres.SetupPGXDB(postgres.TestConfig)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
ind, err = sql.NewStateDiffIndexer(context.Background(), test.LegacyConfig, db)
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
func setupLegacyPGX(t *testing.T) {
|
||||
setupLegacyPGXIndexer(t)
|
||||
test.SetupLegacyTestData(t, ind)
|
||||
}
|
||||
|
||||
func TestLegacyPGXIndexer(t *testing.T) {
|
||||
t.Run("Publish and index header IPLDs", func(t *testing.T) {
|
||||
setupLegacyPGX(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
test.TestLegacyIndexer(t, db)
|
||||
})
|
||||
}
|
245
indexer/database/sql/pgx_indexer_test.go
Normal file
245
indexer/database/sql/pgx_indexer_test.go
Normal file
@ -0,0 +1,245 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package sql_test
|
||||
|
||||
import (
|
||||
"context"
|
||||
"math/big"
|
||||
"testing"
|
||||
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql/postgres"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/mocks"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test"
|
||||
)
|
||||
|
||||
func setupPGXIndexer(t *testing.T, config postgres.Config) {
|
||||
db, err = postgres.SetupPGXDB(config)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
ind, err = sql.NewStateDiffIndexer(context.Background(), mocks.TestConfig, db)
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
func setupPGX(t *testing.T) {
|
||||
setupPGXWithConfig(t, postgres.TestConfig)
|
||||
}
|
||||
|
||||
func setupPGXWithConfig(t *testing.T, config postgres.Config) {
|
||||
setupPGXIndexer(t, config)
|
||||
test.SetupTestData(t, ind)
|
||||
}
|
||||
|
||||
func setupPGXNonCanonical(t *testing.T) {
|
||||
setupPGXIndexer(t, postgres.TestConfig)
|
||||
test.SetupTestDataNonCanonical(t, ind)
|
||||
}
|
||||
|
||||
// Test indexer for a canonical block
|
||||
func TestPGXIndexer(t *testing.T) {
|
||||
t.Run("Publish and index header IPLDs in a single tx", func(t *testing.T) {
|
||||
setupPGX(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
test.TestPublishAndIndexHeaderIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index transaction IPLDs in a single tx", func(t *testing.T) {
|
||||
setupPGX(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
test.TestPublishAndIndexTransactionIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index log IPLDs for multiple receipt of a specific block", func(t *testing.T) {
|
||||
setupPGX(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
test.TestPublishAndIndexLogIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index receipt IPLDs in a single tx", func(t *testing.T) {
|
||||
setupPGX(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
test.TestPublishAndIndexReceiptIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index state IPLDs in a single tx", func(t *testing.T) {
|
||||
setupPGX(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
test.TestPublishAndIndexStateIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index storage IPLDs in a single tx", func(t *testing.T) {
|
||||
setupPGX(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
test.TestPublishAndIndexStorageIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index with CopyFrom enabled.", func(t *testing.T) {
|
||||
config := postgres.TestConfig
|
||||
config.CopyFrom = true
|
||||
|
||||
setupPGXWithConfig(t, config)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
test.TestPublishAndIndexStateIPLDs(t, db)
|
||||
test.TestPublishAndIndexStorageIPLDs(t, db)
|
||||
test.TestPublishAndIndexReceiptIPLDs(t, db)
|
||||
test.TestPublishAndIndexLogIPLDs(t, db)
|
||||
})
|
||||
}
|
||||
|
||||
// Test indexer for a canonical + a non-canonical block at London height + a non-canonical block at London height + 1
|
||||
func TestPGXIndexerNonCanonical(t *testing.T) {
|
||||
t.Run("Publish and index header", func(t *testing.T) {
|
||||
setupPGXNonCanonical(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
test.TestPublishAndIndexHeaderNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index transactions", func(t *testing.T) {
|
||||
setupPGXNonCanonical(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
test.TestPublishAndIndexTransactionsNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index receipts", func(t *testing.T) {
|
||||
setupPGXNonCanonical(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
test.TestPublishAndIndexReceiptsNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index logs", func(t *testing.T) {
|
||||
setupPGXNonCanonical(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
test.TestPublishAndIndexLogsNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index state nodes", func(t *testing.T) {
|
||||
setupPGXNonCanonical(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
test.TestPublishAndIndexStateNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index storage nodes", func(t *testing.T) {
|
||||
setupPGXNonCanonical(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
test.TestPublishAndIndexStorageNonCanonical(t, db)
|
||||
})
|
||||
}
|
||||
|
||||
func TestPGXWatchAddressMethods(t *testing.T) {
|
||||
setupPGXIndexer(t, postgres.TestConfig)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 1, 0, 1)
|
||||
|
||||
t.Run("Load watched addresses (empty table)", func(t *testing.T) {
|
||||
test.TestLoadEmptyWatchedAddresses(t, ind)
|
||||
})
|
||||
|
||||
t.Run("Insert watched addresses", func(t *testing.T) {
|
||||
args := mocks.GetInsertWatchedAddressesArgs()
|
||||
err = ind.InsertWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt1)))
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestInsertWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Insert watched addresses (some already watched)", func(t *testing.T) {
|
||||
args := mocks.GetInsertAlreadyWatchedAddressesArgs()
|
||||
err = ind.InsertWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt2)))
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestInsertAlreadyWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Remove watched addresses", func(t *testing.T) {
|
||||
args := mocks.GetRemoveWatchedAddressesArgs()
|
||||
err = ind.RemoveWatchedAddresses(args)
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestRemoveWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Remove watched addresses (some non-watched)", func(t *testing.T) {
|
||||
args := mocks.GetRemoveNonWatchedAddressesArgs()
|
||||
err = ind.RemoveWatchedAddresses(args)
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestRemoveNonWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Set watched addresses", func(t *testing.T) {
|
||||
args := mocks.GetSetWatchedAddressesArgs()
|
||||
err = ind.SetWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt2)))
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestSetWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Set watched addresses (some already watched)", func(t *testing.T) {
|
||||
args := mocks.GetSetAlreadyWatchedAddressesArgs()
|
||||
err = ind.SetWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt3)))
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestSetAlreadyWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Load watched addresses", func(t *testing.T) {
|
||||
test.TestLoadWatchedAddresses(t, ind)
|
||||
})
|
||||
|
||||
t.Run("Clear watched addresses", func(t *testing.T) {
|
||||
err = ind.ClearWatchedAddresses()
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestClearWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Clear watched addresses (empty table)", func(t *testing.T) {
|
||||
err = ind.ClearWatchedAddresses()
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestClearEmptyWatchedAddresses(t, db)
|
||||
})
|
||||
}
|
140
indexer/database/sql/postgres/config.go
Normal file
140
indexer/database/sql/postgres/config.go
Normal file
@ -0,0 +1,140 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package postgres
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"os"
|
||||
"strconv"
|
||||
"strings"
|
||||
"time"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/shared"
|
||||
)
|
||||
|
||||
// DriverType to explicitly type the kind of sql driver we are using
|
||||
type DriverType string
|
||||
|
||||
const (
|
||||
PGX DriverType = "PGX"
|
||||
SQLX DriverType = "SQLX"
|
||||
Unknown DriverType = "Unknown"
|
||||
)
|
||||
|
||||
// Env variables
|
||||
const (
|
||||
DATABASE_NAME = "DATABASE_NAME"
|
||||
DATABASE_HOSTNAME = "DATABASE_HOSTNAME"
|
||||
DATABASE_PORT = "DATABASE_PORT"
|
||||
DATABASE_USER = "DATABASE_USER"
|
||||
DATABASE_PASSWORD = "DATABASE_PASSWORD"
|
||||
)
|
||||
|
||||
// ResolveDriverType resolves a DriverType from a provided string
|
||||
func ResolveDriverType(str string) (DriverType, error) {
|
||||
switch strings.ToLower(str) {
|
||||
case "pgx", "pgxpool":
|
||||
return PGX, nil
|
||||
case "sqlx":
|
||||
return SQLX, nil
|
||||
default:
|
||||
return Unknown, fmt.Errorf("unrecognized driver type string: %s", str)
|
||||
}
|
||||
}
|
||||
|
||||
// TestConfig specifies default parameters for connecting to a testing DB
|
||||
var TestConfig = Config{
|
||||
Hostname: "localhost",
|
||||
Port: 8077,
|
||||
DatabaseName: "cerc_testing",
|
||||
Username: "vdbm",
|
||||
Password: "password",
|
||||
Driver: SQLX,
|
||||
}
|
||||
|
||||
// Config holds params for a Postgres db
|
||||
type Config struct {
|
||||
// conn string params
|
||||
Hostname string
|
||||
Port int
|
||||
DatabaseName string
|
||||
Username string
|
||||
Password string
|
||||
|
||||
// conn settings
|
||||
MaxConns int
|
||||
MaxIdle int
|
||||
MinConns int
|
||||
MaxConnIdleTime time.Duration
|
||||
MaxConnLifetime time.Duration
|
||||
ConnTimeout time.Duration
|
||||
LogStatements bool
|
||||
|
||||
// node info params
|
||||
ID string
|
||||
ClientName string
|
||||
|
||||
// driver type
|
||||
Driver DriverType
|
||||
|
||||
// toggle on/off upserts
|
||||
Upsert bool
|
||||
|
||||
// toggle on/off CopyFrom
|
||||
CopyFrom bool
|
||||
}
|
||||
|
||||
// Type satisfies interfaces.Config
|
||||
func (c Config) Type() shared.DBType {
|
||||
return shared.POSTGRES
|
||||
}
|
||||
|
||||
// DbConnectionString constructs and returns the connection string from the config
|
||||
func (c Config) DbConnectionString() string {
|
||||
if len(c.Username) > 0 && len(c.Password) > 0 {
|
||||
return fmt.Sprintf("postgresql://%s:%s@%s:%d/%s?sslmode=disable",
|
||||
c.Username, c.Password, c.Hostname, c.Port, c.DatabaseName)
|
||||
}
|
||||
if len(c.Username) > 0 && len(c.Password) == 0 {
|
||||
return fmt.Sprintf("postgresql://%s@%s:%d/%s?sslmode=disable",
|
||||
c.Username, c.Hostname, c.Port, c.DatabaseName)
|
||||
}
|
||||
return fmt.Sprintf("postgresql://%s:%d/%s?sslmode=disable", c.Hostname, c.Port, c.DatabaseName)
|
||||
}
|
||||
|
||||
func (c Config) WithEnv() (Config, error) {
|
||||
if val := os.Getenv(DATABASE_NAME); val != "" {
|
||||
c.DatabaseName = val
|
||||
}
|
||||
if val := os.Getenv(DATABASE_HOSTNAME); val != "" {
|
||||
c.Hostname = val
|
||||
}
|
||||
if val := os.Getenv(DATABASE_PORT); val != "" {
|
||||
port, err := strconv.Atoi(val)
|
||||
if err != nil {
|
||||
return c, err
|
||||
}
|
||||
c.Port = port
|
||||
}
|
||||
if val := os.Getenv(DATABASE_USER); val != "" {
|
||||
c.Username = val
|
||||
}
|
||||
if val := os.Getenv(DATABASE_PASSWORD); val != "" {
|
||||
c.Password = val
|
||||
}
|
||||
return c, nil
|
||||
}
|
126
indexer/database/sql/postgres/database.go
Normal file
126
indexer/database/sql/postgres/database.go
Normal file
@ -0,0 +1,126 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package postgres
|
||||
|
||||
import (
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/shared/schema"
|
||||
)
|
||||
|
||||
var _ sql.Database = &DB{}
|
||||
|
||||
const (
|
||||
createNodeStm = `INSERT INTO nodes (genesis_block, network_id, node_id, client_name, chain_id) VALUES ($1, $2, $3, $4, $5)
|
||||
ON CONFLICT (node_id) DO NOTHING`
|
||||
)
|
||||
|
||||
// NewPostgresDB returns a postgres.DB using the provided driver
|
||||
func NewPostgresDB(driver sql.Driver, upsert bool) *DB {
|
||||
return &DB{upsert, driver}
|
||||
}
|
||||
|
||||
// DB implements sql.Database using a configured driver and Postgres statement syntax
|
||||
type DB struct {
|
||||
upsert bool
|
||||
sql.Driver
|
||||
}
|
||||
|
||||
// InsertHeaderStm satisfies the sql.Statements interface
|
||||
// Stm == Statement
|
||||
func (db *DB) InsertHeaderStm() string {
|
||||
return schema.TableHeader.ToInsertStatement(db.upsert)
|
||||
}
|
||||
|
||||
// InsertUncleStm satisfies the sql.Statements interface
|
||||
func (db *DB) InsertUncleStm() string {
|
||||
return schema.TableUncle.ToInsertStatement(db.upsert)
|
||||
}
|
||||
|
||||
// InsertTxStm satisfies the sql.Statements interface
|
||||
func (db *DB) InsertTxStm() string {
|
||||
return schema.TableTransaction.ToInsertStatement(db.upsert)
|
||||
}
|
||||
|
||||
// InsertRctStm satisfies the sql.Statements interface
|
||||
func (db *DB) InsertRctStm() string {
|
||||
return schema.TableReceipt.ToInsertStatement(db.upsert)
|
||||
}
|
||||
|
||||
// InsertLogStm satisfies the sql.Statements interface
|
||||
func (db *DB) InsertLogStm() string {
|
||||
return schema.TableLog.ToInsertStatement(db.upsert)
|
||||
}
|
||||
|
||||
// InsertStateStm satisfies the sql.Statements interface
|
||||
func (db *DB) InsertStateStm() string {
|
||||
return schema.TableStateNode.ToInsertStatement(db.upsert)
|
||||
}
|
||||
|
||||
// InsertStorageStm satisfies the sql.Statements interface
|
||||
func (db *DB) InsertStorageStm() string {
|
||||
return schema.TableStorageNode.ToInsertStatement(db.upsert)
|
||||
}
|
||||
|
||||
// InsertIPLDStm satisfies the sql.Statements interface
|
||||
func (db *DB) InsertIPLDStm() string {
|
||||
return schema.TableIPLDBlock.ToInsertStatement(db.upsert)
|
||||
}
|
||||
|
||||
// InsertIPLDsStm satisfies the sql.Statements interface
|
||||
func (db *DB) InsertIPLDsStm() string {
|
||||
return `INSERT INTO ipld.blocks (block_number, key, data) VALUES (unnest($1::BIGINT[]), unnest($2::TEXT[]), unnest($3::BYTEA[])) ON CONFLICT DO NOTHING`
|
||||
}
|
||||
|
||||
func (db *DB) LogTableName() []string {
|
||||
return []string{"eth", "log_cids"}
|
||||
}
|
||||
|
||||
func (db *DB) LogColumnNames() []string {
|
||||
return []string{"block_number", "header_id", "cid", "rct_id", "address", "index", "topic0", "topic1", "topic2", "topic3"}
|
||||
}
|
||||
|
||||
func (db *DB) RctTableName() []string {
|
||||
return []string{"eth", "receipt_cids"}
|
||||
}
|
||||
|
||||
func (db *DB) RctColumnNames() []string {
|
||||
return []string{"block_number", "header_id", "tx_id", "cid", "contract", "post_state", "post_status"}
|
||||
}
|
||||
|
||||
func (db *DB) StateTableName() []string {
|
||||
return []string{"eth", "state_cids"}
|
||||
}
|
||||
|
||||
func (db *DB) StateColumnNames() []string {
|
||||
return []string{"block_number", "header_id", "state_leaf_key", "cid", "diff", "balance", "nonce", "code_hash", "storage_root", "removed"}
|
||||
}
|
||||
|
||||
func (db *DB) StorageTableName() []string {
|
||||
return []string{"eth", "storage_cids"}
|
||||
}
|
||||
|
||||
func (db *DB) StorageColumnNames() []string {
|
||||
return []string{"block_number", "header_id", "state_leaf_key", "storage_leaf_key", "cid", "diff", "val", "removed"}
|
||||
}
|
||||
|
||||
func (db *DB) TxTableName() []string {
|
||||
return []string{"eth", "transaction_cids"}
|
||||
}
|
||||
|
||||
func (db *DB) TxColumnNames() []string {
|
||||
return []string{"block_number", "header_id", "tx_hash", "cid", "dst", "src", "index", "tx_type", "value"}
|
||||
}
|
38
indexer/database/sql/postgres/errors.go
Normal file
38
indexer/database/sql/postgres/errors.go
Normal file
@ -0,0 +1,38 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package postgres
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
)
|
||||
|
||||
const (
|
||||
DbConnectionFailedMsg = "db connection failed"
|
||||
SettingNodeFailedMsg = "unable to set db node"
|
||||
)
|
||||
|
||||
func ErrDBConnectionFailed(connectErr error) error {
|
||||
return formatError(DbConnectionFailedMsg, connectErr)
|
||||
}
|
||||
|
||||
func ErrUnableToSetNode(setErr error) error {
|
||||
return formatError(SettingNodeFailedMsg, setErr)
|
||||
}
|
||||
|
||||
func formatError(msg string, err error) error {
|
||||
return fmt.Errorf("%s: %w", msg, err)
|
||||
}
|
61
indexer/database/sql/postgres/log_adapter.go
Normal file
61
indexer/database/sql/postgres/log_adapter.go
Normal file
@ -0,0 +1,61 @@
|
||||
// Copyright © 2023 Cerc
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package postgres
|
||||
|
||||
import (
|
||||
"context"
|
||||
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
"github.com/jackc/pgx/v4"
|
||||
)
|
||||
|
||||
type LogAdapter struct {
|
||||
l log.Logger
|
||||
}
|
||||
|
||||
func NewLogAdapter(l log.Logger) *LogAdapter {
|
||||
return &LogAdapter{l: l}
|
||||
}
|
||||
|
||||
func (l *LogAdapter) Log(ctx context.Context, level pgx.LogLevel, msg string, data map[string]interface{}) {
|
||||
var logger log.Logger
|
||||
if data != nil {
|
||||
var args = make([]interface{}, 0)
|
||||
for key, value := range data {
|
||||
if value != nil {
|
||||
args = append(args, key, value)
|
||||
}
|
||||
}
|
||||
logger = l.l.New(args...)
|
||||
} else {
|
||||
logger = l.l
|
||||
}
|
||||
|
||||
switch level {
|
||||
case pgx.LogLevelTrace:
|
||||
logger.Trace(msg)
|
||||
case pgx.LogLevelDebug:
|
||||
logger.Debug(msg)
|
||||
case pgx.LogLevelInfo:
|
||||
logger.Info(msg)
|
||||
case pgx.LogLevelWarn:
|
||||
logger.Warn(msg)
|
||||
case pgx.LogLevelError:
|
||||
logger.Error(msg)
|
||||
default:
|
||||
logger.New("INVALID_PGX_LOG_LEVEL", level).Error(msg)
|
||||
}
|
||||
}
|
256
indexer/database/sql/postgres/pgx.go
Normal file
256
indexer/database/sql/postgres/pgx.go
Normal file
@ -0,0 +1,256 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package postgres
|
||||
|
||||
import (
|
||||
"context"
|
||||
"time"
|
||||
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
|
||||
"github.com/georgysavva/scany/pgxscan"
|
||||
"github.com/jackc/pgconn"
|
||||
"github.com/jackc/pgx/v4"
|
||||
"github.com/jackc/pgx/v4/pgxpool"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/metrics"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/node"
|
||||
)
|
||||
|
||||
// PGXDriver driver, implements sql.Driver
|
||||
type PGXDriver struct {
|
||||
ctx context.Context
|
||||
pool *pgxpool.Pool
|
||||
nodeInfo node.Info
|
||||
nodeID string
|
||||
config Config
|
||||
}
|
||||
|
||||
// ConnectPGX initializes and returns a PGX connection pool
|
||||
func ConnectPGX(ctx context.Context, config Config) (*pgxpool.Pool, error) {
|
||||
pgConf, err := MakeConfig(config)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return pgxpool.ConnectConfig(ctx, pgConf)
|
||||
}
|
||||
|
||||
// NewPGXDriver returns a new pgx driver
|
||||
// it initializes the connection pool and creates the node info table
|
||||
func NewPGXDriver(ctx context.Context, config Config, node node.Info) (*PGXDriver, error) {
|
||||
dbPool, err := ConnectPGX(ctx, config)
|
||||
if err != nil {
|
||||
return nil, ErrDBConnectionFailed(err)
|
||||
}
|
||||
pg := &PGXDriver{ctx: ctx, pool: dbPool, nodeInfo: node, config: config}
|
||||
nodeErr := pg.createNode()
|
||||
if nodeErr != nil {
|
||||
return &PGXDriver{}, ErrUnableToSetNode(nodeErr)
|
||||
}
|
||||
return pg, nil
|
||||
}
|
||||
|
||||
// MakeConfig creates a pgxpool.Config from the provided Config
|
||||
func MakeConfig(config Config) (*pgxpool.Config, error) {
|
||||
conf, err := pgxpool.ParseConfig("")
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
//conf.ConnConfig.BuildStatementCache = nil
|
||||
conf.ConnConfig.Config.Host = config.Hostname
|
||||
conf.ConnConfig.Config.Port = uint16(config.Port)
|
||||
conf.ConnConfig.Config.Database = config.DatabaseName
|
||||
conf.ConnConfig.Config.User = config.Username
|
||||
conf.ConnConfig.Config.Password = config.Password
|
||||
|
||||
if config.ConnTimeout != 0 {
|
||||
conf.ConnConfig.Config.ConnectTimeout = config.ConnTimeout
|
||||
}
|
||||
if config.MaxConns != 0 {
|
||||
conf.MaxConns = int32(config.MaxConns)
|
||||
}
|
||||
if config.MinConns != 0 {
|
||||
conf.MinConns = int32(config.MinConns)
|
||||
}
|
||||
if config.MaxConnLifetime != 0 {
|
||||
conf.MaxConnLifetime = config.MaxConnLifetime
|
||||
}
|
||||
if config.MaxConnIdleTime != 0 {
|
||||
conf.MaxConnIdleTime = config.MaxConnIdleTime
|
||||
}
|
||||
|
||||
if config.LogStatements {
|
||||
conf.ConnConfig.Logger = NewLogAdapter(log.New())
|
||||
}
|
||||
|
||||
return conf, nil
|
||||
}
|
||||
|
||||
func (pgx *PGXDriver) createNode() error {
|
||||
_, err := pgx.pool.Exec(
|
||||
pgx.ctx,
|
||||
createNodeStm,
|
||||
pgx.nodeInfo.GenesisBlock, pgx.nodeInfo.NetworkID,
|
||||
pgx.nodeInfo.ID, pgx.nodeInfo.ClientName,
|
||||
pgx.nodeInfo.ChainID)
|
||||
if err != nil {
|
||||
return ErrUnableToSetNode(err)
|
||||
}
|
||||
pgx.nodeID = pgx.nodeInfo.ID
|
||||
return nil
|
||||
}
|
||||
|
||||
// QueryRow satisfies sql.Database
|
||||
func (pgx *PGXDriver) QueryRow(ctx context.Context, sql string, args ...interface{}) sql.ScannableRow {
|
||||
return pgx.pool.QueryRow(ctx, sql, args...)
|
||||
}
|
||||
|
||||
// Exec satisfies sql.Database
|
||||
func (pgx *PGXDriver) Exec(ctx context.Context, sql string, args ...interface{}) (sql.Result, error) {
|
||||
res, err := pgx.pool.Exec(ctx, sql, args...)
|
||||
return resultWrapper{ct: res}, err
|
||||
}
|
||||
|
||||
// Select satisfies sql.Database
|
||||
func (pgx *PGXDriver) Select(ctx context.Context, dest interface{}, query string, args ...interface{}) error {
|
||||
return pgxscan.Select(ctx, pgx.pool, dest, query, args...)
|
||||
}
|
||||
|
||||
// Get satisfies sql.Database
|
||||
func (pgx *PGXDriver) Get(ctx context.Context, dest interface{}, query string, args ...interface{}) error {
|
||||
return pgxscan.Get(ctx, pgx.pool, dest, query, args...)
|
||||
}
|
||||
|
||||
// Begin satisfies sql.Database
|
||||
func (pgx *PGXDriver) Begin(ctx context.Context) (sql.Tx, error) {
|
||||
tx, err := pgx.pool.Begin(ctx)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return pgxTxWrapper{tx: tx}, nil
|
||||
}
|
||||
|
||||
func (pgx *PGXDriver) Stats() metrics.DbStats {
|
||||
stats := pgx.pool.Stat()
|
||||
return pgxStatsWrapper{stats: stats}
|
||||
}
|
||||
|
||||
// NodeID satisfies sql.Database
|
||||
func (pgx *PGXDriver) NodeID() string {
|
||||
return pgx.nodeID
|
||||
}
|
||||
|
||||
// Close satisfies sql.Database/io.Closer
|
||||
func (pgx *PGXDriver) Close() error {
|
||||
pgx.pool.Close()
|
||||
return nil
|
||||
}
|
||||
|
||||
// Context satisfies sql.Database
|
||||
func (pgx *PGXDriver) Context() context.Context {
|
||||
return pgx.ctx
|
||||
}
|
||||
|
||||
// HasCopy satisfies sql.Database
|
||||
func (pgx *PGXDriver) UseCopyFrom() bool {
|
||||
return pgx.config.CopyFrom
|
||||
}
|
||||
|
||||
type resultWrapper struct {
|
||||
ct pgconn.CommandTag
|
||||
}
|
||||
|
||||
// RowsAffected satisfies sql.Result
|
||||
func (r resultWrapper) RowsAffected() (int64, error) {
|
||||
return r.ct.RowsAffected(), nil
|
||||
}
|
||||
|
||||
type pgxStatsWrapper struct {
|
||||
stats *pgxpool.Stat
|
||||
}
|
||||
|
||||
// MaxOpen satisfies metrics.DbStats
|
||||
func (s pgxStatsWrapper) MaxOpen() int64 {
|
||||
return int64(s.stats.MaxConns())
|
||||
}
|
||||
|
||||
// Open satisfies metrics.DbStats
|
||||
func (s pgxStatsWrapper) Open() int64 {
|
||||
return int64(s.stats.TotalConns())
|
||||
}
|
||||
|
||||
// InUse satisfies metrics.DbStats
|
||||
func (s pgxStatsWrapper) InUse() int64 {
|
||||
return int64(s.stats.AcquiredConns())
|
||||
}
|
||||
|
||||
// Idle satisfies metrics.DbStats
|
||||
func (s pgxStatsWrapper) Idle() int64 {
|
||||
return int64(s.stats.IdleConns())
|
||||
}
|
||||
|
||||
// WaitCount satisfies metrics.DbStats
|
||||
func (s pgxStatsWrapper) WaitCount() int64 {
|
||||
return s.stats.EmptyAcquireCount()
|
||||
}
|
||||
|
||||
// WaitDuration satisfies metrics.DbStats
|
||||
func (s pgxStatsWrapper) WaitDuration() time.Duration {
|
||||
return s.stats.AcquireDuration()
|
||||
}
|
||||
|
||||
// MaxIdleClosed satisfies metrics.DbStats
|
||||
func (s pgxStatsWrapper) MaxIdleClosed() int64 {
|
||||
// this stat isn't supported by pgxpool, but we don't want to panic
|
||||
return 0
|
||||
}
|
||||
|
||||
// MaxLifetimeClosed satisfies metrics.DbStats
|
||||
func (s pgxStatsWrapper) MaxLifetimeClosed() int64 {
|
||||
return s.stats.CanceledAcquireCount()
|
||||
}
|
||||
|
||||
type pgxTxWrapper struct {
|
||||
tx pgx.Tx
|
||||
}
|
||||
|
||||
// QueryRow satisfies sql.Tx
|
||||
func (t pgxTxWrapper) QueryRow(ctx context.Context, sql string, args ...interface{}) sql.ScannableRow {
|
||||
return t.tx.QueryRow(ctx, sql, args...)
|
||||
}
|
||||
|
||||
// Exec satisfies sql.Tx
|
||||
func (t pgxTxWrapper) Exec(ctx context.Context, sql string, args ...interface{}) (sql.Result, error) {
|
||||
res, err := t.tx.Exec(ctx, sql, args...)
|
||||
return resultWrapper{ct: res}, err
|
||||
}
|
||||
|
||||
// Commit satisfies sql.Tx
|
||||
func (t pgxTxWrapper) Commit(ctx context.Context) error {
|
||||
return t.tx.Commit(ctx)
|
||||
}
|
||||
|
||||
// Rollback satisfies sql.Tx
|
||||
func (t pgxTxWrapper) Rollback(ctx context.Context) error {
|
||||
return t.tx.Rollback(ctx)
|
||||
}
|
||||
|
||||
func (t pgxTxWrapper) CopyFrom(ctx context.Context, tableName []string, columnNames []string, rows [][]interface{}) (int64, error) {
|
||||
return t.tx.CopyFrom(ctx, tableName, columnNames, pgx.CopyFromRows(rows))
|
||||
}
|
121
indexer/database/sql/postgres/pgx_test.go
Normal file
121
indexer/database/sql/postgres/pgx_test.go
Normal file
@ -0,0 +1,121 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package postgres_test
|
||||
|
||||
import (
|
||||
"context"
|
||||
"fmt"
|
||||
"math/big"
|
||||
"strings"
|
||||
"testing"
|
||||
|
||||
"github.com/jackc/pgx/v4/pgxpool"
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql/postgres"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/node"
|
||||
)
|
||||
|
||||
var (
|
||||
pgConfig, _ = postgres.MakeConfig(postgres.TestConfig)
|
||||
ctx = context.Background()
|
||||
)
|
||||
|
||||
func expectContainsSubstring(t *testing.T, full string, sub string) {
|
||||
if !strings.Contains(full, sub) {
|
||||
t.Fatalf("Expected \"%v\" to contain substring \"%v\"\n", full, sub)
|
||||
}
|
||||
}
|
||||
|
||||
func TestPostgresPGX(t *testing.T) {
|
||||
t.Run("connects to the sql", func(t *testing.T) {
|
||||
dbPool, err := pgxpool.ConnectConfig(context.Background(), pgConfig)
|
||||
if err != nil {
|
||||
t.Fatalf("failed to connect to db with connection string: %s err: %v", pgConfig.ConnString(), err)
|
||||
}
|
||||
if dbPool == nil {
|
||||
t.Fatal("DB pool is nil")
|
||||
}
|
||||
dbPool.Close()
|
||||
})
|
||||
|
||||
t.Run("serializes big.Int to db", func(t *testing.T) {
|
||||
// postgres driver doesn't support go big.Int type
|
||||
// various casts in golang uint64, int64, overflow for
|
||||
// transaction value (in wei) even though
|
||||
// postgres numeric can handle an arbitrary
|
||||
// sized int, so use string representation of big.Int
|
||||
// and cast on insert
|
||||
|
||||
dbPool, err := pgxpool.ConnectConfig(context.Background(), pgConfig)
|
||||
if err != nil {
|
||||
t.Fatalf("failed to connect to db with connection string: %s err: %v", pgConfig.ConnString(), err)
|
||||
}
|
||||
defer dbPool.Close()
|
||||
|
||||
bi := new(big.Int)
|
||||
bi.SetString("34940183920000000000", 10)
|
||||
require.Equal(t, "34940183920000000000", bi.String())
|
||||
|
||||
defer dbPool.Exec(ctx, `DROP TABLE IF EXISTS example`)
|
||||
_, err = dbPool.Exec(ctx, "CREATE TABLE example ( id INTEGER, data NUMERIC )")
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
sqlStatement := `
|
||||
INSERT INTO example (id, data)
|
||||
VALUES (1, cast($1 AS NUMERIC))`
|
||||
_, err = dbPool.Exec(ctx, sqlStatement, bi.String())
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
var data string
|
||||
err = dbPool.QueryRow(ctx, `SELECT cast(data AS TEXT) FROM example WHERE id = 1`).Scan(&data)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
require.Equal(t, data, bi.String())
|
||||
actual := new(big.Int)
|
||||
actual.SetString(data, 10)
|
||||
require.Equal(t, bi, actual)
|
||||
})
|
||||
|
||||
t.Run("throws error when can't connect to the database", func(t *testing.T) {
|
||||
goodInfo := node.Info{GenesisBlock: "GENESIS", NetworkID: "1", ID: "x123", ClientName: "geth"}
|
||||
_, err := postgres.NewPGXDriver(ctx, postgres.Config{}, goodInfo)
|
||||
if err == nil {
|
||||
t.Fatal("Expected an error")
|
||||
}
|
||||
|
||||
expectContainsSubstring(t, err.Error(), postgres.DbConnectionFailedMsg)
|
||||
})
|
||||
|
||||
t.Run("throws error when can't create node", func(t *testing.T) {
|
||||
badHash := fmt.Sprintf("x %s", strings.Repeat("1", 100))
|
||||
badInfo := node.Info{GenesisBlock: badHash, NetworkID: "1", ID: "x123", ClientName: "geth"}
|
||||
|
||||
_, err := postgres.NewPGXDriver(ctx, postgres.TestConfig, badInfo)
|
||||
if err == nil {
|
||||
t.Fatal("Expected an error")
|
||||
}
|
||||
|
||||
expectContainsSubstring(t, err.Error(), postgres.SettingNodeFailedMsg)
|
||||
})
|
||||
}
|
33
indexer/database/sql/postgres/postgres_suite_test.go
Normal file
33
indexer/database/sql/postgres/postgres_suite_test.go
Normal file
@ -0,0 +1,33 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package postgres_test
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"os"
|
||||
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
)
|
||||
|
||||
func init() {
|
||||
if os.Getenv("MODE") != "statediff" {
|
||||
fmt.Println("Skipping statediff test")
|
||||
os.Exit(0)
|
||||
}
|
||||
|
||||
log.Root().SetHandler(log.DiscardHandler())
|
||||
}
|
210
indexer/database/sql/postgres/sqlx.go
Normal file
210
indexer/database/sql/postgres/sqlx.go
Normal file
@ -0,0 +1,210 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package postgres
|
||||
|
||||
import (
|
||||
"context"
|
||||
coresql "database/sql"
|
||||
"errors"
|
||||
"time"
|
||||
|
||||
"github.com/jmoiron/sqlx"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/metrics"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/node"
|
||||
)
|
||||
|
||||
// SQLXDriver driver, implements sql.Driver
|
||||
type SQLXDriver struct {
|
||||
ctx context.Context
|
||||
db *sqlx.DB
|
||||
nodeInfo node.Info
|
||||
nodeID string
|
||||
}
|
||||
|
||||
// ConnectSQLX initializes and returns a SQLX connection pool for postgres
|
||||
func ConnectSQLX(ctx context.Context, config Config) (*sqlx.DB, error) {
|
||||
db, err := sqlx.ConnectContext(ctx, "postgres", config.DbConnectionString())
|
||||
if err != nil {
|
||||
return nil, ErrDBConnectionFailed(err)
|
||||
}
|
||||
if config.MaxConns > 0 {
|
||||
db.SetMaxOpenConns(config.MaxConns)
|
||||
}
|
||||
if config.MaxConnLifetime > 0 {
|
||||
db.SetConnMaxLifetime(config.MaxConnLifetime)
|
||||
}
|
||||
db.SetMaxIdleConns(config.MaxIdle)
|
||||
return db, nil
|
||||
}
|
||||
|
||||
// NewSQLXDriver returns a new sqlx driver for Postgres
|
||||
// it initializes the connection pool and creates the node info table
|
||||
func NewSQLXDriver(ctx context.Context, config Config, node node.Info) (*SQLXDriver, error) {
|
||||
db, err := ConnectSQLX(ctx, config)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
driver := &SQLXDriver{ctx: ctx, db: db, nodeInfo: node}
|
||||
if err := driver.createNode(); err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return driver, nil
|
||||
}
|
||||
|
||||
func (driver *SQLXDriver) createNode() error {
|
||||
_, err := driver.db.Exec(
|
||||
createNodeStm,
|
||||
driver.nodeInfo.GenesisBlock,
|
||||
driver.nodeInfo.NetworkID,
|
||||
driver.nodeInfo.ID,
|
||||
driver.nodeInfo.ClientName,
|
||||
driver.nodeInfo.ChainID)
|
||||
if err != nil {
|
||||
return ErrUnableToSetNode(err)
|
||||
}
|
||||
driver.nodeID = driver.nodeInfo.ID
|
||||
return nil
|
||||
}
|
||||
|
||||
// QueryRow satisfies sql.Database
|
||||
func (driver *SQLXDriver) QueryRow(_ context.Context, sql string, args ...interface{}) sql.ScannableRow {
|
||||
return driver.db.QueryRowx(sql, args...)
|
||||
}
|
||||
|
||||
// Exec satisfies sql.Database
|
||||
func (driver *SQLXDriver) Exec(_ context.Context, sql string, args ...interface{}) (sql.Result, error) {
|
||||
return driver.db.Exec(sql, args...)
|
||||
}
|
||||
|
||||
// Select satisfies sql.Database
|
||||
func (driver *SQLXDriver) Select(_ context.Context, dest interface{}, query string, args ...interface{}) error {
|
||||
return driver.db.Select(dest, query, args...)
|
||||
}
|
||||
|
||||
// Get satisfies sql.Database
|
||||
func (driver *SQLXDriver) Get(_ context.Context, dest interface{}, query string, args ...interface{}) error {
|
||||
return driver.db.Get(dest, query, args...)
|
||||
}
|
||||
|
||||
// Begin satisfies sql.Database
|
||||
func (driver *SQLXDriver) Begin(_ context.Context) (sql.Tx, error) {
|
||||
tx, err := driver.db.Beginx()
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return sqlxTxWrapper{tx: tx}, nil
|
||||
}
|
||||
|
||||
func (driver *SQLXDriver) Stats() metrics.DbStats {
|
||||
stats := driver.db.Stats()
|
||||
return sqlxStatsWrapper{stats: stats}
|
||||
}
|
||||
|
||||
// NodeID satisfies sql.Database
|
||||
func (driver *SQLXDriver) NodeID() string {
|
||||
return driver.nodeID
|
||||
}
|
||||
|
||||
// Close satisfies sql.Database/io.Closer
|
||||
func (driver *SQLXDriver) Close() error {
|
||||
return driver.db.Close()
|
||||
}
|
||||
|
||||
// Context satisfies sql.Database
|
||||
func (driver *SQLXDriver) Context() context.Context {
|
||||
return driver.ctx
|
||||
}
|
||||
|
||||
// HasCopy satisfies sql.Database
|
||||
func (driver *SQLXDriver) UseCopyFrom() bool {
|
||||
// sqlx does not currently support COPY.
|
||||
return false
|
||||
}
|
||||
|
||||
type sqlxStatsWrapper struct {
|
||||
stats coresql.DBStats
|
||||
}
|
||||
|
||||
// MaxOpen satisfies metrics.DbStats
|
||||
func (s sqlxStatsWrapper) MaxOpen() int64 {
|
||||
return int64(s.stats.MaxOpenConnections)
|
||||
}
|
||||
|
||||
// Open satisfies metrics.DbStats
|
||||
func (s sqlxStatsWrapper) Open() int64 {
|
||||
return int64(s.stats.OpenConnections)
|
||||
}
|
||||
|
||||
// InUse satisfies metrics.DbStats
|
||||
func (s sqlxStatsWrapper) InUse() int64 {
|
||||
return int64(s.stats.InUse)
|
||||
}
|
||||
|
||||
// Idle satisfies metrics.DbStats
|
||||
func (s sqlxStatsWrapper) Idle() int64 {
|
||||
return int64(s.stats.Idle)
|
||||
}
|
||||
|
||||
// WaitCount satisfies metrics.DbStats
|
||||
func (s sqlxStatsWrapper) WaitCount() int64 {
|
||||
return s.stats.WaitCount
|
||||
}
|
||||
|
||||
// WaitDuration satisfies metrics.DbStats
|
||||
func (s sqlxStatsWrapper) WaitDuration() time.Duration {
|
||||
return s.stats.WaitDuration
|
||||
}
|
||||
|
||||
// MaxIdleClosed satisfies metrics.DbStats
|
||||
func (s sqlxStatsWrapper) MaxIdleClosed() int64 {
|
||||
return s.stats.MaxIdleClosed
|
||||
}
|
||||
|
||||
// MaxLifetimeClosed satisfies metrics.DbStats
|
||||
func (s sqlxStatsWrapper) MaxLifetimeClosed() int64 {
|
||||
return s.stats.MaxLifetimeClosed
|
||||
}
|
||||
|
||||
type sqlxTxWrapper struct {
|
||||
tx *sqlx.Tx
|
||||
}
|
||||
|
||||
// QueryRow satisfies sql.Tx
|
||||
func (t sqlxTxWrapper) QueryRow(ctx context.Context, sql string, args ...interface{}) sql.ScannableRow {
|
||||
return t.tx.QueryRowx(sql, args...)
|
||||
}
|
||||
|
||||
// Exec satisfies sql.Tx
|
||||
func (t sqlxTxWrapper) Exec(ctx context.Context, sql string, args ...interface{}) (sql.Result, error) {
|
||||
return t.tx.Exec(sql, args...)
|
||||
}
|
||||
|
||||
// Commit satisfies sql.Tx
|
||||
func (t sqlxTxWrapper) Commit(ctx context.Context) error {
|
||||
return t.tx.Commit()
|
||||
}
|
||||
|
||||
// Rollback satisfies sql.Tx
|
||||
func (t sqlxTxWrapper) Rollback(ctx context.Context) error {
|
||||
return t.tx.Rollback()
|
||||
}
|
||||
|
||||
func (t sqlxTxWrapper) CopyFrom(ctx context.Context, tableName []string, columnNames []string, rows [][]interface{}) (int64, error) {
|
||||
return 0, errors.New("Unsupported Operation")
|
||||
}
|
119
indexer/database/sql/postgres/sqlx_test.go
Normal file
119
indexer/database/sql/postgres/sqlx_test.go
Normal file
@ -0,0 +1,119 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package postgres_test
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"math/big"
|
||||
"strings"
|
||||
"testing"
|
||||
|
||||
"github.com/jmoiron/sqlx"
|
||||
_ "github.com/lib/pq"
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql/postgres"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/node"
|
||||
)
|
||||
|
||||
func TestPostgresSQLX(t *testing.T) {
|
||||
var sqlxdb *sqlx.DB
|
||||
|
||||
t.Run("connects to the database", func(t *testing.T) {
|
||||
var err error
|
||||
connStr := postgres.TestConfig.DbConnectionString()
|
||||
|
||||
sqlxdb, err = sqlx.Connect("postgres", connStr)
|
||||
if err != nil {
|
||||
t.Fatalf("failed to connect to db with connection string: %s err: %v", connStr, err)
|
||||
}
|
||||
if sqlxdb == nil {
|
||||
t.Fatal("DB is nil")
|
||||
}
|
||||
err = sqlxdb.Close()
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
})
|
||||
|
||||
t.Run("serializes big.Int to db", func(t *testing.T) {
|
||||
// postgres driver doesn't support go big.Int type
|
||||
// various casts in golang uint64, int64, overflow for
|
||||
// transaction value (in wei) even though
|
||||
// postgres numeric can handle an arbitrary
|
||||
// sized int, so use string representation of big.Int
|
||||
// and cast on insert
|
||||
|
||||
connStr := postgres.TestConfig.DbConnectionString()
|
||||
db, err := sqlx.Connect("postgres", connStr)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
defer db.Close()
|
||||
|
||||
bi := new(big.Int)
|
||||
bi.SetString("34940183920000000000", 10)
|
||||
require.Equal(t, "34940183920000000000", bi.String())
|
||||
|
||||
defer db.Exec(`DROP TABLE IF EXISTS example`)
|
||||
_, err = db.Exec("CREATE TABLE example ( id INTEGER, data NUMERIC )")
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
sqlStatement := `
|
||||
INSERT INTO example (id, data)
|
||||
VALUES (1, cast($1 AS NUMERIC))`
|
||||
_, err = db.Exec(sqlStatement, bi.String())
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
var data string
|
||||
err = db.QueryRow(`SELECT data FROM example WHERE id = 1`).Scan(&data)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
require.Equal(t, data, bi.String())
|
||||
actual := new(big.Int)
|
||||
actual.SetString(data, 10)
|
||||
require.Equal(t, bi, actual)
|
||||
})
|
||||
|
||||
t.Run("throws error when can't connect to the database", func(t *testing.T) {
|
||||
goodInfo := node.Info{GenesisBlock: "GENESIS", NetworkID: "1", ID: "x123", ClientName: "geth"}
|
||||
_, err := postgres.NewSQLXDriver(ctx, postgres.Config{}, goodInfo)
|
||||
if err == nil {
|
||||
t.Fatal("Expected an error")
|
||||
}
|
||||
|
||||
expectContainsSubstring(t, err.Error(), postgres.DbConnectionFailedMsg)
|
||||
})
|
||||
|
||||
t.Run("throws error when can't create node", func(t *testing.T) {
|
||||
badHash := fmt.Sprintf("x %s", strings.Repeat("1", 100))
|
||||
badInfo := node.Info{GenesisBlock: badHash, NetworkID: "1", ID: "x123", ClientName: "geth"}
|
||||
|
||||
_, err := postgres.NewSQLXDriver(ctx, postgres.TestConfig, badInfo)
|
||||
if err == nil {
|
||||
t.Fatal("Expected an error")
|
||||
}
|
||||
|
||||
expectContainsSubstring(t, err.Error(), postgres.SettingNodeFailedMsg)
|
||||
})
|
||||
}
|
44
indexer/database/sql/postgres/test_helpers.go
Normal file
44
indexer/database/sql/postgres/test_helpers.go
Normal file
@ -0,0 +1,44 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package postgres
|
||||
|
||||
import (
|
||||
"context"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/node"
|
||||
)
|
||||
|
||||
// SetupSQLXDB is used to setup a sqlx db for tests
|
||||
func SetupSQLXDB() (sql.Database, error) {
|
||||
conf := TestConfig
|
||||
conf.MaxIdle = 0
|
||||
driver, err := NewSQLXDriver(context.Background(), conf, node.Info{})
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return NewPostgresDB(driver, false), nil
|
||||
}
|
||||
|
||||
// SetupPGXDB is used to setup a pgx db for tests
|
||||
func SetupPGXDB(config Config) (sql.Database, error) {
|
||||
driver, err := NewPGXDriver(context.Background(), config, node.Info{})
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return NewPostgresDB(driver, false), nil
|
||||
}
|
52
indexer/database/sql/sqlx_indexer_legacy_test.go
Normal file
52
indexer/database/sql/sqlx_indexer_legacy_test.go
Normal file
@ -0,0 +1,52 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package sql_test
|
||||
|
||||
import (
|
||||
"context"
|
||||
"testing"
|
||||
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql/postgres"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test"
|
||||
)
|
||||
|
||||
func setupLegacySQLXIndexer(t *testing.T) {
|
||||
db, err = postgres.SetupSQLXDB()
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
ind, err = sql.NewStateDiffIndexer(context.Background(), test.LegacyConfig, db)
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
func setupLegacySQLX(t *testing.T) {
|
||||
setupLegacySQLXIndexer(t)
|
||||
test.SetupLegacyTestData(t, ind)
|
||||
}
|
||||
|
||||
func TestLegacySQLXIndexer(t *testing.T) {
|
||||
t.Run("Publish and index header IPLDs", func(t *testing.T) {
|
||||
setupLegacySQLX(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
|
||||
test.TestLegacyIndexer(t, db)
|
||||
})
|
||||
}
|
227
indexer/database/sql/sqlx_indexer_test.go
Normal file
227
indexer/database/sql/sqlx_indexer_test.go
Normal file
@ -0,0 +1,227 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package sql_test
|
||||
|
||||
import (
|
||||
"context"
|
||||
"math/big"
|
||||
"testing"
|
||||
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql/postgres"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/mocks"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/test"
|
||||
)
|
||||
|
||||
func setupSQLXIndexer(t *testing.T) {
|
||||
db, err = postgres.SetupSQLXDB()
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
ind, err = sql.NewStateDiffIndexer(context.Background(), mocks.TestConfig, db)
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
func setupSQLX(t *testing.T) {
|
||||
setupSQLXIndexer(t)
|
||||
test.SetupTestData(t, ind)
|
||||
}
|
||||
|
||||
func setupSQLXNonCanonical(t *testing.T) {
|
||||
setupSQLXIndexer(t)
|
||||
test.SetupTestDataNonCanonical(t, ind)
|
||||
}
|
||||
|
||||
// Test indexer for a canonical block
|
||||
func TestSQLXIndexer(t *testing.T) {
|
||||
t.Run("Publish and index header IPLDs in a single tx", func(t *testing.T) {
|
||||
setupSQLX(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
|
||||
test.TestPublishAndIndexHeaderIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index transaction IPLDs in a single tx", func(t *testing.T) {
|
||||
setupSQLX(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
|
||||
test.TestPublishAndIndexTransactionIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index log IPLDs for multiple receipt of a specific block", func(t *testing.T) {
|
||||
setupSQLX(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
|
||||
test.TestPublishAndIndexLogIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index receipt IPLDs in a single tx", func(t *testing.T) {
|
||||
setupSQLX(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
|
||||
test.TestPublishAndIndexReceiptIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index state IPLDs in a single tx", func(t *testing.T) {
|
||||
setupSQLX(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
|
||||
test.TestPublishAndIndexStateIPLDs(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index storage IPLDs in a single tx", func(t *testing.T) {
|
||||
setupSQLX(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
|
||||
test.TestPublishAndIndexStorageIPLDs(t, db)
|
||||
})
|
||||
}
|
||||
|
||||
// Test indexer for a canonical + a non-canonical block at London height + a non-canonical block at London height + 1
|
||||
func TestSQLXIndexerNonCanonical(t *testing.T) {
|
||||
t.Run("Publish and index header", func(t *testing.T) {
|
||||
setupSQLXNonCanonical(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
|
||||
test.TestPublishAndIndexHeaderNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index transactions", func(t *testing.T) {
|
||||
setupSQLXNonCanonical(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
|
||||
test.TestPublishAndIndexTransactionsNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index receipts", func(t *testing.T) {
|
||||
setupSQLXNonCanonical(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
|
||||
test.TestPublishAndIndexReceiptsNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index logs", func(t *testing.T) {
|
||||
setupSQLXNonCanonical(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
|
||||
test.TestPublishAndIndexLogsNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index state nodes", func(t *testing.T) {
|
||||
setupSQLXNonCanonical(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
|
||||
test.TestPublishAndIndexStateNonCanonical(t, db)
|
||||
})
|
||||
|
||||
t.Run("Publish and index storage nodes", func(t *testing.T) {
|
||||
setupSQLXNonCanonical(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
|
||||
test.TestPublishAndIndexStorageNonCanonical(t, db)
|
||||
})
|
||||
}
|
||||
|
||||
func TestSQLXWatchAddressMethods(t *testing.T) {
|
||||
setupSQLXIndexer(t)
|
||||
defer tearDown(t)
|
||||
defer checkTxClosure(t, 0, 0, 0)
|
||||
|
||||
t.Run("Load watched addresses (empty table)", func(t *testing.T) {
|
||||
test.TestLoadEmptyWatchedAddresses(t, ind)
|
||||
})
|
||||
|
||||
t.Run("Insert watched addresses", func(t *testing.T) {
|
||||
args := mocks.GetInsertWatchedAddressesArgs()
|
||||
err = ind.InsertWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt1)))
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestInsertWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Insert watched addresses (some already watched)", func(t *testing.T) {
|
||||
args := mocks.GetInsertAlreadyWatchedAddressesArgs()
|
||||
err = ind.InsertWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt2)))
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestInsertAlreadyWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Remove watched addresses", func(t *testing.T) {
|
||||
args := mocks.GetRemoveWatchedAddressesArgs()
|
||||
err = ind.RemoveWatchedAddresses(args)
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestRemoveWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Remove watched addresses (some non-watched)", func(t *testing.T) {
|
||||
args := mocks.GetRemoveNonWatchedAddressesArgs()
|
||||
err = ind.RemoveWatchedAddresses(args)
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestRemoveNonWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Set watched addresses", func(t *testing.T) {
|
||||
args := mocks.GetSetWatchedAddressesArgs()
|
||||
err = ind.SetWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt2)))
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestSetWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Set watched addresses (some already watched)", func(t *testing.T) {
|
||||
args := mocks.GetSetAlreadyWatchedAddressesArgs()
|
||||
err = ind.SetWatchedAddresses(args, big.NewInt(int64(mocks.WatchedAt3)))
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestSetAlreadyWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Load watched addresses", func(t *testing.T) {
|
||||
test.TestLoadWatchedAddresses(t, ind)
|
||||
})
|
||||
|
||||
t.Run("Clear watched addresses", func(t *testing.T) {
|
||||
err = ind.ClearWatchedAddresses()
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestClearWatchedAddresses(t, db)
|
||||
})
|
||||
|
||||
t.Run("Clear watched addresses (empty table)", func(t *testing.T) {
|
||||
err = ind.ClearWatchedAddresses()
|
||||
require.NoError(t, err)
|
||||
|
||||
test.TestClearEmptyWatchedAddresses(t, db)
|
||||
})
|
||||
}
|
345
indexer/database/sql/writer.go
Normal file
345
indexer/database/sql/writer.go
Normal file
@ -0,0 +1,345 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package sql
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"strconv"
|
||||
|
||||
"github.com/jackc/pgtype"
|
||||
shopspring "github.com/jackc/pgtype/ext/shopspring-numeric"
|
||||
"github.com/lib/pq"
|
||||
"github.com/shopspring/decimal"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/metrics"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/models"
|
||||
)
|
||||
|
||||
// Writer handles processing and writing of indexed IPLD objects to Postgres
|
||||
type Writer struct {
|
||||
db Database
|
||||
}
|
||||
|
||||
// NewWriter creates a new pointer to a Writer
|
||||
func NewWriter(db Database) *Writer {
|
||||
return &Writer{
|
||||
db: db,
|
||||
}
|
||||
}
|
||||
|
||||
// Close satisfies io.Closer
|
||||
func (w *Writer) Close() error {
|
||||
return w.db.Close()
|
||||
}
|
||||
|
||||
/*
|
||||
INSERT INTO eth.header_cids (block_number, block_hash, parent_hash, cid, td, node_ids, reward, state_root, tx_root, receipt_root, uncles_hash, bloom, timestamp, coinbase)
|
||||
VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14)
|
||||
ON CONFLICT (block_hash, block_number) DO NOTHING
|
||||
*/
|
||||
func (w *Writer) upsertHeaderCID(tx Tx, header models.HeaderModel) error {
|
||||
nodeIDs := pq.StringArray([]string{w.db.NodeID()})
|
||||
_, err := tx.Exec(w.db.Context(), w.db.InsertHeaderStm(),
|
||||
header.BlockNumber,
|
||||
header.BlockHash,
|
||||
header.ParentHash,
|
||||
header.CID,
|
||||
header.TotalDifficulty,
|
||||
nodeIDs,
|
||||
header.Reward,
|
||||
header.StateRoot,
|
||||
header.TxRoot,
|
||||
header.RctRoot,
|
||||
header.UnclesHash,
|
||||
header.Bloom,
|
||||
header.Timestamp,
|
||||
header.Coinbase)
|
||||
if err != nil {
|
||||
return insertError{"eth.header_cids", err, w.db.InsertHeaderStm(), header}
|
||||
}
|
||||
metrics.IndexerMetrics.BlocksCounter.Inc(1)
|
||||
return nil
|
||||
}
|
||||
|
||||
/*
|
||||
INSERT INTO eth.uncle_cids (block_number, block_hash, header_id, parent_hash, cid, reward, index) VALUES ($1, $2, $3, $4, $5, $6, $7)
|
||||
ON CONFLICT (block_hash, block_number) DO NOTHING
|
||||
*/
|
||||
func (w *Writer) upsertUncleCID(tx Tx, uncle models.UncleModel) error {
|
||||
_, err := tx.Exec(w.db.Context(), w.db.InsertUncleStm(),
|
||||
uncle.BlockNumber,
|
||||
uncle.BlockHash,
|
||||
uncle.HeaderID,
|
||||
uncle.ParentHash,
|
||||
uncle.CID,
|
||||
uncle.Reward,
|
||||
uncle.Index)
|
||||
if err != nil {
|
||||
return insertError{"eth.uncle_cids", err, w.db.InsertUncleStm(), uncle}
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
/*
|
||||
INSERT INTO eth.transaction_cids (block_number, header_id, tx_hash, cid, dst, src, index, tx_type, value) VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9)
|
||||
ON CONFLICT (tx_hash, header_id, block_number) DO NOTHING
|
||||
*/
|
||||
func (w *Writer) upsertTransactionCID(tx Tx, transaction models.TxModel) error {
|
||||
val := transaction.Value
|
||||
if val == "" {
|
||||
val = "0"
|
||||
}
|
||||
if w.useCopyForTx(tx) {
|
||||
blockNum, err := strconv.ParseInt(transaction.BlockNumber, 10, 64)
|
||||
if err != nil {
|
||||
return insertError{"eth.transaction_cids", err, "COPY", transaction}
|
||||
}
|
||||
|
||||
value, err := toNumeric(val)
|
||||
if err != nil {
|
||||
return insertError{"eth.transaction_cids", err, "COPY", transaction}
|
||||
}
|
||||
|
||||
_, err = tx.CopyFrom(w.db.Context(), w.db.TxTableName(), w.db.TxColumnNames(),
|
||||
toRows(toRow(blockNum, transaction.HeaderID, transaction.TxHash, transaction.CID, transaction.Dst,
|
||||
transaction.Src, transaction.Index, int(transaction.Type), value)))
|
||||
if err != nil {
|
||||
return insertError{"eth.transaction_cids", err, "COPY", transaction}
|
||||
}
|
||||
} else {
|
||||
_, err := tx.Exec(w.db.Context(), w.db.InsertTxStm(),
|
||||
transaction.BlockNumber,
|
||||
transaction.HeaderID,
|
||||
transaction.TxHash,
|
||||
transaction.CID,
|
||||
transaction.Dst,
|
||||
transaction.Src,
|
||||
transaction.Index,
|
||||
transaction.Type,
|
||||
transaction.Value)
|
||||
if err != nil {
|
||||
return insertError{"eth.transaction_cids", err, w.db.InsertTxStm(), transaction}
|
||||
}
|
||||
}
|
||||
metrics.IndexerMetrics.TransactionsCounter.Inc(1)
|
||||
return nil
|
||||
}
|
||||
|
||||
/*
|
||||
INSERT INTO eth.receipt_cids (block_number, header_id, tx_id, cid, contract, post_state, post_status) VALUES ($1, $2, $3, $4, $5, $6, $7, $8)
|
||||
ON CONFLICT (tx_id, header_id, block_number) DO NOTHING
|
||||
*/
|
||||
func (w *Writer) upsertReceiptCID(tx Tx, rct *models.ReceiptModel) error {
|
||||
if w.useCopyForTx(tx) {
|
||||
blockNum, err := strconv.ParseUint(rct.BlockNumber, 10, 64)
|
||||
if err != nil {
|
||||
return insertError{"eth.receipt_cids", err, "COPY", rct}
|
||||
}
|
||||
|
||||
_, err = tx.CopyFrom(w.db.Context(), w.db.RctTableName(), w.db.RctColumnNames(),
|
||||
toRows(toRow(blockNum, rct.HeaderID, rct.TxID, rct.CID, rct.Contract,
|
||||
rct.PostState, int(rct.PostStatus))))
|
||||
if err != nil {
|
||||
return insertError{"eth.receipt_cids", err, "COPY", rct}
|
||||
}
|
||||
} else {
|
||||
_, err := tx.Exec(w.db.Context(), w.db.InsertRctStm(),
|
||||
rct.BlockNumber,
|
||||
rct.HeaderID,
|
||||
rct.TxID,
|
||||
rct.CID,
|
||||
rct.Contract,
|
||||
rct.PostState,
|
||||
rct.PostStatus)
|
||||
if err != nil {
|
||||
return insertError{"eth.receipt_cids", err, w.db.InsertRctStm(), *rct}
|
||||
}
|
||||
}
|
||||
metrics.IndexerMetrics.ReceiptsCounter.Inc(1)
|
||||
return nil
|
||||
}
|
||||
|
||||
/*
|
||||
INSERT INTO eth.log_cids (block_number, header_id, cid, rct_id, address, index, topic0, topic1, topic2, topic3) VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, $10)
|
||||
ON CONFLICT (rct_id, index, header_id, block_number) DO NOTHING
|
||||
*/
|
||||
func (w *Writer) upsertLogCID(tx Tx, logs []*models.LogsModel) error {
|
||||
if w.useCopyForTx(tx) {
|
||||
var rows [][]interface{}
|
||||
for _, log := range logs {
|
||||
blockNum, err := strconv.ParseUint(log.BlockNumber, 10, 64)
|
||||
if err != nil {
|
||||
return insertError{"eth.log_cids", err, "COPY", log}
|
||||
}
|
||||
|
||||
rows = append(rows, toRow(blockNum, log.HeaderID, log.CID, log.ReceiptID,
|
||||
log.Address, log.Index, log.Topic0, log.Topic1, log.Topic2, log.Topic3))
|
||||
}
|
||||
if nil != rows && len(rows) >= 0 {
|
||||
_, err := tx.CopyFrom(w.db.Context(), w.db.LogTableName(), w.db.LogColumnNames(), rows)
|
||||
if err != nil {
|
||||
return insertError{"eth.log_cids", err, "COPY", rows}
|
||||
}
|
||||
metrics.IndexerMetrics.LogsCounter.Inc(int64(len(rows)))
|
||||
}
|
||||
} else {
|
||||
for _, log := range logs {
|
||||
_, err := tx.Exec(w.db.Context(), w.db.InsertLogStm(),
|
||||
log.BlockNumber,
|
||||
log.HeaderID,
|
||||
log.CID,
|
||||
log.ReceiptID,
|
||||
log.Address,
|
||||
log.Index,
|
||||
log.Topic0,
|
||||
log.Topic1,
|
||||
log.Topic2,
|
||||
log.Topic3)
|
||||
if err != nil {
|
||||
return insertError{"eth.log_cids", err, w.db.InsertLogStm(), *log}
|
||||
}
|
||||
metrics.IndexerMetrics.LogsCounter.Inc(1)
|
||||
}
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
/*
|
||||
INSERT INTO eth.state_cids (block_number, header_id, state_leaf_key, cid, removed, diff, balance, nonce, code_hash, storage_root) VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, $10)
|
||||
ON CONFLICT (header_id, state_leaf_key, block_number) DO NOTHING
|
||||
*/
|
||||
func (w *Writer) upsertStateCID(tx Tx, stateNode models.StateNodeModel) error {
|
||||
bal := stateNode.Balance
|
||||
if stateNode.Removed {
|
||||
bal = "0"
|
||||
}
|
||||
|
||||
if w.useCopyForTx(tx) {
|
||||
blockNum, err := strconv.ParseUint(stateNode.BlockNumber, 10, 64)
|
||||
if err != nil {
|
||||
return insertError{"eth.state_cids", err, "COPY", stateNode}
|
||||
}
|
||||
|
||||
balance, err := toNumeric(bal)
|
||||
if err != nil {
|
||||
return insertError{"eth.state_cids", err, "COPY", stateNode}
|
||||
}
|
||||
|
||||
_, err = tx.CopyFrom(w.db.Context(), w.db.StateTableName(), w.db.StateColumnNames(),
|
||||
toRows(toRow(blockNum, stateNode.HeaderID, stateNode.StateKey, stateNode.CID,
|
||||
true, balance, stateNode.Nonce, stateNode.CodeHash, stateNode.StorageRoot, stateNode.Removed)))
|
||||
if err != nil {
|
||||
return insertError{"eth.state_cids", err, "COPY", stateNode}
|
||||
}
|
||||
} else {
|
||||
_, err := tx.Exec(w.db.Context(), w.db.InsertStateStm(),
|
||||
stateNode.BlockNumber,
|
||||
stateNode.HeaderID,
|
||||
stateNode.StateKey,
|
||||
stateNode.CID,
|
||||
true,
|
||||
bal,
|
||||
stateNode.Nonce,
|
||||
stateNode.CodeHash,
|
||||
stateNode.StorageRoot,
|
||||
stateNode.Removed,
|
||||
)
|
||||
if err != nil {
|
||||
return insertError{"eth.state_cids", err, w.db.InsertStateStm(), stateNode}
|
||||
}
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
/*
|
||||
INSERT INTO eth.storage_cids (block_number, header_id, state_leaf_key, storage_leaf_key, cid, removed, diff, val) VALUES ($1, $2, $3, $4, $5, $6, $7, $8)
|
||||
ON CONFLICT (header_id, state_leaf_key, storage_leaf_key, block_number) DO NOTHING
|
||||
*/
|
||||
func (w *Writer) upsertStorageCID(tx Tx, storageCID models.StorageNodeModel) error {
|
||||
if w.useCopyForTx(tx) {
|
||||
blockNum, err := strconv.ParseUint(storageCID.BlockNumber, 10, 64)
|
||||
if err != nil {
|
||||
return insertError{"eth.storage_cids", err, "COPY", storageCID}
|
||||
}
|
||||
|
||||
_, err = tx.CopyFrom(w.db.Context(), w.db.StorageTableName(), w.db.StorageColumnNames(),
|
||||
toRows(toRow(blockNum, storageCID.HeaderID, storageCID.StateKey, storageCID.StorageKey, storageCID.CID,
|
||||
true, storageCID.Value, storageCID.Removed)))
|
||||
if err != nil {
|
||||
return insertError{"eth.storage_cids", err, "COPY", storageCID}
|
||||
}
|
||||
} else {
|
||||
_, err := tx.Exec(w.db.Context(), w.db.InsertStorageStm(),
|
||||
storageCID.BlockNumber,
|
||||
storageCID.HeaderID,
|
||||
storageCID.StateKey,
|
||||
storageCID.StorageKey,
|
||||
storageCID.CID,
|
||||
true,
|
||||
storageCID.Value,
|
||||
storageCID.Removed,
|
||||
)
|
||||
if err != nil {
|
||||
return insertError{"eth.storage_cids", err, w.db.InsertStorageStm(), storageCID}
|
||||
}
|
||||
}
|
||||
return nil
|
||||
}
|
||||
|
||||
func (w *Writer) useCopyForTx(tx Tx) bool {
|
||||
// Using COPY instead of INSERT only makes much sense if also using a DelayedTx, so that operations
|
||||
// can be collected over time and then all submitted within in a single TX.
|
||||
if _, ok := tx.(*DelayedTx); ok {
|
||||
return w.db.UseCopyFrom()
|
||||
}
|
||||
return false
|
||||
}
|
||||
|
||||
// combine args into a row
|
||||
func toRow(args ...interface{}) []interface{} {
|
||||
var row []interface{}
|
||||
row = append(row, args...)
|
||||
return row
|
||||
}
|
||||
|
||||
func toNumeric(value string) (*shopspring.Numeric, error) {
|
||||
decimalValue, err := decimal.NewFromString(value)
|
||||
if nil != err {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
return &shopspring.Numeric{Decimal: decimalValue, Status: pgtype.Present}, nil
|
||||
}
|
||||
|
||||
// combine row (or rows) into a slice of rows for CopyFrom
|
||||
func toRows(rows ...[]interface{}) [][]interface{} {
|
||||
return rows
|
||||
}
|
||||
|
||||
type insertError struct {
|
||||
table string
|
||||
err error
|
||||
stmt string
|
||||
arguments interface{}
|
||||
}
|
||||
|
||||
var _ error = insertError{}
|
||||
|
||||
func (dbe insertError) Error() string {
|
||||
return fmt.Sprintf("error inserting %s entry: %v\r\nstatement: %s\r\narguments: %+v",
|
||||
dbe.table, dbe.err, dbe.stmt, dbe.arguments)
|
||||
}
|
55
indexer/interfaces/interfaces.go
Normal file
55
indexer/interfaces/interfaces.go
Normal file
@ -0,0 +1,55 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package interfaces
|
||||
|
||||
import (
|
||||
"io"
|
||||
"math/big"
|
||||
"time"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/shared"
|
||||
sdtypes "github.com/ethereum/go-ethereum/statediff/types"
|
||||
)
|
||||
|
||||
// StateDiffIndexer interface required to index statediff data
|
||||
type StateDiffIndexer interface {
|
||||
PushBlock(block *types.Block, receipts types.Receipts, totalDifficulty *big.Int) (Batch, error)
|
||||
PushStateNode(tx Batch, stateNode sdtypes.StateLeafNode, headerID string) error
|
||||
PushIPLD(tx Batch, ipld sdtypes.IPLD) error
|
||||
ReportDBMetrics(delay time.Duration, quit <-chan bool)
|
||||
|
||||
// Methods used by WatchAddress API/functionality
|
||||
LoadWatchedAddresses() ([]common.Address, error)
|
||||
InsertWatchedAddresses(addresses []sdtypes.WatchAddressArg, currentBlock *big.Int) error
|
||||
RemoveWatchedAddresses(addresses []sdtypes.WatchAddressArg) error
|
||||
SetWatchedAddresses(args []sdtypes.WatchAddressArg, currentBlockNumber *big.Int) error
|
||||
ClearWatchedAddresses() error
|
||||
|
||||
io.Closer
|
||||
}
|
||||
|
||||
// Batch required for indexing data atomically
|
||||
type Batch interface {
|
||||
Submit(err error) error
|
||||
}
|
||||
|
||||
// Config used to configure different underlying implementations
|
||||
type Config interface {
|
||||
Type() shared.DBType
|
||||
}
|
BIN
indexer/ipld/eip2930_test_data/eth-block-12252078
Normal file
BIN
indexer/ipld/eip2930_test_data/eth-block-12252078
Normal file
Binary file not shown.
BIN
indexer/ipld/eip2930_test_data/eth-block-12365585
Normal file
BIN
indexer/ipld/eip2930_test_data/eth-block-12365585
Normal file
Binary file not shown.
BIN
indexer/ipld/eip2930_test_data/eth-block-12365586
Normal file
BIN
indexer/ipld/eip2930_test_data/eth-block-12365586
Normal file
Binary file not shown.
BIN
indexer/ipld/eip2930_test_data/eth-receipts-12252078
Normal file
BIN
indexer/ipld/eip2930_test_data/eth-receipts-12252078
Normal file
Binary file not shown.
BIN
indexer/ipld/eip2930_test_data/eth-receipts-12365585
Normal file
BIN
indexer/ipld/eip2930_test_data/eth-receipts-12365585
Normal file
Binary file not shown.
BIN
indexer/ipld/eip2930_test_data/eth-receipts-12365586
Normal file
BIN
indexer/ipld/eip2930_test_data/eth-receipts-12365586
Normal file
Binary file not shown.
60
indexer/ipld/eth_header.go
Normal file
60
indexer/ipld/eth_header.go
Normal file
@ -0,0 +1,60 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package ipld
|
||||
|
||||
import (
|
||||
"github.com/ipfs/go-cid"
|
||||
mh "github.com/multiformats/go-multihash"
|
||||
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/rlp"
|
||||
)
|
||||
|
||||
// EthHeader (eth-block, codec 0x90), represents an ethereum block header
|
||||
type EthHeader struct {
|
||||
cid cid.Cid
|
||||
rawdata []byte
|
||||
}
|
||||
|
||||
// Static (compile time) check that EthHeader satisfies the node.Node interface.
|
||||
var _ IPLD = (*EthHeader)(nil)
|
||||
|
||||
// NewEthHeader converts a *types.Header into an EthHeader IPLD node
|
||||
func NewEthHeader(header *types.Header) (*EthHeader, error) {
|
||||
headerRLP, err := rlp.EncodeToBytes(header)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
c, err := RawdataToCid(MEthHeader, headerRLP, mh.KECCAK_256)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return &EthHeader{
|
||||
cid: c,
|
||||
rawdata: headerRLP,
|
||||
}, nil
|
||||
}
|
||||
|
||||
// RawData returns the binary of the RLP encode of the block header.
|
||||
func (b *EthHeader) RawData() []byte {
|
||||
return b.rawdata
|
||||
}
|
||||
|
||||
// Cid returns the cid of the block header.
|
||||
func (b *EthHeader) Cid() cid.Cid {
|
||||
return b.cid
|
||||
}
|
44
indexer/ipld/eth_log.go
Normal file
44
indexer/ipld/eth_log.go
Normal file
@ -0,0 +1,44 @@
|
||||
package ipld
|
||||
|
||||
import (
|
||||
"github.com/ipfs/go-cid"
|
||||
mh "github.com/multiformats/go-multihash"
|
||||
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/rlp"
|
||||
)
|
||||
|
||||
// EthLog (eth-log, codec 0x9a), represents an ethereum block header
|
||||
type EthLog struct {
|
||||
rawData []byte
|
||||
cid cid.Cid
|
||||
}
|
||||
|
||||
// Static (compile time) check that EthLog satisfies the node.Node interface.
|
||||
var _ IPLD = (*EthLog)(nil)
|
||||
|
||||
// NewLog create a new EthLog IPLD node
|
||||
func NewLog(log *types.Log) (*EthLog, error) {
|
||||
logRaw, err := rlp.EncodeToBytes(log)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
c, err := RawdataToCid(MEthLog, logRaw, mh.KECCAK_256)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return &EthLog{
|
||||
cid: c,
|
||||
rawData: logRaw,
|
||||
}, nil
|
||||
}
|
||||
|
||||
// RawData returns the binary of the RLP encode of the log.
|
||||
func (l *EthLog) RawData() []byte {
|
||||
return l.rawData
|
||||
}
|
||||
|
||||
// Cid returns the cid of the receipt log.
|
||||
func (l *EthLog) Cid() cid.Cid {
|
||||
return l.cid
|
||||
}
|
94
indexer/ipld/eth_parser.go
Normal file
94
indexer/ipld/eth_parser.go
Normal file
@ -0,0 +1,94 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package ipld
|
||||
|
||||
import (
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
)
|
||||
|
||||
// FromBlockAndReceipts takes a block and processes it
|
||||
// to return it a set of IPLD nodes for further processing.
|
||||
func FromBlockAndReceipts(block *types.Block, receipts []*types.Receipt) (*EthHeader, []*EthTx, []*EthReceipt, [][]*EthLog, error) {
|
||||
// Process the header
|
||||
headerNode, err := NewEthHeader(block.Header())
|
||||
if err != nil {
|
||||
return nil, nil, nil, nil, err
|
||||
}
|
||||
|
||||
// Process the txs
|
||||
txNodes, err := processTransactions(block.Transactions())
|
||||
if err != nil {
|
||||
return nil, nil, nil, nil, err
|
||||
}
|
||||
|
||||
// Process the receipts and logs
|
||||
rctNodes, logNodes, err := processReceiptsAndLogs(receipts)
|
||||
|
||||
return headerNode, txNodes, rctNodes, logNodes, err
|
||||
}
|
||||
|
||||
// processTransactions will take the found transactions in a parsed block body
|
||||
// to return IPLD node slices for eth-tx
|
||||
func processTransactions(txs []*types.Transaction) ([]*EthTx, error) {
|
||||
var ethTxNodes []*EthTx
|
||||
for _, tx := range txs {
|
||||
ethTx, err := NewEthTx(tx)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
ethTxNodes = append(ethTxNodes, ethTx)
|
||||
}
|
||||
|
||||
return ethTxNodes, nil
|
||||
}
|
||||
|
||||
// processReceiptsAndLogs will take in receipts
|
||||
// to return IPLD node slices for eth-rct and eth-log
|
||||
func processReceiptsAndLogs(rcts []*types.Receipt) ([]*EthReceipt, [][]*EthLog, error) {
|
||||
// Pre allocating memory.
|
||||
ethRctNodes := make([]*EthReceipt, len(rcts))
|
||||
ethLogNodes := make([][]*EthLog, len(rcts))
|
||||
|
||||
for idx, rct := range rcts {
|
||||
logNodes, err := processLogs(rct.Logs)
|
||||
if err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
|
||||
ethRct, err := NewReceipt(rct)
|
||||
if err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
|
||||
ethRctNodes[idx] = ethRct
|
||||
ethLogNodes[idx] = logNodes
|
||||
}
|
||||
|
||||
return ethRctNodes, ethLogNodes, nil
|
||||
}
|
||||
|
||||
func processLogs(logs []*types.Log) ([]*EthLog, error) {
|
||||
logNodes := make([]*EthLog, len(logs))
|
||||
for idx, log := range logs {
|
||||
logNode, err := NewLog(log)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
logNodes[idx] = logNode
|
||||
}
|
||||
return logNodes, nil
|
||||
}
|
126
indexer/ipld/eth_parser_test.go
Normal file
126
indexer/ipld/eth_parser_test.go
Normal file
@ -0,0 +1,126 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package ipld
|
||||
|
||||
import (
|
||||
"os"
|
||||
"path/filepath"
|
||||
"testing"
|
||||
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/rlp"
|
||||
)
|
||||
|
||||
type kind string
|
||||
|
||||
const (
|
||||
legacy kind = "legacy"
|
||||
eip1559 kind = "eip2930"
|
||||
)
|
||||
|
||||
var blockFileNames = []string{
|
||||
"eth-block-12252078",
|
||||
"eth-block-12365585",
|
||||
"eth-block-12365586",
|
||||
}
|
||||
|
||||
var receiptsFileNames = []string{
|
||||
"eth-receipts-12252078",
|
||||
"eth-receipts-12365585",
|
||||
"eth-receipts-12365586",
|
||||
}
|
||||
|
||||
var kinds = []kind{
|
||||
eip1559,
|
||||
eip1559,
|
||||
legacy,
|
||||
}
|
||||
|
||||
type testCase struct {
|
||||
kind kind
|
||||
block *types.Block
|
||||
receipts types.Receipts
|
||||
}
|
||||
|
||||
func loadBlockData(t *testing.T) []testCase {
|
||||
fileDir := "./eip2930_test_data"
|
||||
testCases := make([]testCase, len(blockFileNames))
|
||||
for i, blockFileName := range blockFileNames {
|
||||
blockRLP, err := os.ReadFile(filepath.Join(fileDir, blockFileName))
|
||||
if err != nil {
|
||||
t.Fatalf("failed to load blockRLP from file, err %v", err)
|
||||
}
|
||||
block := new(types.Block)
|
||||
if err := rlp.DecodeBytes(blockRLP, block); err != nil {
|
||||
t.Fatalf("failed to decode blockRLP, err %v", err)
|
||||
}
|
||||
receiptsFileName := receiptsFileNames[i]
|
||||
receiptsRLP, err := os.ReadFile(filepath.Join(fileDir, receiptsFileName))
|
||||
if err != nil {
|
||||
t.Fatalf("failed to load receiptsRLP from file, err %s", err)
|
||||
}
|
||||
receipts := make(types.Receipts, 0)
|
||||
if err := rlp.DecodeBytes(receiptsRLP, &receipts); err != nil {
|
||||
t.Fatalf("failed to decode receiptsRLP, err %s", err)
|
||||
}
|
||||
testCases[i] = testCase{
|
||||
block: block,
|
||||
receipts: receipts,
|
||||
kind: kinds[i],
|
||||
}
|
||||
}
|
||||
return testCases
|
||||
}
|
||||
|
||||
func TestFromBlockAndReceipts(t *testing.T) {
|
||||
testCases := loadBlockData(t)
|
||||
for _, tc := range testCases {
|
||||
_, _, _, _, err := FromBlockAndReceipts(tc.block, tc.receipts)
|
||||
if err != nil {
|
||||
t.Fatalf("error generating IPLDs from block and receipts, err %v, kind %s, block hash %s", err, tc.kind, tc.block.Hash())
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
func TestProcessLogs(t *testing.T) {
|
||||
logs := []*types.Log{mockLog1, mockLog2}
|
||||
nodes, err := processLogs(logs)
|
||||
require.NoError(t, err)
|
||||
require.GreaterOrEqual(t, len(nodes), len(logs))
|
||||
}
|
||||
|
||||
var (
|
||||
address = common.HexToAddress("0xaE9BEa628c4Ce503DcFD7E305CaB4e29E7476592")
|
||||
anotherAddress = common.HexToAddress("0xaE9BEa628c4Ce503DcFD7E305CaB4e29E7476593")
|
||||
mockTopic11 = common.HexToHash("0x04")
|
||||
mockTopic12 = common.HexToHash("0x06")
|
||||
mockTopic21 = common.HexToHash("0x05")
|
||||
mockTopic22 = common.HexToHash("0x07")
|
||||
mockLog1 = &types.Log{
|
||||
Address: address,
|
||||
Topics: []common.Hash{mockTopic11, mockTopic12},
|
||||
Data: []byte{},
|
||||
}
|
||||
mockLog2 = &types.Log{
|
||||
Address: anotherAddress,
|
||||
Topics: []common.Hash{mockTopic21, mockTopic22},
|
||||
Data: []byte{},
|
||||
}
|
||||
)
|
58
indexer/ipld/eth_receipt.go
Normal file
58
indexer/ipld/eth_receipt.go
Normal file
@ -0,0 +1,58 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package ipld
|
||||
|
||||
import (
|
||||
"github.com/ipfs/go-cid"
|
||||
mh "github.com/multiformats/go-multihash"
|
||||
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
)
|
||||
|
||||
type EthReceipt struct {
|
||||
rawdata []byte
|
||||
cid cid.Cid
|
||||
}
|
||||
|
||||
// Static (compile time) check that EthReceipt satisfies the node.Node interface.
|
||||
var _ IPLD = (*EthReceipt)(nil)
|
||||
|
||||
// NewReceipt converts a types.ReceiptForStorage to an EthReceipt IPLD node
|
||||
func NewReceipt(receipt *types.Receipt) (*EthReceipt, error) {
|
||||
rctRaw, err := receipt.MarshalBinary()
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
c, err := RawdataToCid(MEthTxReceipt, rctRaw, mh.KECCAK_256)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return &EthReceipt{
|
||||
cid: c,
|
||||
rawdata: rctRaw,
|
||||
}, nil
|
||||
}
|
||||
|
||||
// RawData returns the binary of the RLP encode of the receipt.
|
||||
func (r *EthReceipt) RawData() []byte {
|
||||
return r.rawdata
|
||||
}
|
||||
|
||||
// Cid returns the cid of the receipt.
|
||||
func (r *EthReceipt) Cid() cid.Cid {
|
||||
return r.cid
|
||||
}
|
59
indexer/ipld/eth_tx.go
Normal file
59
indexer/ipld/eth_tx.go
Normal file
@ -0,0 +1,59 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package ipld
|
||||
|
||||
import (
|
||||
"github.com/ipfs/go-cid"
|
||||
mh "github.com/multiformats/go-multihash"
|
||||
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
)
|
||||
|
||||
// EthTx (eth-tx codec 0x93) represents an ethereum transaction
|
||||
type EthTx struct {
|
||||
cid cid.Cid
|
||||
rawdata []byte
|
||||
}
|
||||
|
||||
// Static (compile time) check that EthTx satisfies the node.Node interface.
|
||||
var _ IPLD = (*EthTx)(nil)
|
||||
|
||||
// NewEthTx converts a *types.Transaction to an EthTx IPLD node
|
||||
func NewEthTx(tx *types.Transaction) (*EthTx, error) {
|
||||
txRaw, err := tx.MarshalBinary()
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
c, err := RawdataToCid(MEthTx, txRaw, mh.KECCAK_256)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
return &EthTx{
|
||||
cid: c,
|
||||
rawdata: txRaw,
|
||||
}, nil
|
||||
}
|
||||
|
||||
// RawData returns the binary of the RLP encode of the transaction.
|
||||
func (t *EthTx) RawData() []byte {
|
||||
return t.rawdata
|
||||
}
|
||||
|
||||
// Cid returns the cid of the transaction.
|
||||
func (t *EthTx) Cid() cid.Cid {
|
||||
return t.cid
|
||||
}
|
8
indexer/ipld/interface.go
Normal file
8
indexer/ipld/interface.go
Normal file
@ -0,0 +1,8 @@
|
||||
package ipld
|
||||
|
||||
import "github.com/ipfs/go-cid"
|
||||
|
||||
type IPLD interface {
|
||||
Cid() cid.Cid
|
||||
RawData() []byte
|
||||
}
|
66
indexer/ipld/shared.go
Normal file
66
indexer/ipld/shared.go
Normal file
@ -0,0 +1,66 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package ipld
|
||||
|
||||
import (
|
||||
"github.com/ipfs/go-cid"
|
||||
mh "github.com/multiformats/go-multihash"
|
||||
)
|
||||
|
||||
// IPLD Codecs for Ethereum
|
||||
// See the authoritative document:
|
||||
// https://github.com/multiformats/multicodec/blob/master/table.csv
|
||||
const (
|
||||
RawBinary = 0x55
|
||||
MEthHeader = 0x90
|
||||
MEthHeaderList = 0x91
|
||||
MEthTxTrie = 0x92
|
||||
MEthTx = 0x93
|
||||
MEthTxReceiptTrie = 0x94
|
||||
MEthTxReceipt = 0x95
|
||||
MEthStateTrie = 0x96
|
||||
MEthAccountSnapshot = 0x97
|
||||
MEthStorageTrie = 0x98
|
||||
MEthLogTrie = 0x99
|
||||
MEthLog = 0x9a
|
||||
)
|
||||
|
||||
// RawdataToCid takes the desired codec and a slice of bytes
|
||||
// and returns the proper cid of the object.
|
||||
func RawdataToCid(codec uint64, rawdata []byte, multiHash uint64) (cid.Cid, error) {
|
||||
c, err := cid.Prefix{
|
||||
Codec: codec,
|
||||
Version: 1,
|
||||
MhType: multiHash,
|
||||
MhLength: -1,
|
||||
}.Sum(rawdata)
|
||||
if err != nil {
|
||||
return cid.Cid{}, err
|
||||
}
|
||||
return c, nil
|
||||
}
|
||||
|
||||
// Keccak256ToCid takes a keccak256 hash and returns its cid based on
|
||||
// the codec given.
|
||||
func Keccak256ToCid(codec uint64, h []byte) cid.Cid {
|
||||
buf, err := mh.Encode(h, mh.KECCAK_256)
|
||||
if err != nil {
|
||||
panic(err)
|
||||
}
|
||||
|
||||
return cid.NewCidV1(codec, buf)
|
||||
}
|
BIN
indexer/mainnet_data/block_12579670.rlp
Normal file
BIN
indexer/mainnet_data/block_12579670.rlp
Normal file
Binary file not shown.
BIN
indexer/mainnet_data/block_12600011.rlp
Normal file
BIN
indexer/mainnet_data/block_12600011.rlp
Normal file
Binary file not shown.
BIN
indexer/mainnet_data/block_12619985.rlp
Normal file
BIN
indexer/mainnet_data/block_12619985.rlp
Normal file
Binary file not shown.
BIN
indexer/mainnet_data/block_12625121.rlp
Normal file
BIN
indexer/mainnet_data/block_12625121.rlp
Normal file
Binary file not shown.
BIN
indexer/mainnet_data/block_12655432.rlp
Normal file
BIN
indexer/mainnet_data/block_12655432.rlp
Normal file
Binary file not shown.
BIN
indexer/mainnet_data/block_12914664.rlp
Normal file
BIN
indexer/mainnet_data/block_12914664.rlp
Normal file
Binary file not shown.
BIN
indexer/mainnet_data/receipts_12579670.rlp
Normal file
BIN
indexer/mainnet_data/receipts_12579670.rlp
Normal file
Binary file not shown.
BIN
indexer/mainnet_data/receipts_12600011.rlp
Normal file
BIN
indexer/mainnet_data/receipts_12600011.rlp
Normal file
Binary file not shown.
BIN
indexer/mainnet_data/receipts_12619985.rlp
Normal file
BIN
indexer/mainnet_data/receipts_12619985.rlp
Normal file
Binary file not shown.
BIN
indexer/mainnet_data/receipts_12625121.rlp
Normal file
BIN
indexer/mainnet_data/receipts_12625121.rlp
Normal file
Binary file not shown.
BIN
indexer/mainnet_data/receipts_12655432.rlp
Normal file
BIN
indexer/mainnet_data/receipts_12655432.rlp
Normal file
Binary file not shown.
BIN
indexer/mainnet_data/receipts_12914664.rlp
Normal file
BIN
indexer/mainnet_data/receipts_12914664.rlp
Normal file
Binary file not shown.
638
indexer/mocks/test_data.go
Normal file
638
indexer/mocks/test_data.go
Normal file
@ -0,0 +1,638 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package mocks
|
||||
|
||||
import (
|
||||
"crypto/ecdsa"
|
||||
"crypto/elliptic"
|
||||
"crypto/rand"
|
||||
"math/big"
|
||||
|
||||
ipld2 "github.com/ethereum/go-ethereum/statediff/indexer/ipld"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/shared"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/crypto"
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
"github.com/ethereum/go-ethereum/params"
|
||||
"github.com/ethereum/go-ethereum/rlp"
|
||||
"github.com/ethereum/go-ethereum/statediff/test_helpers"
|
||||
sdtypes "github.com/ethereum/go-ethereum/statediff/types"
|
||||
"github.com/ethereum/go-ethereum/trie"
|
||||
)
|
||||
|
||||
// Test variables
|
||||
var (
|
||||
// block data
|
||||
TestConfig = params.MainnetChainConfig
|
||||
BlockNumber = TestConfig.LondonBlock
|
||||
|
||||
// canonical block at London height
|
||||
// includes 5 transactions: 3 Legacy + 1 EIP-2930 + 1 EIP-1559
|
||||
MockHeader = types.Header{
|
||||
Time: 0,
|
||||
Number: new(big.Int).Set(BlockNumber),
|
||||
Root: common.HexToHash("0x0"),
|
||||
TxHash: common.HexToHash("0x0"),
|
||||
ReceiptHash: common.HexToHash("0x0"),
|
||||
Difficulty: big.NewInt(5000000),
|
||||
Extra: []byte{},
|
||||
BaseFee: big.NewInt(params.InitialBaseFee),
|
||||
Coinbase: common.HexToAddress("0xaE9BEa628c4Ce503DcFD7E305CaB4e29E7476777"),
|
||||
}
|
||||
MockTransactions, MockReceipts, SenderAddr = createTransactionsAndReceipts(TestConfig, BlockNumber)
|
||||
MockBlock = types.NewBlock(&MockHeader, MockTransactions, nil, MockReceipts, trie.NewEmpty(nil))
|
||||
MockHeaderRlp, _ = rlp.EncodeToBytes(MockBlock.Header())
|
||||
|
||||
// non-canonical block at London height
|
||||
// includes 2nd and 5th transactions from the canonical block
|
||||
MockNonCanonicalHeader = MockHeader
|
||||
MockNonCanonicalBlockTransactions = types.Transactions{MockTransactions[1], MockTransactions[4]}
|
||||
MockNonCanonicalBlockReceipts = createNonCanonicalBlockReceipts(TestConfig, BlockNumber, MockNonCanonicalBlockTransactions)
|
||||
MockNonCanonicalBlock = types.NewBlock(&MockNonCanonicalHeader, MockNonCanonicalBlockTransactions, nil, MockNonCanonicalBlockReceipts, trie.NewEmpty(nil))
|
||||
MockNonCanonicalHeaderRlp, _ = rlp.EncodeToBytes(MockNonCanonicalBlock.Header())
|
||||
|
||||
// non-canonical block at London height + 1
|
||||
// includes 3rd and 5th transactions from the canonical block
|
||||
Block2Number = big.NewInt(BlockNumber.Int64() + 1)
|
||||
MockNonCanonicalHeader2 = types.Header{
|
||||
Time: 0,
|
||||
Number: new(big.Int).Set(Block2Number),
|
||||
Root: common.HexToHash("0x0"),
|
||||
TxHash: common.HexToHash("0x0"),
|
||||
ReceiptHash: common.HexToHash("0x0"),
|
||||
Difficulty: big.NewInt(6000000),
|
||||
Extra: []byte{},
|
||||
BaseFee: big.NewInt(params.InitialBaseFee),
|
||||
Coinbase: common.HexToAddress("0xaE9BEa628c4Ce503DcFD7E305CaB4e29E7476777"),
|
||||
}
|
||||
MockNonCanonicalBlock2Transactions = types.Transactions{MockTransactions[2], MockTransactions[4]}
|
||||
MockNonCanonicalBlock2Receipts = createNonCanonicalBlockReceipts(TestConfig, Block2Number, MockNonCanonicalBlock2Transactions)
|
||||
MockNonCanonicalBlock2 = types.NewBlock(&MockNonCanonicalHeader2, MockNonCanonicalBlock2Transactions, nil, MockNonCanonicalBlock2Receipts, trie.NewEmpty(nil))
|
||||
MockNonCanonicalHeader2Rlp, _ = rlp.EncodeToBytes(MockNonCanonicalBlock2.Header())
|
||||
|
||||
Address = common.HexToAddress("0xaE9BEa628c4Ce503DcFD7E305CaB4e29E7476592")
|
||||
AnotherAddress = common.HexToAddress("0xaE9BEa628c4Ce503DcFD7E305CaB4e29E7476593")
|
||||
ContractAddress = crypto.CreateAddress(SenderAddr, MockTransactions[2].Nonce())
|
||||
ContractAddress2 = crypto.CreateAddress(SenderAddr, MockTransactions[3].Nonce())
|
||||
MockContractByteCode = []byte{0, 1, 2, 3, 4, 5}
|
||||
mockTopic11 = common.HexToHash("0x04")
|
||||
mockTopic12 = common.HexToHash("0x06")
|
||||
mockTopic21 = common.HexToHash("0x05")
|
||||
mockTopic22 = common.HexToHash("0x07")
|
||||
ExpectedPostStatus uint64 = 1
|
||||
ExpectedPostState1 = common.HexToHash("0x1").String()
|
||||
ExpectedPostState2 = common.HexToHash("0x2").String()
|
||||
ExpectedPostState3 = common.HexToHash("0x3").String()
|
||||
MockLog1 = &types.Log{
|
||||
Address: Address,
|
||||
Topics: []common.Hash{mockTopic11, mockTopic12},
|
||||
Data: []byte{},
|
||||
}
|
||||
MockLog2 = &types.Log{
|
||||
Address: AnotherAddress,
|
||||
Topics: []common.Hash{mockTopic21, mockTopic22},
|
||||
Data: []byte{},
|
||||
}
|
||||
MockLog3 = &types.Log{
|
||||
Address: Address,
|
||||
Topics: []common.Hash{mockTopic11, mockTopic22},
|
||||
Data: []byte{},
|
||||
}
|
||||
MockLog4 = &types.Log{
|
||||
Address: AnotherAddress,
|
||||
Topics: []common.Hash{mockTopic21, mockTopic12},
|
||||
Data: []byte{},
|
||||
}
|
||||
ShortLog1 = &types.Log{
|
||||
Address: AnotherAddress,
|
||||
Topics: []common.Hash{},
|
||||
Data: []byte{},
|
||||
}
|
||||
ShortLog2 = &types.Log{
|
||||
Address: Address,
|
||||
Topics: []common.Hash{},
|
||||
Data: []byte{},
|
||||
}
|
||||
|
||||
// access list entries
|
||||
AccessListEntry1 = types.AccessTuple{
|
||||
Address: Address,
|
||||
}
|
||||
AccessListEntry2 = types.AccessTuple{
|
||||
Address: AnotherAddress,
|
||||
StorageKeys: []common.Hash{common.BytesToHash(StorageLeafKey), common.BytesToHash(MockStorageLeafKey)},
|
||||
}
|
||||
|
||||
// statediff data
|
||||
storageLocation = common.HexToHash("0")
|
||||
StorageLeafKey = crypto.Keccak256Hash(storageLocation[:]).Bytes()
|
||||
mockStorageLocation = common.HexToHash("1")
|
||||
MockStorageLeafKey = crypto.Keccak256Hash(mockStorageLocation[:]).Bytes()
|
||||
StorageValue = common.Hex2Bytes("01")
|
||||
StoragePartialPath = common.Hex2Bytes("20290decd9548b62a8d60345a988386fc84ba6bc95484008f6362f93160ef3e563")
|
||||
StorageLeafNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
StoragePartialPath,
|
||||
StorageValue,
|
||||
})
|
||||
StorageLeafNodeCID = ipld2.Keccak256ToCid(ipld2.MEthStorageTrie, crypto.Keccak256(StorageLeafNode)).String()
|
||||
|
||||
nonce1 = uint64(1)
|
||||
ContractRoot = "0x821e2556a290c86405f8160a2d662042a431ba456b9db265c79bb837c04be5f0"
|
||||
ContractCodeHash = common.HexToHash("0x753f98a8d4328b15636e46f66f2cb4bc860100aa17967cc145fcd17d1d4710ea")
|
||||
ContractLeafKey = test_helpers.AddressToLeafKey(ContractAddress)
|
||||
ContractAccount = &types.StateAccount{
|
||||
Nonce: nonce1,
|
||||
Balance: big.NewInt(0),
|
||||
CodeHash: ContractCodeHash.Bytes(),
|
||||
Root: common.HexToHash(ContractRoot),
|
||||
}
|
||||
ContractAccountRLP, _ = rlp.EncodeToBytes(ContractAccount)
|
||||
|
||||
ContractPartialPath = common.Hex2Bytes("3114658a74d9cc9f7acf2c5cd696c3494d7c344d78bfec3add0d91ec4e8d1c45")
|
||||
ContractLeafNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
ContractPartialPath,
|
||||
ContractAccount,
|
||||
})
|
||||
ContractLeafNodeCID = ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(ContractLeafNode)).String()
|
||||
|
||||
Contract2LeafKey = test_helpers.AddressToLeafKey(ContractAddress2)
|
||||
storage2Location = common.HexToHash("2")
|
||||
Storage2LeafKey = crypto.Keccak256Hash(storage2Location[:]).Bytes()
|
||||
storage3Location = common.HexToHash("3")
|
||||
Storage3LeafKey = crypto.Keccak256Hash(storage3Location[:]).Bytes()
|
||||
|
||||
nonce0 = uint64(0)
|
||||
AccountRoot = "0x56e81f171bcc55a6ff8345e692c0f86e5b48e01b996cadc001622fb5e363b421"
|
||||
AccountCodeHash = common.HexToHash("0xc5d2460186f7233c927e7db2dcc703c0e500b653ca82273b7bfad8045d85a470")
|
||||
AccountLeafKey = test_helpers.Account2LeafKey
|
||||
RemovedLeafKey = test_helpers.Account1LeafKey
|
||||
Balance, _ = new(big.Int).SetString("106387458790507306766", 10)
|
||||
Account = &types.StateAccount{
|
||||
Nonce: nonce0,
|
||||
Balance: Balance,
|
||||
CodeHash: AccountCodeHash.Bytes(),
|
||||
Root: common.HexToHash(AccountRoot),
|
||||
}
|
||||
AccountRLP, _ = rlp.EncodeToBytes(Account)
|
||||
AccountPartialPath = common.Hex2Bytes("3957f3e2f04a0764c3a0491b175f69926da61efbcc8f61fa1455fd2d2b4cdd45")
|
||||
AccountLeafNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
AccountPartialPath,
|
||||
Account,
|
||||
})
|
||||
AccountLeafNodeCID = ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(AccountLeafNode)).String()
|
||||
|
||||
StateDiffs = []sdtypes.StateLeafNode{
|
||||
{
|
||||
AccountWrapper: sdtypes.AccountWrapper{
|
||||
Account: ContractAccount,
|
||||
LeafKey: ContractLeafKey,
|
||||
CID: ContractLeafNodeCID,
|
||||
},
|
||||
Removed: false,
|
||||
StorageDiff: []sdtypes.StorageLeafNode{
|
||||
{
|
||||
Removed: false,
|
||||
LeafKey: StorageLeafKey,
|
||||
Value: StorageValue,
|
||||
CID: StorageLeafNodeCID,
|
||||
},
|
||||
{
|
||||
Removed: true,
|
||||
LeafKey: RemovedLeafKey,
|
||||
CID: shared.RemovedNodeStorageCID,
|
||||
Value: []byte{},
|
||||
},
|
||||
},
|
||||
},
|
||||
{
|
||||
AccountWrapper: sdtypes.AccountWrapper{
|
||||
Account: Account,
|
||||
LeafKey: AccountLeafKey,
|
||||
CID: AccountLeafNodeCID,
|
||||
},
|
||||
Removed: false,
|
||||
StorageDiff: []sdtypes.StorageLeafNode{},
|
||||
},
|
||||
{
|
||||
AccountWrapper: sdtypes.AccountWrapper{
|
||||
Account: nil,
|
||||
LeafKey: RemovedLeafKey,
|
||||
CID: shared.RemovedNodeStateCID,
|
||||
},
|
||||
Removed: true,
|
||||
StorageDiff: []sdtypes.StorageLeafNode{},
|
||||
},
|
||||
{
|
||||
AccountWrapper: sdtypes.AccountWrapper{
|
||||
Account: nil,
|
||||
LeafKey: Contract2LeafKey,
|
||||
CID: shared.RemovedNodeStateCID,
|
||||
},
|
||||
Removed: true,
|
||||
StorageDiff: []sdtypes.StorageLeafNode{
|
||||
{
|
||||
Removed: true,
|
||||
CID: shared.RemovedNodeStorageCID,
|
||||
LeafKey: Storage2LeafKey,
|
||||
Value: []byte{},
|
||||
},
|
||||
{
|
||||
Removed: true,
|
||||
CID: shared.RemovedNodeStorageCID,
|
||||
LeafKey: Storage3LeafKey,
|
||||
Value: []byte{},
|
||||
},
|
||||
},
|
||||
},
|
||||
}
|
||||
|
||||
IPLDs = []sdtypes.IPLD{
|
||||
{
|
||||
CID: ContractLeafNodeCID,
|
||||
Content: ContractLeafNode,
|
||||
},
|
||||
{
|
||||
CID: StorageLeafNodeCID,
|
||||
Content: StorageLeafNode,
|
||||
},
|
||||
{
|
||||
CID: shared.RemovedNodeStorageCID,
|
||||
Content: []byte{},
|
||||
},
|
||||
{
|
||||
CID: AccountLeafNodeCID,
|
||||
Content: AccountLeafNode,
|
||||
},
|
||||
{
|
||||
CID: shared.RemovedNodeStateCID,
|
||||
Content: []byte{},
|
||||
},
|
||||
}
|
||||
|
||||
// Mock data for testing watched addresses methods
|
||||
Contract1Address = "0x5d663F5269090bD2A7DC2390c911dF6083D7b28F"
|
||||
Contract2Address = "0x6Eb7e5C66DB8af2E96159AC440cbc8CDB7fbD26B"
|
||||
Contract3Address = "0xcfeB164C328CA13EFd3C77E1980d94975aDfedfc"
|
||||
Contract4Address = "0x0Edf0c4f393a628DE4828B228C48175b3EA297fc"
|
||||
Contract1CreatedAt = uint64(1)
|
||||
Contract2CreatedAt = uint64(2)
|
||||
Contract3CreatedAt = uint64(3)
|
||||
Contract4CreatedAt = uint64(4)
|
||||
|
||||
LastFilledAt = uint64(0)
|
||||
WatchedAt1 = uint64(10)
|
||||
WatchedAt2 = uint64(15)
|
||||
WatchedAt3 = uint64(20)
|
||||
)
|
||||
|
||||
type LegacyData struct {
|
||||
Config *params.ChainConfig
|
||||
BlockNumber *big.Int
|
||||
MockHeader types.Header
|
||||
MockTransactions types.Transactions
|
||||
MockReceipts types.Receipts
|
||||
SenderAddr common.Address
|
||||
MockBlock *types.Block
|
||||
MockHeaderRlp []byte
|
||||
Address []byte
|
||||
AnotherAddress []byte
|
||||
ContractAddress common.Address
|
||||
MockContractByteCode []byte
|
||||
MockLog1 *types.Log
|
||||
MockLog2 *types.Log
|
||||
StorageLeafKey []byte
|
||||
MockStorageLeafKey []byte
|
||||
StorageLeafNode []byte
|
||||
ContractLeafKey []byte
|
||||
ContractAccount []byte
|
||||
ContractPartialPath []byte
|
||||
ContractLeafNode []byte
|
||||
AccountRoot string
|
||||
AccountLeafNode []byte
|
||||
StateDiffs []sdtypes.StateLeafNode
|
||||
}
|
||||
|
||||
func NewLegacyData(config *params.ChainConfig) *LegacyData {
|
||||
// Block number before london fork.
|
||||
blockNumber := config.EIP155Block
|
||||
|
||||
mockHeader := types.Header{
|
||||
Time: 0,
|
||||
Number: new(big.Int).Set(blockNumber),
|
||||
Root: common.HexToHash("0x0"),
|
||||
TxHash: common.HexToHash("0x0"),
|
||||
ReceiptHash: common.HexToHash("0x0"),
|
||||
Difficulty: big.NewInt(5000000),
|
||||
Extra: []byte{},
|
||||
Coinbase: common.HexToAddress("0xaE9BEa628c4Ce503DcFD7E305CaB4e29E7476888"),
|
||||
}
|
||||
|
||||
mockTransactions, mockReceipts, senderAddr := createLegacyTransactionsAndReceipts(config, blockNumber)
|
||||
mockBlock := types.NewBlock(&mockHeader, mockTransactions, nil, mockReceipts, trie.NewEmpty(nil))
|
||||
mockHeaderRlp, _ := rlp.EncodeToBytes(mockBlock.Header())
|
||||
contractAddress := crypto.CreateAddress(senderAddr, mockTransactions[2].Nonce())
|
||||
|
||||
return &LegacyData{
|
||||
Config: config,
|
||||
BlockNumber: blockNumber,
|
||||
MockHeader: mockHeader,
|
||||
MockTransactions: mockTransactions,
|
||||
MockReceipts: mockReceipts,
|
||||
SenderAddr: senderAddr,
|
||||
MockBlock: mockBlock,
|
||||
MockHeaderRlp: mockHeaderRlp,
|
||||
ContractAddress: contractAddress,
|
||||
MockContractByteCode: MockContractByteCode,
|
||||
MockLog1: MockLog1,
|
||||
MockLog2: MockLog2,
|
||||
StorageLeafKey: StorageLeafKey,
|
||||
MockStorageLeafKey: MockStorageLeafKey,
|
||||
StorageLeafNode: StorageLeafNode,
|
||||
ContractLeafKey: ContractLeafKey,
|
||||
ContractAccount: ContractAccountRLP,
|
||||
ContractPartialPath: ContractPartialPath,
|
||||
ContractLeafNode: ContractLeafNode,
|
||||
AccountRoot: AccountRoot,
|
||||
AccountLeafNode: AccountLeafKey,
|
||||
StateDiffs: StateDiffs,
|
||||
}
|
||||
}
|
||||
|
||||
// createLegacyTransactionsAndReceipts is a helper function to generate signed mock legacy transactions and mock receipts with mock logs
|
||||
func createLegacyTransactionsAndReceipts(config *params.ChainConfig, blockNumber *big.Int) (types.Transactions, types.Receipts, common.Address) {
|
||||
// make transactions
|
||||
trx1 := types.NewTransaction(0, Address, big.NewInt(1000), 50, big.NewInt(100), []byte{})
|
||||
trx2 := types.NewTransaction(1, AnotherAddress, big.NewInt(2000), 100, big.NewInt(200), []byte{})
|
||||
trx3 := types.NewContractCreation(2, big.NewInt(1500), 75, big.NewInt(150), MockContractByteCode)
|
||||
|
||||
transactionSigner := types.MakeSigner(config, blockNumber)
|
||||
mockCurve := elliptic.P256()
|
||||
mockPrvKey, err := ecdsa.GenerateKey(mockCurve, rand.Reader)
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
signedTrx1, err := types.SignTx(trx1, transactionSigner, mockPrvKey)
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
signedTrx2, err := types.SignTx(trx2, transactionSigner, mockPrvKey)
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
signedTrx3, err := types.SignTx(trx3, transactionSigner, mockPrvKey)
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
|
||||
senderAddr, err := types.Sender(transactionSigner, signedTrx1) // same for both trx
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
|
||||
// make receipts
|
||||
mockReceipt1 := types.NewReceipt(nil, false, 50)
|
||||
mockReceipt1.Logs = []*types.Log{MockLog1}
|
||||
mockReceipt1.TxHash = signedTrx1.Hash()
|
||||
mockReceipt2 := types.NewReceipt(common.HexToHash("0x1").Bytes(), false, 100)
|
||||
mockReceipt2.Logs = []*types.Log{MockLog2, ShortLog1}
|
||||
mockReceipt2.TxHash = signedTrx2.Hash()
|
||||
mockReceipt3 := types.NewReceipt(common.HexToHash("0x2").Bytes(), false, 75)
|
||||
mockReceipt3.Logs = []*types.Log{}
|
||||
mockReceipt3.TxHash = signedTrx3.Hash()
|
||||
|
||||
return types.Transactions{signedTrx1, signedTrx2, signedTrx3}, types.Receipts{mockReceipt1, mockReceipt2, mockReceipt3}, senderAddr
|
||||
}
|
||||
|
||||
// createTransactionsAndReceipts is a helper function to generate signed mock transactions and mock receipts with mock logs
|
||||
func createTransactionsAndReceipts(config *params.ChainConfig, blockNumber *big.Int) (types.Transactions, types.Receipts, common.Address) {
|
||||
// make transactions
|
||||
trx1 := types.NewTransaction(0, Address, big.NewInt(1000), 50, big.NewInt(100), []byte{})
|
||||
trx2 := types.NewTransaction(1, AnotherAddress, big.NewInt(2000), 100, big.NewInt(200), []byte{})
|
||||
trx3 := types.NewContractCreation(2, big.NewInt(1500), 75, big.NewInt(150), MockContractByteCode)
|
||||
trx4 := types.NewTx(&types.AccessListTx{
|
||||
ChainID: config.ChainID,
|
||||
Nonce: 0,
|
||||
GasPrice: big.NewInt(100),
|
||||
Gas: 50,
|
||||
To: &AnotherAddress,
|
||||
Value: big.NewInt(999),
|
||||
Data: []byte{},
|
||||
AccessList: types.AccessList{
|
||||
AccessListEntry1,
|
||||
AccessListEntry2,
|
||||
},
|
||||
})
|
||||
trx5 := types.NewTx(&types.DynamicFeeTx{
|
||||
ChainID: config.ChainID,
|
||||
Nonce: 0,
|
||||
GasTipCap: big.NewInt(100),
|
||||
GasFeeCap: big.NewInt(100),
|
||||
Gas: 50,
|
||||
To: &AnotherAddress,
|
||||
Value: big.NewInt(1000),
|
||||
Data: []byte{},
|
||||
AccessList: types.AccessList{
|
||||
AccessListEntry1,
|
||||
AccessListEntry2,
|
||||
},
|
||||
})
|
||||
|
||||
transactionSigner := types.MakeSigner(config, blockNumber)
|
||||
mockCurve := elliptic.P256()
|
||||
mockPrvKey, err := ecdsa.GenerateKey(mockCurve, rand.Reader)
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
signedTrx1, err := types.SignTx(trx1, transactionSigner, mockPrvKey)
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
signedTrx2, err := types.SignTx(trx2, transactionSigner, mockPrvKey)
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
signedTrx3, err := types.SignTx(trx3, transactionSigner, mockPrvKey)
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
signedTrx4, err := types.SignTx(trx4, transactionSigner, mockPrvKey)
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
signedTrx5, err := types.SignTx(trx5, transactionSigner, mockPrvKey)
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
|
||||
senderAddr, err := types.Sender(transactionSigner, signedTrx1) // same for both trx
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
|
||||
// make receipts
|
||||
mockReceipt1 := types.NewReceipt(nil, false, 50)
|
||||
mockReceipt1.Logs = []*types.Log{MockLog1}
|
||||
mockReceipt1.TxHash = signedTrx1.Hash()
|
||||
mockReceipt2 := types.NewReceipt(common.HexToHash("0x1").Bytes(), false, 100)
|
||||
mockReceipt2.Logs = []*types.Log{MockLog2, ShortLog1}
|
||||
mockReceipt2.TxHash = signedTrx2.Hash()
|
||||
mockReceipt3 := types.NewReceipt(common.HexToHash("0x2").Bytes(), false, 75)
|
||||
mockReceipt3.Logs = []*types.Log{}
|
||||
mockReceipt3.TxHash = signedTrx3.Hash()
|
||||
mockReceipt4 := &types.Receipt{
|
||||
Type: types.AccessListTxType,
|
||||
PostState: common.HexToHash("0x3").Bytes(),
|
||||
Status: types.ReceiptStatusSuccessful,
|
||||
CumulativeGasUsed: 175,
|
||||
Logs: []*types.Log{MockLog3, MockLog4, ShortLog2},
|
||||
TxHash: signedTrx4.Hash(),
|
||||
}
|
||||
mockReceipt5 := &types.Receipt{
|
||||
Type: types.DynamicFeeTxType,
|
||||
PostState: common.HexToHash("0x3").Bytes(),
|
||||
Status: types.ReceiptStatusSuccessful,
|
||||
CumulativeGasUsed: 175,
|
||||
Logs: []*types.Log{},
|
||||
TxHash: signedTrx5.Hash(),
|
||||
}
|
||||
|
||||
return types.Transactions{signedTrx1, signedTrx2, signedTrx3, signedTrx4, signedTrx5}, types.Receipts{mockReceipt1, mockReceipt2, mockReceipt3, mockReceipt4, mockReceipt5}, senderAddr
|
||||
}
|
||||
|
||||
// createNonCanonicalBlockReceipts is a helper function to generate mock receipts with mock logs for non-canonical blocks
|
||||
func createNonCanonicalBlockReceipts(config *params.ChainConfig, blockNumber *big.Int, transactions types.Transactions) types.Receipts {
|
||||
transactionSigner := types.MakeSigner(config, blockNumber)
|
||||
mockCurve := elliptic.P256()
|
||||
mockPrvKey, err := ecdsa.GenerateKey(mockCurve, rand.Reader)
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
|
||||
signedTrx0, err := types.SignTx(transactions[0], transactionSigner, mockPrvKey)
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
|
||||
signedTrx1, err := types.SignTx(transactions[1], transactionSigner, mockPrvKey)
|
||||
if err != nil {
|
||||
log.Crit(err.Error())
|
||||
}
|
||||
|
||||
mockReceipt0 := types.NewReceipt(common.HexToHash("0x3").Bytes(), false, 300)
|
||||
mockReceipt0.Logs = []*types.Log{MockLog1, ShortLog1}
|
||||
mockReceipt0.TxHash = signedTrx0.Hash()
|
||||
|
||||
mockReceipt1 := &types.Receipt{
|
||||
Type: types.DynamicFeeTxType,
|
||||
PostState: common.HexToHash("0x4").Bytes(),
|
||||
Status: types.ReceiptStatusSuccessful,
|
||||
CumulativeGasUsed: 300,
|
||||
Logs: []*types.Log{},
|
||||
TxHash: signedTrx1.Hash(),
|
||||
}
|
||||
|
||||
return types.Receipts{mockReceipt0, mockReceipt1}
|
||||
}
|
||||
|
||||
// Helper methods for testing watched addresses methods
|
||||
func GetInsertWatchedAddressesArgs() []sdtypes.WatchAddressArg {
|
||||
return []sdtypes.WatchAddressArg{
|
||||
{
|
||||
Address: Contract1Address,
|
||||
CreatedAt: Contract1CreatedAt,
|
||||
},
|
||||
{
|
||||
Address: Contract2Address,
|
||||
CreatedAt: Contract2CreatedAt,
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
func GetInsertAlreadyWatchedAddressesArgs() []sdtypes.WatchAddressArg {
|
||||
return []sdtypes.WatchAddressArg{
|
||||
{
|
||||
Address: Contract3Address,
|
||||
CreatedAt: Contract3CreatedAt,
|
||||
},
|
||||
{
|
||||
Address: Contract2Address,
|
||||
CreatedAt: Contract2CreatedAt,
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
func GetRemoveWatchedAddressesArgs() []sdtypes.WatchAddressArg {
|
||||
return []sdtypes.WatchAddressArg{
|
||||
{
|
||||
Address: Contract3Address,
|
||||
CreatedAt: Contract3CreatedAt,
|
||||
},
|
||||
{
|
||||
Address: Contract2Address,
|
||||
CreatedAt: Contract2CreatedAt,
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
func GetRemoveNonWatchedAddressesArgs() []sdtypes.WatchAddressArg {
|
||||
return []sdtypes.WatchAddressArg{
|
||||
{
|
||||
Address: Contract1Address,
|
||||
CreatedAt: Contract1CreatedAt,
|
||||
},
|
||||
{
|
||||
Address: Contract2Address,
|
||||
CreatedAt: Contract2CreatedAt,
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
func GetSetWatchedAddressesArgs() []sdtypes.WatchAddressArg {
|
||||
return []sdtypes.WatchAddressArg{
|
||||
{
|
||||
Address: Contract1Address,
|
||||
CreatedAt: Contract1CreatedAt,
|
||||
},
|
||||
{
|
||||
Address: Contract2Address,
|
||||
CreatedAt: Contract2CreatedAt,
|
||||
},
|
||||
{
|
||||
Address: Contract3Address,
|
||||
CreatedAt: Contract3CreatedAt,
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
func GetSetAlreadyWatchedAddressesArgs() []sdtypes.WatchAddressArg {
|
||||
return []sdtypes.WatchAddressArg{
|
||||
{
|
||||
Address: Contract4Address,
|
||||
CreatedAt: Contract4CreatedAt,
|
||||
},
|
||||
{
|
||||
Address: Contract2Address,
|
||||
CreatedAt: Contract2CreatedAt,
|
||||
},
|
||||
{
|
||||
Address: Contract3Address,
|
||||
CreatedAt: Contract3CreatedAt,
|
||||
},
|
||||
}
|
||||
}
|
24
indexer/models/batch.go
Normal file
24
indexer/models/batch.go
Normal file
@ -0,0 +1,24 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package models
|
||||
|
||||
// IPLDBatch holds the arguments for a batch insert of IPLD data
|
||||
type IPLDBatch struct {
|
||||
BlockNumbers []string
|
||||
Keys []string
|
||||
Values [][]byte
|
||||
}
|
119
indexer/models/models.go
Normal file
119
indexer/models/models.go
Normal file
@ -0,0 +1,119 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package models
|
||||
|
||||
import "github.com/lib/pq"
|
||||
|
||||
// IPLDModel is the db model for ipld.blocks
|
||||
type IPLDModel struct {
|
||||
BlockNumber string `db:"block_number"`
|
||||
Key string `db:"key"`
|
||||
Data []byte `db:"data"`
|
||||
}
|
||||
|
||||
// HeaderModel is the db model for eth.header_cids
|
||||
type HeaderModel struct {
|
||||
BlockNumber string `db:"block_number"`
|
||||
BlockHash string `db:"block_hash"`
|
||||
ParentHash string `db:"parent_hash"`
|
||||
CID string `db:"cid"`
|
||||
TotalDifficulty string `db:"td"`
|
||||
NodeIDs pq.StringArray `db:"node_ids"`
|
||||
Reward string `db:"reward"`
|
||||
StateRoot string `db:"state_root"`
|
||||
UnclesHash string `db:"uncles_hash"`
|
||||
TxRoot string `db:"tx_root"`
|
||||
RctRoot string `db:"receipt_root"`
|
||||
Bloom []byte `db:"bloom"`
|
||||
Timestamp uint64 `db:"timestamp"`
|
||||
Coinbase string `db:"coinbase"`
|
||||
}
|
||||
|
||||
// UncleModel is the db model for eth.uncle_cids
|
||||
type UncleModel struct {
|
||||
BlockNumber string `db:"block_number"`
|
||||
HeaderID string `db:"header_id"`
|
||||
BlockHash string `db:"block_hash"`
|
||||
ParentHash string `db:"parent_hash"`
|
||||
CID string `db:"cid"`
|
||||
Reward string `db:"reward"`
|
||||
Index int64 `db:"index"`
|
||||
}
|
||||
|
||||
// TxModel is the db model for eth.transaction_cids
|
||||
type TxModel struct {
|
||||
BlockNumber string `db:"block_number"`
|
||||
HeaderID string `db:"header_id"`
|
||||
Index int64 `db:"index"`
|
||||
TxHash string `db:"tx_hash"`
|
||||
CID string `db:"cid"`
|
||||
Dst string `db:"dst"`
|
||||
Src string `db:"src"`
|
||||
Type uint8 `db:"tx_type"`
|
||||
Value string `db:"value"`
|
||||
}
|
||||
|
||||
// ReceiptModel is the db model for eth.receipt_cids
|
||||
type ReceiptModel struct {
|
||||
BlockNumber string `db:"block_number"`
|
||||
HeaderID string `db:"header_id"`
|
||||
TxID string `db:"tx_id"`
|
||||
CID string `db:"cid"`
|
||||
PostStatus uint64 `db:"post_status"`
|
||||
PostState string `db:"post_state"`
|
||||
Contract string `db:"contract"`
|
||||
}
|
||||
|
||||
// StateNodeModel is the db model for eth.state_cids
|
||||
type StateNodeModel struct {
|
||||
BlockNumber string `db:"block_number"`
|
||||
HeaderID string `db:"header_id"`
|
||||
StateKey string `db:"state_leaf_key"`
|
||||
Removed bool `db:"removed"`
|
||||
CID string `db:"cid"`
|
||||
Diff bool `db:"diff"`
|
||||
Balance string `db:"balance"`
|
||||
Nonce uint64 `db:"nonce"`
|
||||
CodeHash string `db:"code_hash"`
|
||||
StorageRoot string `db:"storage_root"`
|
||||
}
|
||||
|
||||
// StorageNodeModel is the db model for eth.storage_cids
|
||||
type StorageNodeModel struct {
|
||||
BlockNumber string `db:"block_number"`
|
||||
HeaderID string `db:"header_id"`
|
||||
StateKey string `db:"state_leaf_key"`
|
||||
StorageKey string `db:"storage_leaf_key"`
|
||||
Removed bool `db:"removed"`
|
||||
CID string `db:"cid"`
|
||||
Diff bool `db:"diff"`
|
||||
Value []byte `db:"val"`
|
||||
}
|
||||
|
||||
// LogsModel is the db model for eth.logs
|
||||
type LogsModel struct {
|
||||
BlockNumber string `db:"block_number"`
|
||||
HeaderID string `db:"header_id"`
|
||||
ReceiptID string `db:"rct_id"`
|
||||
CID string `db:"cid"`
|
||||
Address string `db:"address"`
|
||||
Index int64 `db:"index"`
|
||||
Topic0 string `db:"topic0"`
|
||||
Topic1 string `db:"topic1"`
|
||||
Topic2 string `db:"topic2"`
|
||||
Topic3 string `db:"topic3"`
|
||||
}
|
25
indexer/node/node.go
Normal file
25
indexer/node/node.go
Normal file
@ -0,0 +1,25 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package node
|
||||
|
||||
type Info struct {
|
||||
GenesisBlock string
|
||||
NetworkID string
|
||||
ChainID uint64
|
||||
ID string
|
||||
ClientName string
|
||||
}
|
22
indexer/shared/constants.go
Normal file
22
indexer/shared/constants.go
Normal file
@ -0,0 +1,22 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package shared
|
||||
|
||||
const (
|
||||
RemovedNodeStorageCID = "bagmacgzayxjemamg64rtzet6pwznzrydydsqbnstzkbcoo337lmaixmfurya"
|
||||
RemovedNodeStateCID = "baglacgzayxjemamg64rtzet6pwznzrydydsqbnstzkbcoo337lmaixmfurya"
|
||||
)
|
46
indexer/shared/db_kind.go
Normal file
46
indexer/shared/db_kind.go
Normal file
@ -0,0 +1,46 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package shared
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"strings"
|
||||
)
|
||||
|
||||
// DBType to explicitly type the kind of DB
|
||||
type DBType string
|
||||
|
||||
const (
|
||||
POSTGRES DBType = "Postgres"
|
||||
DUMP DBType = "Dump"
|
||||
FILE DBType = "File"
|
||||
UNKNOWN DBType = "Unknown"
|
||||
)
|
||||
|
||||
// ResolveDBType resolves a DBType from a provided string
|
||||
func ResolveDBType(str string) (DBType, error) {
|
||||
switch strings.ToLower(str) {
|
||||
case "postgres", "pg":
|
||||
return POSTGRES, nil
|
||||
case "dump", "d":
|
||||
return DUMP, nil
|
||||
case "file", "f", "fs":
|
||||
return FILE, nil
|
||||
default:
|
||||
return UNKNOWN, fmt.Errorf("unrecognized db type string: %s", str)
|
||||
}
|
||||
}
|
37
indexer/shared/functions.go
Normal file
37
indexer/shared/functions.go
Normal file
@ -0,0 +1,37 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package shared
|
||||
|
||||
import (
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
)
|
||||
|
||||
// HandleZeroAddrPointer will return an empty string for a nil address pointer
|
||||
func HandleZeroAddrPointer(to *common.Address) string {
|
||||
if to == nil {
|
||||
return ""
|
||||
}
|
||||
return to.Hex()
|
||||
}
|
||||
|
||||
// HandleZeroAddr will return an empty string for a 0 value address
|
||||
func HandleZeroAddr(to common.Address) string {
|
||||
if to.Hex() == "0x0000000000000000000000000000000000000000" {
|
||||
return ""
|
||||
}
|
||||
return to.Hex()
|
||||
}
|
76
indexer/shared/reward.go
Normal file
76
indexer/shared/reward.go
Normal file
@ -0,0 +1,76 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package shared
|
||||
|
||||
import (
|
||||
"math/big"
|
||||
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
)
|
||||
|
||||
func CalcEthBlockReward(header *types.Header, uncles []*types.Header, txs types.Transactions, receipts types.Receipts) *big.Int {
|
||||
staticBlockReward := staticRewardByBlockNumber(header.Number.Uint64())
|
||||
transactionFees := calcEthTransactionFees(txs, receipts)
|
||||
uncleInclusionRewards := calcEthUncleInclusionRewards(header, uncles)
|
||||
tmp := transactionFees.Add(transactionFees, uncleInclusionRewards)
|
||||
return tmp.Add(tmp, staticBlockReward)
|
||||
}
|
||||
|
||||
func CalcUncleMinerReward(blockNumber, uncleBlockNumber uint64) *big.Int {
|
||||
staticBlockReward := staticRewardByBlockNumber(blockNumber)
|
||||
rewardDiv8 := staticBlockReward.Div(staticBlockReward, big.NewInt(8))
|
||||
mainBlock := new(big.Int).SetUint64(blockNumber)
|
||||
uncleBlock := new(big.Int).SetUint64(uncleBlockNumber)
|
||||
uncleBlockPlus8 := uncleBlock.Add(uncleBlock, big.NewInt(8))
|
||||
uncleBlockPlus8MinusMainBlock := uncleBlockPlus8.Sub(uncleBlockPlus8, mainBlock)
|
||||
return rewardDiv8.Mul(rewardDiv8, uncleBlockPlus8MinusMainBlock)
|
||||
}
|
||||
|
||||
func staticRewardByBlockNumber(blockNumber uint64) *big.Int {
|
||||
staticBlockReward := new(big.Int)
|
||||
//https://blog.ethereum.org/2017/10/12/byzantium-hf-announcement/
|
||||
if blockNumber >= 7280000 {
|
||||
staticBlockReward.SetString("2000000000000000000", 10)
|
||||
} else if blockNumber >= 4370000 {
|
||||
staticBlockReward.SetString("3000000000000000000", 10)
|
||||
} else {
|
||||
staticBlockReward.SetString("5000000000000000000", 10)
|
||||
}
|
||||
return staticBlockReward
|
||||
}
|
||||
|
||||
func calcEthTransactionFees(txs types.Transactions, receipts types.Receipts) *big.Int {
|
||||
transactionFees := new(big.Int)
|
||||
for i, transaction := range txs {
|
||||
receipt := receipts[i]
|
||||
gasPrice := big.NewInt(transaction.GasPrice().Int64())
|
||||
gasUsed := big.NewInt(int64(receipt.GasUsed))
|
||||
transactionFee := gasPrice.Mul(gasPrice, gasUsed)
|
||||
transactionFees = transactionFees.Add(transactionFees, transactionFee)
|
||||
}
|
||||
return transactionFees
|
||||
}
|
||||
|
||||
func calcEthUncleInclusionRewards(header *types.Header, uncles []*types.Header) *big.Int {
|
||||
uncleInclusionRewards := new(big.Int)
|
||||
for range uncles {
|
||||
staticBlockReward := staticRewardByBlockNumber(header.Number.Uint64())
|
||||
staticBlockReward.Div(staticBlockReward, big.NewInt(32))
|
||||
uncleInclusionRewards.Add(uncleInclusionRewards, staticBlockReward)
|
||||
}
|
||||
return uncleInclusionRewards
|
||||
}
|
174
indexer/shared/schema/schema.go
Normal file
174
indexer/shared/schema/schema.go
Normal file
@ -0,0 +1,174 @@
|
||||
// Copyright 2022 The go-ethereum Authors
|
||||
// This file is part of the go-ethereum library.
|
||||
//
|
||||
// The go-ethereum library is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Lesser General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
//
|
||||
// The go-ethereum library is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Lesser General Public License for more details.
|
||||
//
|
||||
// You should have received a copy of the GNU Lesser General Public License
|
||||
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package schema
|
||||
|
||||
var TableIPLDBlock = Table{
|
||||
Name: `ipld.blocks`,
|
||||
Columns: []Column{
|
||||
{Name: "block_number", Type: Dbigint},
|
||||
{Name: "key", Type: Dtext},
|
||||
{Name: "data", Type: Dbytea},
|
||||
},
|
||||
UpsertClause: OnConflict("block_number", "key"),
|
||||
}
|
||||
|
||||
var TableNodeInfo = Table{
|
||||
Name: `public.nodes`,
|
||||
Columns: []Column{
|
||||
{Name: "genesis_block", Type: Dvarchar},
|
||||
{Name: "network_id", Type: Dvarchar},
|
||||
{Name: "node_id", Type: Dvarchar},
|
||||
{Name: "client_name", Type: Dvarchar},
|
||||
{Name: "chain_id", Type: Dinteger},
|
||||
},
|
||||
}
|
||||
|
||||
var TableHeader = Table{
|
||||
Name: "eth.header_cids",
|
||||
Columns: []Column{
|
||||
{Name: "block_number", Type: Dbigint},
|
||||
{Name: "block_hash", Type: Dvarchar},
|
||||
{Name: "parent_hash", Type: Dvarchar},
|
||||
{Name: "cid", Type: Dtext},
|
||||
{Name: "td", Type: Dnumeric},
|
||||
{Name: "node_ids", Type: Dvarchar, Array: true},
|
||||
{Name: "reward", Type: Dnumeric},
|
||||
{Name: "state_root", Type: Dvarchar},
|
||||
{Name: "tx_root", Type: Dvarchar},
|
||||
{Name: "receipt_root", Type: Dvarchar},
|
||||
{Name: "uncles_hash", Type: Dvarchar},
|
||||
{Name: "bloom", Type: Dbytea},
|
||||
{Name: "timestamp", Type: Dnumeric},
|
||||
{Name: "coinbase", Type: Dvarchar},
|
||||
},
|
||||
UpsertClause: OnConflict("block_number", "block_hash").Set(
|
||||
"parent_hash",
|
||||
"cid",
|
||||
"td",
|
||||
"node_ids",
|
||||
"reward",
|
||||
"state_root",
|
||||
"tx_root",
|
||||
"receipt_root",
|
||||
"uncles_hash",
|
||||
"bloom",
|
||||
"timestamp",
|
||||
"coinbase",
|
||||
)}
|
||||
|
||||
var TableStateNode = Table{
|
||||
Name: "eth.state_cids",
|
||||
Columns: []Column{
|
||||
{Name: "block_number", Type: Dbigint},
|
||||
{Name: "header_id", Type: Dvarchar},
|
||||
{Name: "state_leaf_key", Type: Dvarchar},
|
||||
{Name: "cid", Type: Dtext},
|
||||
{Name: "diff", Type: Dboolean},
|
||||
{Name: "balance", Type: Dnumeric},
|
||||
{Name: "nonce", Type: Dbigint},
|
||||
{Name: "code_hash", Type: Dvarchar},
|
||||
{Name: "storage_root", Type: Dvarchar},
|
||||
{Name: "removed", Type: Dboolean},
|
||||
},
|
||||
UpsertClause: OnConflict("block_number", "header_id", "state_leaf_key"),
|
||||
}
|
||||
|
||||
var TableStorageNode = Table{
|
||||
Name: "eth.storage_cids",
|
||||
Columns: []Column{
|
||||
{Name: "block_number", Type: Dbigint},
|
||||
{Name: "header_id", Type: Dvarchar},
|
||||
{Name: "state_leaf_key", Type: Dvarchar},
|
||||
{Name: "storage_leaf_key", Type: Dvarchar},
|
||||
{Name: "cid", Type: Dtext},
|
||||
{Name: "diff", Type: Dboolean},
|
||||
{Name: "val", Type: Dbytea},
|
||||
{Name: "removed", Type: Dboolean},
|
||||
},
|
||||
UpsertClause: OnConflict("block_number", "header_id", "state_leaf_key", "storage_leaf_key"),
|
||||
}
|
||||
|
||||
var TableUncle = Table{
|
||||
Name: "eth.uncle_cids",
|
||||
Columns: []Column{
|
||||
{Name: "block_number", Type: Dbigint},
|
||||
{Name: "block_hash", Type: Dvarchar},
|
||||
{Name: "header_id", Type: Dvarchar},
|
||||
{Name: "parent_hash", Type: Dvarchar},
|
||||
{Name: "cid", Type: Dtext},
|
||||
{Name: "reward", Type: Dnumeric},
|
||||
{Name: "index", Type: Dinteger},
|
||||
},
|
||||
UpsertClause: OnConflict("block_number", "block_hash"),
|
||||
}
|
||||
|
||||
var TableTransaction = Table{
|
||||
Name: "eth.transaction_cids",
|
||||
Columns: []Column{
|
||||
{Name: "block_number", Type: Dbigint},
|
||||
{Name: "header_id", Type: Dvarchar},
|
||||
{Name: "tx_hash", Type: Dvarchar},
|
||||
{Name: "cid", Type: Dtext},
|
||||
{Name: "dst", Type: Dvarchar},
|
||||
{Name: "src", Type: Dvarchar},
|
||||
{Name: "index", Type: Dinteger},
|
||||
{Name: "tx_type", Type: Dinteger},
|
||||
{Name: "value", Type: Dnumeric},
|
||||
},
|
||||
UpsertClause: OnConflict("block_number", "header_id", "tx_hash"),
|
||||
}
|
||||
|
||||
var TableReceipt = Table{
|
||||
Name: "eth.receipt_cids",
|
||||
Columns: []Column{
|
||||
{Name: "block_number", Type: Dbigint},
|
||||
{Name: "header_id", Type: Dvarchar},
|
||||
{Name: "tx_id", Type: Dvarchar},
|
||||
{Name: "cid", Type: Dtext},
|
||||
{Name: "contract", Type: Dvarchar},
|
||||
{Name: "post_state", Type: Dvarchar},
|
||||
{Name: "post_status", Type: Dinteger},
|
||||
},
|
||||
UpsertClause: OnConflict("block_number", "header_id", "tx_id"),
|
||||
}
|
||||
|
||||
var TableLog = Table{
|
||||
Name: "eth.log_cids",
|
||||
Columns: []Column{
|
||||
{Name: "block_number", Type: Dbigint},
|
||||
{Name: "header_id", Type: Dvarchar},
|
||||
{Name: "cid", Type: Dtext},
|
||||
{Name: "rct_id", Type: Dvarchar},
|
||||
{Name: "address", Type: Dvarchar},
|
||||
{Name: "index", Type: Dinteger},
|
||||
{Name: "topic0", Type: Dvarchar},
|
||||
{Name: "topic1", Type: Dvarchar},
|
||||
{Name: "topic2", Type: Dvarchar},
|
||||
{Name: "topic3", Type: Dvarchar},
|
||||
},
|
||||
UpsertClause: OnConflict("block_number", "header_id", "rct_id", "index"),
|
||||
}
|
||||
|
||||
var TableWatchedAddresses = Table{
|
||||
Name: "eth_meta.watched_addresses",
|
||||
Columns: []Column{
|
||||
{Name: "address", Type: Dvarchar},
|
||||
{Name: "created_at", Type: Dbigint},
|
||||
{Name: "watched_at", Type: Dbigint},
|
||||
{Name: "last_filled_at", Type: Dbigint},
|
||||
},
|
||||
}
|
147
indexer/shared/schema/table.go
Normal file
147
indexer/shared/schema/table.go
Normal file
@ -0,0 +1,147 @@
|
||||
// Copyright 2022 The go-ethereum Authors
|
||||
// This file is part of the go-ethereum library.
|
||||
//
|
||||
// The go-ethereum library is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Lesser General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
//
|
||||
// The go-ethereum library is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Lesser General Public License for more details.
|
||||
//
|
||||
// You should have received a copy of the GNU Lesser General Public License
|
||||
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package schema
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"strings"
|
||||
|
||||
"github.com/thoas/go-funk"
|
||||
)
|
||||
|
||||
type colType int
|
||||
|
||||
const (
|
||||
Dinteger colType = iota
|
||||
Dboolean
|
||||
Dbigint
|
||||
Dnumeric
|
||||
Dbytea
|
||||
Dvarchar
|
||||
Dtext
|
||||
)
|
||||
|
||||
type ConflictClause struct {
|
||||
Target []string
|
||||
Update []string
|
||||
}
|
||||
|
||||
type Column struct {
|
||||
Name string
|
||||
Type colType
|
||||
Array bool
|
||||
}
|
||||
type Table struct {
|
||||
Name string
|
||||
Columns []Column
|
||||
|
||||
UpsertClause ConflictClause
|
||||
}
|
||||
|
||||
type colfmt = func(interface{}) string
|
||||
|
||||
func (tbl *Table) ToCsvRow(args ...interface{}) []string {
|
||||
var row []string
|
||||
for i, col := range tbl.Columns {
|
||||
value := col.Type.formatter()(args[i])
|
||||
|
||||
if col.Array {
|
||||
valueList := funk.Map(args[i], col.Type.formatter()).([]string)
|
||||
value = fmt.Sprintf("{%s}", strings.Join(valueList, ","))
|
||||
}
|
||||
|
||||
row = append(row, value)
|
||||
}
|
||||
return row
|
||||
}
|
||||
|
||||
func (tbl *Table) VarcharColumns() []string {
|
||||
columns := funk.Filter(tbl.Columns, func(col Column) bool {
|
||||
return col.Type == Dvarchar
|
||||
}).([]Column)
|
||||
|
||||
columnNames := funk.Map(columns, func(col Column) string {
|
||||
return col.Name
|
||||
}).([]string)
|
||||
return columnNames
|
||||
}
|
||||
|
||||
func OnConflict(target ...string) ConflictClause {
|
||||
return ConflictClause{Target: target}
|
||||
}
|
||||
func (c ConflictClause) Set(fields ...string) ConflictClause {
|
||||
c.Update = fields
|
||||
return c
|
||||
}
|
||||
|
||||
// ToInsertStatement returns a Postgres-compatible SQL insert statement for the table
|
||||
// using positional placeholders
|
||||
func (tbl *Table) ToInsertStatement(upsert bool) string {
|
||||
var colnames, placeholders []string
|
||||
for i, col := range tbl.Columns {
|
||||
colnames = append(colnames, col.Name)
|
||||
placeholders = append(placeholders, fmt.Sprintf("$%d", i+1))
|
||||
}
|
||||
suffix := fmt.Sprintf("ON CONFLICT (%s)", strings.Join(tbl.UpsertClause.Target, ", "))
|
||||
if upsert && len(tbl.UpsertClause.Update) != 0 {
|
||||
var update_placeholders []string
|
||||
for _, name := range tbl.UpsertClause.Update {
|
||||
i := funk.IndexOf(tbl.Columns, func(col Column) bool { return col.Name == name })
|
||||
update_placeholders = append(update_placeholders, fmt.Sprintf("$%d", i+1))
|
||||
}
|
||||
suffix += fmt.Sprintf(
|
||||
" DO UPDATE SET (%s) = (%s)",
|
||||
strings.Join(tbl.UpsertClause.Update, ", "), strings.Join(update_placeholders, ", "),
|
||||
)
|
||||
} else {
|
||||
suffix += " DO NOTHING"
|
||||
}
|
||||
|
||||
return fmt.Sprintf(
|
||||
"INSERT INTO %s (%s) VALUES (%s) %s",
|
||||
tbl.Name, strings.Join(colnames, ", "), strings.Join(placeholders, ", "), suffix,
|
||||
)
|
||||
}
|
||||
|
||||
func sprintf(f string) colfmt {
|
||||
return func(x interface{}) string { return fmt.Sprintf(f, x) }
|
||||
}
|
||||
|
||||
func (typ colType) formatter() colfmt {
|
||||
switch typ {
|
||||
case Dinteger:
|
||||
return sprintf("%d")
|
||||
case Dboolean:
|
||||
return func(x interface{}) string {
|
||||
if x.(bool) {
|
||||
return "t"
|
||||
}
|
||||
return "f"
|
||||
}
|
||||
case Dbigint:
|
||||
return sprintf("%s")
|
||||
case Dnumeric:
|
||||
return sprintf("%s")
|
||||
case Dbytea:
|
||||
return sprintf(`\x%x`)
|
||||
case Dvarchar:
|
||||
return sprintf("%s")
|
||||
case Dtext:
|
||||
return sprintf("%s")
|
||||
}
|
||||
panic("unreachable")
|
||||
}
|
53
indexer/shared/schema/table_test.go
Normal file
53
indexer/shared/schema/table_test.go
Normal file
@ -0,0 +1,53 @@
|
||||
package schema_test
|
||||
|
||||
import (
|
||||
"testing"
|
||||
|
||||
"github.com/stretchr/testify/require"
|
||||
|
||||
. "github.com/ethereum/go-ethereum/statediff/indexer/shared/schema"
|
||||
)
|
||||
|
||||
var testHeaderTable = Table{
|
||||
Name: "eth.header_cids",
|
||||
Columns: []Column{
|
||||
{Name: "block_number", Type: Dbigint},
|
||||
{Name: "block_hash", Type: Dvarchar},
|
||||
{Name: "parent_hash", Type: Dvarchar},
|
||||
{Name: "cid", Type: Dtext},
|
||||
{Name: "td", Type: Dnumeric},
|
||||
{Name: "node_id", Type: Dvarchar},
|
||||
{Name: "reward", Type: Dnumeric},
|
||||
{Name: "state_root", Type: Dvarchar},
|
||||
{Name: "tx_root", Type: Dvarchar},
|
||||
{Name: "receipt_root", Type: Dvarchar},
|
||||
{Name: "uncle_root", Type: Dvarchar},
|
||||
{Name: "bloom", Type: Dbytea},
|
||||
{Name: "timestamp", Type: Dnumeric},
|
||||
{Name: "mh_key", Type: Dtext},
|
||||
{Name: "times_validated", Type: Dinteger},
|
||||
{Name: "coinbase", Type: Dvarchar},
|
||||
},
|
||||
UpsertClause: OnConflict("block_hash", "block_number").Set(
|
||||
"parent_hash",
|
||||
"cid",
|
||||
"td",
|
||||
"node_id",
|
||||
"reward",
|
||||
"state_root",
|
||||
"tx_root",
|
||||
"receipt_root",
|
||||
"uncle_root",
|
||||
"bloom",
|
||||
"timestamp",
|
||||
"mh_key",
|
||||
"times_validated",
|
||||
"coinbase",
|
||||
)}
|
||||
|
||||
func TestTable(t *testing.T) {
|
||||
headerUpsert := `INSERT INTO eth.header_cids (block_number, block_hash, parent_hash, cid, td, node_id, reward, state_root, tx_root, receipt_root, uncle_root, bloom, timestamp, mh_key, times_validated, coinbase) VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16) ON CONFLICT (block_hash, block_number) DO UPDATE SET (parent_hash, cid, td, node_id, reward, state_root, tx_root, receipt_root, uncle_root, bloom, timestamp, mh_key, times_validated, coinbase) = ($3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16)`
|
||||
headerNoUpsert := `INSERT INTO eth.header_cids (block_number, block_hash, parent_hash, cid, td, node_id, reward, state_root, tx_root, receipt_root, uncle_root, bloom, timestamp, mh_key, times_validated, coinbase) VALUES ($1, $2, $3, $4, $5, $6, $7, $8, $9, $10, $11, $12, $13, $14, $15, $16) ON CONFLICT (block_hash, block_number) DO NOTHING`
|
||||
require.Equal(t, headerNoUpsert, testHeaderTable.ToInsertStatement(false))
|
||||
require.Equal(t, headerUpsert, testHeaderTable.ToInsertStatement(true))
|
||||
}
|
1157
indexer/test/test.go
Normal file
1157
indexer/test/test.go
Normal file
File diff suppressed because it is too large
Load Diff
194
indexer/test/test_init.go
Normal file
194
indexer/test/test_init.go
Normal file
@ -0,0 +1,194 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2022 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package test
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"fmt"
|
||||
"os"
|
||||
"testing"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/ipld"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/mocks"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/models"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/shared"
|
||||
"github.com/ipfs/go-cid"
|
||||
"github.com/multiformats/go-multihash"
|
||||
)
|
||||
|
||||
var (
|
||||
err error
|
||||
ipfsPgGet = `SELECT data FROM ipld.blocks
|
||||
WHERE key = $1 AND block_number = $2`
|
||||
watchedAddressesPgGet = `SELECT *
|
||||
FROM eth_meta.watched_addresses`
|
||||
tx1, tx2, tx3, tx4, tx5, rct1, rct2, rct3, rct4, rct5 []byte
|
||||
nonCanonicalBlockRct1, nonCanonicalBlockRct2 []byte
|
||||
nonCanonicalBlock2Rct1, nonCanonicalBlock2Rct2 []byte
|
||||
mockBlock, mockNonCanonicalBlock, mockNonCanonicalBlock2 *types.Block
|
||||
headerCID, mockNonCanonicalHeaderCID, mockNonCanonicalHeader2CID cid.Cid
|
||||
trx1CID, trx2CID, trx3CID, trx4CID, trx5CID cid.Cid
|
||||
rct1CID, rct2CID, rct3CID, rct4CID, rct5CID cid.Cid
|
||||
nonCanonicalBlockRct1CID, nonCanonicalBlockRct2CID cid.Cid
|
||||
nonCanonicalBlock2Rct1CID, nonCanonicalBlock2Rct2CID cid.Cid
|
||||
state1CID, state2CID, storageCID cid.Cid
|
||||
)
|
||||
|
||||
func init() {
|
||||
if os.Getenv("MODE") != "statediff" {
|
||||
fmt.Println("Skipping statediff test")
|
||||
os.Exit(0)
|
||||
}
|
||||
|
||||
// canonical block at LondonBlock height
|
||||
mockBlock = mocks.MockBlock
|
||||
txs, rcts := mocks.MockBlock.Transactions(), mocks.MockReceipts
|
||||
|
||||
// non-canonical block at LondonBlock height
|
||||
mockNonCanonicalBlock = mocks.MockNonCanonicalBlock
|
||||
nonCanonicalBlockRcts := mocks.MockNonCanonicalBlockReceipts
|
||||
|
||||
// non-canonical block at LondonBlock height + 1
|
||||
mockNonCanonicalBlock2 = mocks.MockNonCanonicalBlock2
|
||||
nonCanonicalBlock2Rcts := mocks.MockNonCanonicalBlock2Receipts
|
||||
|
||||
// encode mock receipts
|
||||
buf := new(bytes.Buffer)
|
||||
txs.EncodeIndex(0, buf)
|
||||
tx1 = make([]byte, buf.Len())
|
||||
copy(tx1, buf.Bytes())
|
||||
buf.Reset()
|
||||
|
||||
txs.EncodeIndex(1, buf)
|
||||
tx2 = make([]byte, buf.Len())
|
||||
copy(tx2, buf.Bytes())
|
||||
buf.Reset()
|
||||
|
||||
txs.EncodeIndex(2, buf)
|
||||
tx3 = make([]byte, buf.Len())
|
||||
copy(tx3, buf.Bytes())
|
||||
buf.Reset()
|
||||
|
||||
txs.EncodeIndex(3, buf)
|
||||
tx4 = make([]byte, buf.Len())
|
||||
copy(tx4, buf.Bytes())
|
||||
buf.Reset()
|
||||
|
||||
txs.EncodeIndex(4, buf)
|
||||
tx5 = make([]byte, buf.Len())
|
||||
copy(tx5, buf.Bytes())
|
||||
buf.Reset()
|
||||
|
||||
rcts.EncodeIndex(0, buf)
|
||||
rct1 = make([]byte, buf.Len())
|
||||
copy(rct1, buf.Bytes())
|
||||
buf.Reset()
|
||||
|
||||
rcts.EncodeIndex(1, buf)
|
||||
rct2 = make([]byte, buf.Len())
|
||||
copy(rct2, buf.Bytes())
|
||||
buf.Reset()
|
||||
|
||||
rcts.EncodeIndex(2, buf)
|
||||
rct3 = make([]byte, buf.Len())
|
||||
copy(rct3, buf.Bytes())
|
||||
buf.Reset()
|
||||
|
||||
rcts.EncodeIndex(3, buf)
|
||||
rct4 = make([]byte, buf.Len())
|
||||
copy(rct4, buf.Bytes())
|
||||
buf.Reset()
|
||||
|
||||
rcts.EncodeIndex(4, buf)
|
||||
rct5 = make([]byte, buf.Len())
|
||||
copy(rct5, buf.Bytes())
|
||||
buf.Reset()
|
||||
|
||||
// encode mock receipts for non-canonical blocks
|
||||
nonCanonicalBlockRcts.EncodeIndex(0, buf)
|
||||
nonCanonicalBlockRct1 = make([]byte, buf.Len())
|
||||
copy(nonCanonicalBlockRct1, buf.Bytes())
|
||||
buf.Reset()
|
||||
|
||||
nonCanonicalBlockRcts.EncodeIndex(1, buf)
|
||||
nonCanonicalBlockRct2 = make([]byte, buf.Len())
|
||||
copy(nonCanonicalBlockRct2, buf.Bytes())
|
||||
buf.Reset()
|
||||
|
||||
nonCanonicalBlock2Rcts.EncodeIndex(0, buf)
|
||||
nonCanonicalBlock2Rct1 = make([]byte, buf.Len())
|
||||
copy(nonCanonicalBlock2Rct1, buf.Bytes())
|
||||
buf.Reset()
|
||||
|
||||
nonCanonicalBlock2Rcts.EncodeIndex(1, buf)
|
||||
nonCanonicalBlock2Rct2 = make([]byte, buf.Len())
|
||||
copy(nonCanonicalBlock2Rct2, buf.Bytes())
|
||||
buf.Reset()
|
||||
|
||||
headerCID, _ = ipld.RawdataToCid(ipld.MEthHeader, mocks.MockHeaderRlp, multihash.KECCAK_256)
|
||||
mockNonCanonicalHeaderCID, _ = ipld.RawdataToCid(ipld.MEthHeader, mocks.MockNonCanonicalHeaderRlp, multihash.KECCAK_256)
|
||||
mockNonCanonicalHeader2CID, _ = ipld.RawdataToCid(ipld.MEthHeader, mocks.MockNonCanonicalHeader2Rlp, multihash.KECCAK_256)
|
||||
trx1CID, _ = ipld.RawdataToCid(ipld.MEthTx, tx1, multihash.KECCAK_256)
|
||||
trx2CID, _ = ipld.RawdataToCid(ipld.MEthTx, tx2, multihash.KECCAK_256)
|
||||
trx3CID, _ = ipld.RawdataToCid(ipld.MEthTx, tx3, multihash.KECCAK_256)
|
||||
trx4CID, _ = ipld.RawdataToCid(ipld.MEthTx, tx4, multihash.KECCAK_256)
|
||||
trx5CID, _ = ipld.RawdataToCid(ipld.MEthTx, tx5, multihash.KECCAK_256)
|
||||
state1CID, _ = ipld.RawdataToCid(ipld.MEthStateTrie, mocks.ContractLeafNode, multihash.KECCAK_256)
|
||||
state2CID, _ = ipld.RawdataToCid(ipld.MEthStateTrie, mocks.AccountLeafNode, multihash.KECCAK_256)
|
||||
storageCID, _ = ipld.RawdataToCid(ipld.MEthStorageTrie, mocks.StorageLeafNode, multihash.KECCAK_256)
|
||||
rct1CID, _ = ipld.RawdataToCid(ipld.MEthTxReceipt, rct1, multihash.KECCAK_256)
|
||||
rct2CID, _ = ipld.RawdataToCid(ipld.MEthTxReceipt, rct2, multihash.KECCAK_256)
|
||||
rct3CID, _ = ipld.RawdataToCid(ipld.MEthTxReceipt, rct3, multihash.KECCAK_256)
|
||||
rct4CID, _ = ipld.RawdataToCid(ipld.MEthTxReceipt, rct4, multihash.KECCAK_256)
|
||||
rct5CID, _ = ipld.RawdataToCid(ipld.MEthTxReceipt, rct5, multihash.KECCAK_256)
|
||||
|
||||
// create raw receipts for non-canonical blocks
|
||||
nonCanonicalBlockRct1CID, _ = ipld.RawdataToCid(ipld.MEthTxReceipt, nonCanonicalBlockRct1, multihash.KECCAK_256)
|
||||
nonCanonicalBlockRct2CID, _ = ipld.RawdataToCid(ipld.MEthTxReceipt, nonCanonicalBlockRct2, multihash.KECCAK_256)
|
||||
|
||||
nonCanonicalBlock2Rct1CID, _ = ipld.RawdataToCid(ipld.MEthTxReceipt, nonCanonicalBlock2Rct1, multihash.KECCAK_256)
|
||||
nonCanonicalBlock2Rct2CID, _ = ipld.RawdataToCid(ipld.MEthTxReceipt, nonCanonicalBlock2Rct2, multihash.KECCAK_256)
|
||||
}
|
||||
|
||||
// createRctModel creates a models.ReceiptModel object from a given ethereum receipt
|
||||
func createRctModel(rct *types.Receipt, cid cid.Cid, blockNumber string) models.ReceiptModel {
|
||||
rctModel := models.ReceiptModel{
|
||||
BlockNumber: blockNumber,
|
||||
HeaderID: rct.BlockHash.String(),
|
||||
TxID: rct.TxHash.String(),
|
||||
CID: cid.String(),
|
||||
}
|
||||
|
||||
contract := shared.HandleZeroAddr(rct.ContractAddress)
|
||||
rctModel.Contract = contract
|
||||
|
||||
if len(rct.PostState) == 0 {
|
||||
rctModel.PostStatus = rct.Status
|
||||
} else {
|
||||
rctModel.PostState = common.BytesToHash(rct.PostState).String()
|
||||
}
|
||||
|
||||
return rctModel
|
||||
}
|
||||
|
||||
func expectTrue(t *testing.T, value bool) {
|
||||
if !value {
|
||||
t.Fatalf("Assertion failed")
|
||||
}
|
||||
}
|
96
indexer/test/test_legacy.go
Normal file
96
indexer/test/test_legacy.go
Normal file
@ -0,0 +1,96 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2022 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package test
|
||||
|
||||
import (
|
||||
"context"
|
||||
"testing"
|
||||
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/params"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/file"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/interfaces"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/ipld"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/mocks"
|
||||
"github.com/ipfs/go-cid"
|
||||
"github.com/multiformats/go-multihash"
|
||||
"github.com/stretchr/testify/require"
|
||||
)
|
||||
|
||||
var (
|
||||
LegacyConfig = params.MainnetChainConfig
|
||||
legacyData = mocks.NewLegacyData(LegacyConfig)
|
||||
mockLegacyBlock *types.Block
|
||||
legacyHeaderCID cid.Cid
|
||||
)
|
||||
|
||||
func SetupLegacyTestData(t *testing.T, ind interfaces.StateDiffIndexer) {
|
||||
mockLegacyBlock = legacyData.MockBlock
|
||||
legacyHeaderCID, _ = ipld.RawdataToCid(ipld.MEthHeader, legacyData.MockHeaderRlp, multihash.KECCAK_256)
|
||||
|
||||
var tx interfaces.Batch
|
||||
tx, err = ind.PushBlock(
|
||||
mockLegacyBlock,
|
||||
legacyData.MockReceipts,
|
||||
legacyData.MockBlock.Difficulty())
|
||||
require.NoError(t, err)
|
||||
|
||||
defer func() {
|
||||
if err := tx.Submit(err); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
}()
|
||||
for _, node := range legacyData.StateDiffs {
|
||||
err = ind.PushStateNode(tx, node, mockLegacyBlock.Hash().String())
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
if batchTx, ok := tx.(*sql.BatchTx); ok {
|
||||
require.Equal(t, legacyData.BlockNumber.String(), batchTx.BlockNumber)
|
||||
} else if batchTx, ok := tx.(*file.BatchTx); ok {
|
||||
require.Equal(t, legacyData.BlockNumber.String(), batchTx.BlockNumber)
|
||||
}
|
||||
}
|
||||
|
||||
func TestLegacyIndexer(t *testing.T, db sql.Database) {
|
||||
pgStr := `SELECT cid, cast(td AS TEXT), cast(reward AS TEXT), block_hash, coinbase
|
||||
FROM eth.header_cids
|
||||
WHERE block_number = $1`
|
||||
// check header was properly indexed
|
||||
type res struct {
|
||||
CID string
|
||||
TD string
|
||||
Reward string
|
||||
BlockHash string `db:"block_hash"`
|
||||
Coinbase string `db:"coinbase"`
|
||||
}
|
||||
header := new(res)
|
||||
err = db.QueryRow(context.Background(), pgStr, legacyData.BlockNumber.Uint64()).Scan(
|
||||
&header.CID,
|
||||
&header.TD,
|
||||
&header.Reward,
|
||||
&header.BlockHash,
|
||||
&header.Coinbase)
|
||||
require.NoError(t, err)
|
||||
|
||||
require.Equal(t, legacyHeaderCID.String(), header.CID)
|
||||
require.Equal(t, legacyData.MockBlock.Difficulty().String(), header.TD)
|
||||
require.Equal(t, "5000000000000011250", header.Reward)
|
||||
require.Equal(t, legacyData.MockHeader.Coinbase.String(), header.Coinbase)
|
||||
require.Nil(t, legacyData.MockHeader.BaseFee)
|
||||
}
|
53
indexer/test/test_mainnet.go
Normal file
53
indexer/test/test_mainnet.go
Normal file
@ -0,0 +1,53 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2022 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package test
|
||||
|
||||
import (
|
||||
"testing"
|
||||
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/file"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/interfaces"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/mocks"
|
||||
"github.com/stretchr/testify/require"
|
||||
)
|
||||
|
||||
func TestBlock(t *testing.T, ind interfaces.StateDiffIndexer, testBlock *types.Block, testReceipts types.Receipts) {
|
||||
var tx interfaces.Batch
|
||||
tx, err = ind.PushBlock(
|
||||
testBlock,
|
||||
testReceipts,
|
||||
testBlock.Difficulty())
|
||||
require.NoError(t, err)
|
||||
|
||||
defer func() {
|
||||
if err := tx.Submit(err); err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
}()
|
||||
for _, node := range mocks.StateDiffs {
|
||||
err = ind.PushStateNode(tx, node, testBlock.Hash().String())
|
||||
require.NoError(t, err)
|
||||
}
|
||||
|
||||
if batchTx, ok := tx.(*sql.BatchTx); ok {
|
||||
require.Equal(t, testBlock.Number().String(), batchTx.BlockNumber)
|
||||
} else if batchTx, ok := tx.(*file.BatchTx); ok {
|
||||
require.Equal(t, testBlock.Number().String(), batchTx.BlockNumber)
|
||||
}
|
||||
}
|
258
indexer/test/test_watched_addresses.go
Normal file
258
indexer/test/test_watched_addresses.go
Normal file
@ -0,0 +1,258 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2022 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package test
|
||||
|
||||
import (
|
||||
"context"
|
||||
"testing"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/interfaces"
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/mocks"
|
||||
"github.com/stretchr/testify/require"
|
||||
)
|
||||
|
||||
type res struct {
|
||||
Address string `db:"address"`
|
||||
CreatedAt uint64 `db:"created_at"`
|
||||
WatchedAt uint64 `db:"watched_at"`
|
||||
LastFilledAt uint64 `db:"last_filled_at"`
|
||||
}
|
||||
|
||||
func TestLoadEmptyWatchedAddresses(t *testing.T, ind interfaces.StateDiffIndexer) {
|
||||
expectedData := []common.Address{}
|
||||
|
||||
rows, err := ind.LoadWatchedAddresses()
|
||||
require.NoError(t, err)
|
||||
|
||||
require.Equal(t, len(expectedData), len(rows))
|
||||
for idx, row := range rows {
|
||||
require.Equal(t, expectedData[idx], row)
|
||||
}
|
||||
}
|
||||
|
||||
func TestInsertWatchedAddresses(t *testing.T, db sql.Database) {
|
||||
expectedData := []res{
|
||||
{
|
||||
Address: mocks.Contract1Address,
|
||||
CreatedAt: mocks.Contract1CreatedAt,
|
||||
WatchedAt: mocks.WatchedAt1,
|
||||
LastFilledAt: mocks.LastFilledAt,
|
||||
},
|
||||
{
|
||||
Address: mocks.Contract2Address,
|
||||
CreatedAt: mocks.Contract2CreatedAt,
|
||||
WatchedAt: mocks.WatchedAt1,
|
||||
LastFilledAt: mocks.LastFilledAt,
|
||||
},
|
||||
}
|
||||
|
||||
rows := []res{}
|
||||
err = db.Select(context.Background(), &rows, watchedAddressesPgGet)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
require.Equal(t, len(expectedData), len(rows))
|
||||
for idx, row := range rows {
|
||||
require.Equal(t, expectedData[idx], row)
|
||||
}
|
||||
}
|
||||
|
||||
func TestInsertAlreadyWatchedAddresses(t *testing.T, db sql.Database) {
|
||||
expectedData := []res{
|
||||
{
|
||||
Address: mocks.Contract1Address,
|
||||
CreatedAt: mocks.Contract1CreatedAt,
|
||||
WatchedAt: mocks.WatchedAt1,
|
||||
LastFilledAt: mocks.LastFilledAt,
|
||||
},
|
||||
{
|
||||
Address: mocks.Contract2Address,
|
||||
CreatedAt: mocks.Contract2CreatedAt,
|
||||
WatchedAt: mocks.WatchedAt1,
|
||||
LastFilledAt: mocks.LastFilledAt,
|
||||
},
|
||||
{
|
||||
Address: mocks.Contract3Address,
|
||||
CreatedAt: mocks.Contract3CreatedAt,
|
||||
WatchedAt: mocks.WatchedAt2,
|
||||
LastFilledAt: mocks.LastFilledAt,
|
||||
},
|
||||
}
|
||||
|
||||
rows := []res{}
|
||||
err = db.Select(context.Background(), &rows, watchedAddressesPgGet)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
require.Equal(t, len(expectedData), len(rows))
|
||||
for idx, row := range rows {
|
||||
require.Equal(t, expectedData[idx], row)
|
||||
}
|
||||
}
|
||||
|
||||
func TestRemoveWatchedAddresses(t *testing.T, db sql.Database) {
|
||||
expectedData := []res{
|
||||
{
|
||||
Address: mocks.Contract1Address,
|
||||
CreatedAt: mocks.Contract1CreatedAt,
|
||||
WatchedAt: mocks.WatchedAt1,
|
||||
LastFilledAt: mocks.LastFilledAt,
|
||||
},
|
||||
}
|
||||
|
||||
rows := []res{}
|
||||
err = db.Select(context.Background(), &rows, watchedAddressesPgGet)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
require.Equal(t, len(expectedData), len(rows))
|
||||
for idx, row := range rows {
|
||||
require.Equal(t, expectedData[idx], row)
|
||||
}
|
||||
}
|
||||
|
||||
func TestRemoveNonWatchedAddresses(t *testing.T, db sql.Database) {
|
||||
expectedData := []res{}
|
||||
|
||||
rows := []res{}
|
||||
err = db.Select(context.Background(), &rows, watchedAddressesPgGet)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
require.Equal(t, len(expectedData), len(rows))
|
||||
for idx, row := range rows {
|
||||
require.Equal(t, expectedData[idx], row)
|
||||
}
|
||||
}
|
||||
|
||||
func TestSetWatchedAddresses(t *testing.T, db sql.Database) {
|
||||
expectedData := []res{
|
||||
{
|
||||
Address: mocks.Contract1Address,
|
||||
CreatedAt: mocks.Contract1CreatedAt,
|
||||
WatchedAt: mocks.WatchedAt2,
|
||||
LastFilledAt: mocks.LastFilledAt,
|
||||
},
|
||||
{
|
||||
Address: mocks.Contract2Address,
|
||||
CreatedAt: mocks.Contract2CreatedAt,
|
||||
WatchedAt: mocks.WatchedAt2,
|
||||
LastFilledAt: mocks.LastFilledAt,
|
||||
},
|
||||
{
|
||||
Address: mocks.Contract3Address,
|
||||
CreatedAt: mocks.Contract3CreatedAt,
|
||||
WatchedAt: mocks.WatchedAt2,
|
||||
LastFilledAt: mocks.LastFilledAt,
|
||||
},
|
||||
}
|
||||
|
||||
rows := []res{}
|
||||
err = db.Select(context.Background(), &rows, watchedAddressesPgGet)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
require.Equal(t, len(expectedData), len(rows))
|
||||
for idx, row := range rows {
|
||||
require.Equal(t, expectedData[idx], row)
|
||||
}
|
||||
}
|
||||
|
||||
func TestSetAlreadyWatchedAddresses(t *testing.T, db sql.Database) {
|
||||
expectedData := []res{
|
||||
{
|
||||
Address: mocks.Contract4Address,
|
||||
CreatedAt: mocks.Contract4CreatedAt,
|
||||
WatchedAt: mocks.WatchedAt3,
|
||||
LastFilledAt: mocks.LastFilledAt,
|
||||
},
|
||||
{
|
||||
Address: mocks.Contract2Address,
|
||||
CreatedAt: mocks.Contract2CreatedAt,
|
||||
WatchedAt: mocks.WatchedAt3,
|
||||
LastFilledAt: mocks.LastFilledAt,
|
||||
},
|
||||
{
|
||||
Address: mocks.Contract3Address,
|
||||
CreatedAt: mocks.Contract3CreatedAt,
|
||||
WatchedAt: mocks.WatchedAt3,
|
||||
LastFilledAt: mocks.LastFilledAt,
|
||||
},
|
||||
}
|
||||
|
||||
rows := []res{}
|
||||
err = db.Select(context.Background(), &rows, watchedAddressesPgGet)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
require.Equal(t, len(expectedData), len(rows))
|
||||
for idx, row := range rows {
|
||||
require.Equal(t, expectedData[idx], row)
|
||||
}
|
||||
}
|
||||
|
||||
func TestLoadWatchedAddresses(t *testing.T, ind interfaces.StateDiffIndexer) {
|
||||
expectedData := []common.Address{
|
||||
common.HexToAddress(mocks.Contract4Address),
|
||||
common.HexToAddress(mocks.Contract2Address),
|
||||
common.HexToAddress(mocks.Contract3Address),
|
||||
}
|
||||
|
||||
rows, err := ind.LoadWatchedAddresses()
|
||||
require.NoError(t, err)
|
||||
|
||||
require.Equal(t, len(expectedData), len(rows))
|
||||
for idx, row := range rows {
|
||||
require.Equal(t, expectedData[idx], row)
|
||||
}
|
||||
}
|
||||
|
||||
func TestClearWatchedAddresses(t *testing.T, db sql.Database) {
|
||||
expectedData := []res{}
|
||||
rows := []res{}
|
||||
err = db.Select(context.Background(), &rows, watchedAddressesPgGet)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
require.Equal(t, len(expectedData), len(rows))
|
||||
for idx, row := range rows {
|
||||
require.Equal(t, expectedData[idx], row)
|
||||
}
|
||||
}
|
||||
|
||||
func TestClearEmptyWatchedAddresses(t *testing.T, db sql.Database) {
|
||||
expectedData := []res{}
|
||||
rows := []res{}
|
||||
err = db.Select(context.Background(), &rows, watchedAddressesPgGet)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
require.Equal(t, len(expectedData), len(rows))
|
||||
for idx, row := range rows {
|
||||
require.Equal(t, expectedData[idx], row)
|
||||
}
|
||||
}
|
248
indexer/test_helpers/mainnet_test_helpers.go
Normal file
248
indexer/test_helpers/mainnet_test_helpers.go
Normal file
@ -0,0 +1,248 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package test_helpers
|
||||
|
||||
import (
|
||||
"context"
|
||||
"errors"
|
||||
"fmt"
|
||||
"math/big"
|
||||
"os"
|
||||
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/ethclient"
|
||||
"github.com/ethereum/go-ethereum/rlp"
|
||||
)
|
||||
|
||||
const (
|
||||
defaultBlockFilePath = "../../../mainnet_data/block"
|
||||
defaultReceiptsFilePath = "../../../mainnet_data/receipts"
|
||||
)
|
||||
|
||||
const (
|
||||
TEST_RAW_URL = "TEST_RAW_URL"
|
||||
TEST_BLOCK_NUMBER = "TEST_BLOCK_NUMBER"
|
||||
)
|
||||
|
||||
// ProblemBlocks list of known problem blocks, with funky edge cases
|
||||
var ProblemBlocks = []int64{
|
||||
12600011,
|
||||
12619985,
|
||||
12625121,
|
||||
12655432,
|
||||
12579670,
|
||||
12914664,
|
||||
}
|
||||
|
||||
// TestConfig holds configuration params for mainnet tests
|
||||
type TestConfig struct {
|
||||
RawURL string
|
||||
BlockNumber *big.Int
|
||||
LocalCache bool
|
||||
}
|
||||
|
||||
// DefaultTestConfig is the default TestConfig
|
||||
var DefaultTestConfig = TestConfig{
|
||||
RawURL: "http://127.0.0.1:8545",
|
||||
BlockNumber: big.NewInt(12914664),
|
||||
LocalCache: true,
|
||||
}
|
||||
|
||||
func GetTestConfig() TestConfig {
|
||||
conf := DefaultTestConfig
|
||||
rawURL := os.Getenv(TEST_RAW_URL)
|
||||
if rawURL == "" {
|
||||
fmt.Printf("Warning: no raw url configured for statediffing mainnet tests, will look for local file and"+
|
||||
"then try default endpoint (%s)\r\n", DefaultTestConfig.RawURL)
|
||||
} else {
|
||||
conf.RawURL = rawURL
|
||||
}
|
||||
return conf
|
||||
}
|
||||
|
||||
// TestBlockAndReceiptsFromEnv retrieves the block and receipts using env variables to override default config block number
|
||||
func TestBlockAndReceiptsFromEnv(conf TestConfig) (*types.Block, types.Receipts, error) {
|
||||
blockNumberStr := os.Getenv(TEST_BLOCK_NUMBER)
|
||||
blockNumber, ok := new(big.Int).SetString(blockNumberStr, 10)
|
||||
if !ok {
|
||||
fmt.Printf("Warning: no blockNumber configured for statediffing mainnet tests, using default (%d)\r\n",
|
||||
DefaultTestConfig.BlockNumber)
|
||||
} else {
|
||||
conf.BlockNumber = blockNumber
|
||||
}
|
||||
return TestBlockAndReceipts(conf)
|
||||
}
|
||||
|
||||
// TestBlockAndReceipts retrieves the block and receipts for the provided test config
|
||||
// It first tries to load files from the local system before setting up and using an ethclient.Client to pull the data
|
||||
func TestBlockAndReceipts(conf TestConfig) (*types.Block, types.Receipts, error) {
|
||||
var cli *ethclient.Client
|
||||
var err error
|
||||
var block *types.Block
|
||||
var receipts types.Receipts
|
||||
blockFilePath := fmt.Sprintf("%s_%s.rlp", defaultBlockFilePath, conf.BlockNumber.String())
|
||||
if _, err = os.Stat(blockFilePath); !errors.Is(err, os.ErrNotExist) {
|
||||
fmt.Printf("local file (%s) found for block %s\n", blockFilePath, conf.BlockNumber.String())
|
||||
block, err = LoadBlockRLP(blockFilePath)
|
||||
if err != nil {
|
||||
fmt.Printf("loading local file (%s) failed (%s), dialing remote client at %s\n", blockFilePath, err.Error(), conf.RawURL)
|
||||
cli, err = ethclient.Dial(conf.RawURL)
|
||||
if err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
block, err = FetchBlock(cli, conf.BlockNumber)
|
||||
if err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
if conf.LocalCache {
|
||||
if err := WriteBlockRLP(blockFilePath, block); err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
}
|
||||
}
|
||||
} else {
|
||||
fmt.Printf("no local file found for block %s, dialing remote client at %s\n", conf.BlockNumber.String(), conf.RawURL)
|
||||
cli, err = ethclient.Dial(conf.RawURL)
|
||||
if err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
block, err = FetchBlock(cli, conf.BlockNumber)
|
||||
if err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
if conf.LocalCache {
|
||||
if err := WriteBlockRLP(blockFilePath, block); err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
}
|
||||
}
|
||||
receiptsFilePath := fmt.Sprintf("%s_%s.rlp", defaultReceiptsFilePath, conf.BlockNumber.String())
|
||||
if _, err = os.Stat(receiptsFilePath); !errors.Is(err, os.ErrNotExist) {
|
||||
fmt.Printf("local file (%s) found for block %s receipts\n", receiptsFilePath, conf.BlockNumber.String())
|
||||
receipts, err = LoadReceiptsEncoding(receiptsFilePath, len(block.Transactions()))
|
||||
if err != nil {
|
||||
fmt.Printf("loading local file (%s) failed (%s), dialing remote client at %s\n", receiptsFilePath, err.Error(), conf.RawURL)
|
||||
if cli == nil {
|
||||
cli, err = ethclient.Dial(conf.RawURL)
|
||||
if err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
}
|
||||
receipts, err = FetchReceipts(cli, block)
|
||||
if err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
if conf.LocalCache {
|
||||
if err := WriteReceiptsEncoding(receiptsFilePath, block.Number(), receipts); err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
}
|
||||
}
|
||||
} else {
|
||||
fmt.Printf("no local file found for block %s receipts, dialing remote client at %s\n", conf.BlockNumber.String(), conf.RawURL)
|
||||
if cli == nil {
|
||||
cli, err = ethclient.Dial(conf.RawURL)
|
||||
if err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
}
|
||||
receipts, err = FetchReceipts(cli, block)
|
||||
if err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
if conf.LocalCache {
|
||||
if err := WriteReceiptsEncoding(receiptsFilePath, block.Number(), receipts); err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
}
|
||||
}
|
||||
return block, receipts, nil
|
||||
}
|
||||
|
||||
// FetchBlock fetches the block at the provided height using the ethclient.Client
|
||||
func FetchBlock(cli *ethclient.Client, blockNumber *big.Int) (*types.Block, error) {
|
||||
return cli.BlockByNumber(context.Background(), blockNumber)
|
||||
}
|
||||
|
||||
// FetchReceipts fetches the receipts for the provided block using the ethclient.Client
|
||||
func FetchReceipts(cli *ethclient.Client, block *types.Block) (types.Receipts, error) {
|
||||
receipts := make(types.Receipts, len(block.Transactions()))
|
||||
for i, tx := range block.Transactions() {
|
||||
rct, err := cli.TransactionReceipt(context.Background(), tx.Hash())
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
receipts[i] = rct
|
||||
}
|
||||
return receipts, nil
|
||||
}
|
||||
|
||||
// WriteBlockRLP writes out the RLP encoding of the block to the provided filePath
|
||||
func WriteBlockRLP(filePath string, block *types.Block) error {
|
||||
if filePath == "" {
|
||||
filePath = fmt.Sprintf("%s_%s.rlp", defaultBlockFilePath, block.Number().String())
|
||||
}
|
||||
if _, err := os.Stat(filePath); !errors.Is(err, os.ErrNotExist) {
|
||||
return fmt.Errorf("cannot create file, file (%s) already exists", filePath)
|
||||
}
|
||||
file, err := os.Create(filePath)
|
||||
if err != nil {
|
||||
return fmt.Errorf("unable to create file (%s), err: %v", filePath, err)
|
||||
}
|
||||
fmt.Printf("writing block rlp to file at %s\r\n", filePath)
|
||||
if err := block.EncodeRLP(file); err != nil {
|
||||
return err
|
||||
}
|
||||
return file.Close()
|
||||
}
|
||||
|
||||
// LoadBlockRLP loads block from the rlp at filePath
|
||||
func LoadBlockRLP(filePath string) (*types.Block, error) {
|
||||
blockBytes, err := os.ReadFile(filePath)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
block := new(types.Block)
|
||||
return block, rlp.DecodeBytes(blockBytes, block)
|
||||
}
|
||||
|
||||
// LoadReceiptsEncoding loads receipts from the encoding at filePath
|
||||
func LoadReceiptsEncoding(filePath string, cap int) (types.Receipts, error) {
|
||||
rctsBytes, err := os.ReadFile(filePath)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
receipts := new(types.Receipts)
|
||||
return *receipts, rlp.DecodeBytes(rctsBytes, receipts)
|
||||
}
|
||||
|
||||
// WriteReceiptsEncoding writes out the consensus encoding of the receipts to the provided io.WriteCloser
|
||||
func WriteReceiptsEncoding(filePath string, blockNumber *big.Int, receipts types.Receipts) error {
|
||||
if filePath == "" {
|
||||
filePath = fmt.Sprintf("%s_%s.rlp", defaultReceiptsFilePath, blockNumber.String())
|
||||
}
|
||||
if _, err := os.Stat(filePath); !errors.Is(err, os.ErrNotExist) {
|
||||
return fmt.Errorf("cannot create file, file (%s) already exists", filePath)
|
||||
}
|
||||
file, err := os.Create(filePath)
|
||||
if err != nil {
|
||||
return fmt.Errorf("unable to create file (%s), err: %v", filePath, err)
|
||||
}
|
||||
defer file.Close()
|
||||
fmt.Printf("writing receipts rlp to file at %s\r\n", filePath)
|
||||
return rlp.Encode(file, receipts)
|
||||
}
|
123
indexer/test_helpers/test_helpers.go
Normal file
123
indexer/test_helpers/test_helpers.go
Normal file
@ -0,0 +1,123 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2019 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package test_helpers
|
||||
|
||||
import (
|
||||
"bufio"
|
||||
"context"
|
||||
"os"
|
||||
"testing"
|
||||
|
||||
"github.com/ethereum/go-ethereum/statediff/indexer/database/sql"
|
||||
)
|
||||
|
||||
// ListContainsString used to check if a list of strings contains a particular string
|
||||
func ListContainsString(sss []string, s string) bool {
|
||||
for _, str := range sss {
|
||||
if s == str {
|
||||
return true
|
||||
}
|
||||
}
|
||||
return false
|
||||
}
|
||||
|
||||
// DedupFile removes duplicates from the given file
|
||||
func DedupFile(filePath string) error {
|
||||
f, err := os.OpenFile(filePath, os.O_CREATE|os.O_RDONLY, os.ModePerm)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
stmts := make(map[string]struct{}, 0)
|
||||
sc := bufio.NewScanner(f)
|
||||
|
||||
for sc.Scan() {
|
||||
s := sc.Text()
|
||||
stmts[s] = struct{}{}
|
||||
}
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
f.Close()
|
||||
|
||||
f, err = os.Create(filePath)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
defer f.Close()
|
||||
|
||||
for stmt := range stmts {
|
||||
f.Write([]byte(stmt + "\n"))
|
||||
}
|
||||
|
||||
return nil
|
||||
}
|
||||
|
||||
// TearDownDB is used to tear down the watcher dbs after tests
|
||||
func TearDownDB(t *testing.T, db sql.Database) {
|
||||
ctx := context.Background()
|
||||
tx, err := db.Begin(ctx)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
|
||||
_, err = tx.Exec(ctx, `DELETE FROM eth.header_cids`)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
_, err = tx.Exec(ctx, `DELETE FROM eth.uncle_cids`)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
_, err = tx.Exec(ctx, `DELETE FROM eth.transaction_cids`)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
_, err = tx.Exec(ctx, `DELETE FROM eth.receipt_cids`)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
_, err = tx.Exec(ctx, `DELETE FROM eth.state_cids`)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
_, err = tx.Exec(ctx, `DELETE FROM eth.storage_cids`)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
_, err = tx.Exec(ctx, `DELETE FROM eth.log_cids`)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
_, err = tx.Exec(ctx, `DELETE FROM ipld.blocks`)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
_, err = tx.Exec(ctx, `DELETE FROM nodes`)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
_, err = tx.Exec(ctx, `DELETE FROM eth_meta.watched_addresses`)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
err = tx.Commit(ctx)
|
||||
if err != nil {
|
||||
t.Fatal(err)
|
||||
}
|
||||
}
|
BIN
mainnet_tests/block0_rlp
Normal file
BIN
mainnet_tests/block0_rlp
Normal file
Binary file not shown.
BIN
mainnet_tests/block1_rlp
Normal file
BIN
mainnet_tests/block1_rlp
Normal file
Binary file not shown.
BIN
mainnet_tests/block2_rlp
Normal file
BIN
mainnet_tests/block2_rlp
Normal file
Binary file not shown.
BIN
mainnet_tests/block3_rlp
Normal file
BIN
mainnet_tests/block3_rlp
Normal file
Binary file not shown.
704
mainnet_tests/builder_test.go
Normal file
704
mainnet_tests/builder_test.go
Normal file
@ -0,0 +1,704 @@
|
||||
// Copyright 2019 The go-ethereum Authors
|
||||
// This file is part of the go-ethereum library.
|
||||
//
|
||||
// The go-ethereum library is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Lesser General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
//
|
||||
// The go-ethereum library is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Lesser General Public License for more details.
|
||||
//
|
||||
// You should have received a copy of the GNU Lesser General Public License
|
||||
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package statediff_test
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
"io"
|
||||
"log"
|
||||
"math/big"
|
||||
"os"
|
||||
"sort"
|
||||
"testing"
|
||||
|
||||
"github.com/ethereum/go-ethereum/common"
|
||||
"github.com/ethereum/go-ethereum/consensus/ethash"
|
||||
"github.com/ethereum/go-ethereum/core"
|
||||
"github.com/ethereum/go-ethereum/core/rawdb"
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/core/vm"
|
||||
"github.com/ethereum/go-ethereum/crypto"
|
||||
"github.com/ethereum/go-ethereum/ethdb"
|
||||
"github.com/ethereum/go-ethereum/rlp"
|
||||
"github.com/ethereum/go-ethereum/statediff"
|
||||
ipld2 "github.com/ethereum/go-ethereum/statediff/indexer/ipld"
|
||||
"github.com/ethereum/go-ethereum/statediff/test_helpers"
|
||||
sdtypes "github.com/ethereum/go-ethereum/statediff/types"
|
||||
)
|
||||
|
||||
var (
|
||||
db ethdb.Database
|
||||
genesisBlock, block0, block1, block2, block3 *types.Block
|
||||
block1CoinbaseAddr, block2CoinbaseAddr, block3CoinbaseAddr common.Address
|
||||
block1CoinbaseHash, block2CoinbaseHash, block3CoinbaseHash common.Hash
|
||||
builder statediff.Builder
|
||||
emptyStorage = make([]sdtypes.StorageLeafNode, 0)
|
||||
|
||||
// block 1 data
|
||||
block1CoinbaseAccount = &types.StateAccount{
|
||||
Nonce: 0,
|
||||
Balance: big.NewInt(5000000000000000000),
|
||||
CodeHash: test_helpers.NullCodeHash.Bytes(),
|
||||
Root: test_helpers.EmptyContractRoot,
|
||||
}
|
||||
block1CoinbaseAccountRLP, _ = rlp.EncodeToBytes(block1CoinbaseAccount)
|
||||
block1CoinbaseLeafNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("38251692195afc818c92b485fcb8a4691af89cbe5a2ab557b83a4261be2a9a"),
|
||||
block1CoinbaseAccountRLP,
|
||||
})
|
||||
block1CoinbaseLeafNodeHash = crypto.Keccak256(block1CoinbaseLeafNode)
|
||||
block1x040bBranchNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("cc947d5ebb80600bad471f12c6ad5e4981e3525ecf8a2d982cc032536ae8b66d"),
|
||||
common.Hex2Bytes("e80e52462e635a834e90e86ccf7673a6430384aac17004d626f4db831f0624bc"),
|
||||
common.Hex2Bytes("59a8f11f60cb0a8488831f242da02944a26fd269d0608a44b8b873ded9e59e1b"),
|
||||
common.Hex2Bytes("1ffb51e987e3cbd2e1dc1a64508d2e2b265477e21698b0d10fdf137f35027f40"),
|
||||
[]byte{},
|
||||
common.Hex2Bytes("ce5077f49a13ff8199d0e77715fdd7bfd6364774effcd5499bd93cba54b3c644"),
|
||||
common.Hex2Bytes("f5146783c048e66ce1a776ae990b4255e5fba458ece77fcb83ff6e91d6637a88"),
|
||||
common.Hex2Bytes("6a0558b6c38852e985cf01c2156517c1c6a1e64c787a953c347825f050b236c6"),
|
||||
common.Hex2Bytes("56b6e93958b99aaae158cc2329e71a1865ba6f39c67b096922c5cf3ed86b0ae5"),
|
||||
[]byte{},
|
||||
common.Hex2Bytes("50d317a89a3405367d66668902f2c9f273a8d0d7d5d790dc516bca142f4a84af"),
|
||||
common.Hex2Bytes("c72ca72750fdc1af3e6da5c7c5d82c54e4582f15b488a8aa1674058a99825dae"),
|
||||
common.Hex2Bytes("e1a489df7b18cde818da6d38e235b026c2e61bcd3d34880b3ed0d67e0e4f0159"),
|
||||
common.Hex2Bytes("b58d5062f2609fd2d68f00d14ab33fef2b373853877cf40bf64729e85b8fdc54"),
|
||||
block1CoinbaseLeafNodeHash,
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
})
|
||||
block1x040bBranchNodeHash = crypto.Keccak256(block1x040bBranchNode)
|
||||
block1x04BranchNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("a9317a59365ca09cefcd384018696590afffc432e35a97e8f85aa48907bf3247"),
|
||||
common.Hex2Bytes("e0bc229254ce7a6a736c3953e570ab18b4a7f5f2a9aa3c3057b5f17d250a1cad"),
|
||||
common.Hex2Bytes("a2484ec8884dbe0cf24ece99d67df0d1fe78992d67cc777636a817cb2ef205aa"),
|
||||
common.Hex2Bytes("12b78d4078c607747f06bb88bd08f839eaae0e3ac6854e5f65867d4f78abb84e"),
|
||||
common.Hex2Bytes("359a51862df5462e4cd302f69cb338512f21eb37ce0791b9a562e72ec48b7dbf"),
|
||||
common.Hex2Bytes("13f8d617b6a734da9235b6ac80bdd7aeaff6120c39aa223638d88f22d4ba4007"),
|
||||
common.Hex2Bytes("02055c6400e0ec3440a8bb8fdfd7d6b6c57b7bf83e37d7e4e983d416fdd8314e"),
|
||||
common.Hex2Bytes("4b1cca9eb3e47e805e7f4c80671a9fcd589fd6ddbe1790c3f3e177e8ede01b9e"),
|
||||
common.Hex2Bytes("70c3815efb23b986018089e009a38e6238b8850b3efd33831913ca6fa9240249"),
|
||||
common.Hex2Bytes("7084699d2e72a193fd75bb6108ae797b4661696eba2d631d521fc94acc7b3247"),
|
||||
common.Hex2Bytes("b2b3cd9f1e46eb583a6185d9a96b4e80125e3d75e6191fdcf684892ef52935cb"),
|
||||
block1x040bBranchNodeHash,
|
||||
common.Hex2Bytes("34d9ff0fee6c929424e52268dedbc596d10786e909c5a68d6466c2aba17387ce"),
|
||||
common.Hex2Bytes("7484d5e44b6ee6b10000708c37e035b42b818475620f9316beffc46531d1eebf"),
|
||||
common.Hex2Bytes("30c8a283adccf2742272563cd3d6710c89ba21eac0118bf5310cfb231bcca77f"),
|
||||
common.Hex2Bytes("4bae8558d2385b8d3bc6e6ede20bdbc5dbb0b5384c316ba8985682f88d2e506d"),
|
||||
[]byte{},
|
||||
})
|
||||
block1x04BranchNodeHash = crypto.Keccak256(block1x04BranchNode)
|
||||
block1RootBranchNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("90dcaf88c40c7bbc95a912cbdde67c175767b31173df9ee4b0d733bfdd511c43"),
|
||||
common.Hex2Bytes("babe369f6b12092f49181ae04ca173fb68d1a5456f18d20fa32cba73954052bd"),
|
||||
common.Hex2Bytes("473ecf8a7e36a829e75039a3b055e51b8332cbf03324ab4af2066bbd6fbf0021"),
|
||||
common.Hex2Bytes("bbda34753d7aa6c38e603f360244e8f59611921d9e1f128372fec0d586d4f9e0"),
|
||||
block1x04BranchNodeHash,
|
||||
common.Hex2Bytes("a5f3f2f7542148c973977c8a1e154c4300fec92f755f7846f1b734d3ab1d90e7"),
|
||||
common.Hex2Bytes("e823850f50bf72baae9d1733a36a444ab65d0a6faaba404f0583ce0ca4dad92d"),
|
||||
common.Hex2Bytes("f7a00cbe7d4b30b11faea3ae61b7f1f2b315b61d9f6bd68bfe587ad0eeceb721"),
|
||||
common.Hex2Bytes("7117ef9fc932f1a88e908eaead8565c19b5645dc9e5b1b6e841c5edbdfd71681"),
|
||||
common.Hex2Bytes("69eb2de283f32c11f859d7bcf93da23990d3e662935ed4d6b39ce3673ec84472"),
|
||||
common.Hex2Bytes("203d26456312bbc4da5cd293b75b840fc5045e493d6f904d180823ec22bfed8e"),
|
||||
common.Hex2Bytes("9287b5c21f2254af4e64fca76acc5cd87399c7f1ede818db4326c98ce2dc2208"),
|
||||
common.Hex2Bytes("6fc2d754e304c48ce6a517753c62b1a9c1d5925b89707486d7fc08919e0a94ec"),
|
||||
common.Hex2Bytes("7b1c54f15e299bd58bdfef9741538c7828b5d7d11a489f9c20d052b3471df475"),
|
||||
common.Hex2Bytes("51f9dd3739a927c89e357580a4c97b40234aa01ed3d5e0390dc982a7975880a0"),
|
||||
common.Hex2Bytes("89d613f26159af43616fd9455bb461f4869bfede26f2130835ed067a8b967bfb"),
|
||||
[]byte{},
|
||||
})
|
||||
|
||||
// block 2 data
|
||||
block2CoinbaseAccount = &types.StateAccount{
|
||||
Nonce: 0,
|
||||
Balance: big.NewInt(5000000000000000000),
|
||||
CodeHash: test_helpers.NullCodeHash.Bytes(),
|
||||
Root: test_helpers.EmptyContractRoot,
|
||||
}
|
||||
block2CoinbaseAccountRLP, _ = rlp.EncodeToBytes(block2CoinbaseAccount)
|
||||
block2CoinbaseLeafNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("20679cbcf198c1741a6f4e4473845659a30caa8b26f8d37a0be2e2bc0d8892"),
|
||||
block2CoinbaseAccountRLP,
|
||||
})
|
||||
block2CoinbaseLeafNodeHash = crypto.Keccak256(block2CoinbaseLeafNode)
|
||||
block2MovedPremineBalance, _ = new(big.Int).SetString("4000000000000000000000", 10)
|
||||
block2MovedPremineAccount = &types.StateAccount{
|
||||
Nonce: 0,
|
||||
Balance: block2MovedPremineBalance,
|
||||
CodeHash: test_helpers.NullCodeHash.Bytes(),
|
||||
Root: test_helpers.EmptyContractRoot,
|
||||
}
|
||||
block2MovedPremineAccountRLP, _ = rlp.EncodeToBytes(block2MovedPremineAccount)
|
||||
block2MovedPremineLeafNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("20f2e24db7943eab4415f99e109698863b0fecca1cf9ffc500f38cefbbe29e"),
|
||||
block2MovedPremineAccountRLP,
|
||||
})
|
||||
block2MovedPremineLeafNodeHash = crypto.Keccak256(block2MovedPremineLeafNode)
|
||||
block2x00080dBranchNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
block2MovedPremineLeafNodeHash,
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
block2CoinbaseLeafNodeHash,
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
})
|
||||
block2x00080dBranchNodeHash = crypto.Keccak256(block2x00080dBranchNode)
|
||||
block2x0008BranchNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("def97a26f824fc3911cf7f8c41dfc9bc93cc36ae2248de22ecae01d6950b2dc9"),
|
||||
common.Hex2Bytes("234a575e2c5badab8de0f6515b6723195323a0562fbe1316255888637043f1c1"),
|
||||
common.Hex2Bytes("29659740af1c23306ee8f8294c71a5632ace8c80b1eb61cfdf7022f47ff52305"),
|
||||
common.Hex2Bytes("cf2681d23bb666d89dec8123bce9e626240a7e2ce7a1e8316b1ee88181c9471c"),
|
||||
common.Hex2Bytes("18d8de6967fe34b9fd411c74fecc45f8a737961791e70d8ece967bb07cf4d4dc"),
|
||||
common.Hex2Bytes("7cad60c7cbca8c79c2db5a8fc1baa9381484d43d6c37dfb97718c3a109d47dfc"),
|
||||
common.Hex2Bytes("2138f5a9062b750b6320e5fac5b134da90a9edbda06ef3e1ae64fb1366ca998c"),
|
||||
common.Hex2Bytes("532826502a9661fcae7c0f5d2a4c8cb287dfc521e828349543c5a461a9d591ed"),
|
||||
common.Hex2Bytes("30543537413dd086d4b1560f46b90e8da0f43de5584a138ab036d74e84657523"),
|
||||
common.Hex2Bytes("c98042928af640bfa1142aca895cd76e146332dce94ddad3426e74ed519ca1e0"),
|
||||
common.Hex2Bytes("43de3e62cc3148193899d018dff813c04c5b636ce95bd7e828416204292d9ff9"),
|
||||
[]byte{},
|
||||
common.Hex2Bytes("78d533b9182bb42f6c16e9ebd5734f0d280179ba1c9b6316c2c1df73f7dd8a54"),
|
||||
block2x00080dBranchNodeHash,
|
||||
common.Hex2Bytes("934b736b57a892aaa15a03c7e37746bb096313727135f9841cb64c263785cf81"),
|
||||
common.Hex2Bytes("38ce97150e90dfd7258901a0ddee72d8e30760a3d0419dbb80135c66588739a2"),
|
||||
[]byte{},
|
||||
})
|
||||
block2x0008BranchNodeHash = crypto.Keccak256(block2x0008BranchNode)
|
||||
block2x00BranchNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("e45a9e85cab1b6eb18b30df2c6acc448bbac6a30d81646823b31223e16e5063e"),
|
||||
common.Hex2Bytes("33bd7171d556b981f6849064eb09412b24fedc0812127db936067043f53db1b9"),
|
||||
common.Hex2Bytes("ca56945f074da4f15587404593faf3a50d17ea0e21a418ad6ec99bdf4bf3f914"),
|
||||
common.Hex2Bytes("da23e9004f782df128eea1adff77952dc85f91b7f7ca4893aac5f21d24c3a1c9"),
|
||||
common.Hex2Bytes("ba5ec61fa780ee02af19db99677c37560fc4f0df5c278d9dfa2837f30f72bc6b"),
|
||||
common.Hex2Bytes("8310ad91625c2e3429a74066b7e2e0c958325e4e7fa3ec486b73b7c8300cfef7"),
|
||||
common.Hex2Bytes("732e5c103bf4d5adfef83773026809d9405539b67e93293a02342e83ad2fb766"),
|
||||
common.Hex2Bytes("30d14ff0c2aab57d1fbaf498ab14519b4e9d94f149a3dc15f0eec5adf8df25e1"),
|
||||
block2x0008BranchNodeHash,
|
||||
common.Hex2Bytes("5a43bd92e55aa78df60e70b6b53b6366c4080fd6a5bdd7b533b46aff4a75f6f2"),
|
||||
common.Hex2Bytes("a0c410aa59efe416b1213166fab680ce330bd46c3ebf877ff14609ee6a383600"),
|
||||
common.Hex2Bytes("2f41e918786e557293068b1eda9b3f9f86ed4e65a6a5363ee3262109f6e08b17"),
|
||||
common.Hex2Bytes("01f42a40f02f6f24bb97b09c4d3934e8b03be7cfbb902acc1c8fd67a7a5abace"),
|
||||
common.Hex2Bytes("0acbdce2787a6ea177209bd13bfc9d0779d7e2b5249e0211a2974164e14312f5"),
|
||||
common.Hex2Bytes("dadbe113e4132e0c0c3cd4867e0a2044d0e5a3d44b350677ed42fc9244d004d4"),
|
||||
common.Hex2Bytes("aa7441fefc17d76aedfcaf692fe71014b94c1547b6d129562b34fc5995ca0d1a"),
|
||||
[]byte{},
|
||||
})
|
||||
block2x00BranchNodeHash = crypto.Keccak256(block2x00BranchNode)
|
||||
block2RootBranchNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
block2x00BranchNodeHash,
|
||||
common.Hex2Bytes("babe369f6b12092f49181ae04ca173fb68d1a5456f18d20fa32cba73954052bd"),
|
||||
common.Hex2Bytes("473ecf8a7e36a829e75039a3b055e51b8332cbf03324ab4af2066bbd6fbf0021"),
|
||||
common.Hex2Bytes("bbda34753d7aa6c38e603f360244e8f59611921d9e1f128372fec0d586d4f9e0"),
|
||||
block1x04BranchNodeHash,
|
||||
common.Hex2Bytes("a5f3f2f7542148c973977c8a1e154c4300fec92f755f7846f1b734d3ab1d90e7"),
|
||||
common.Hex2Bytes("e823850f50bf72baae9d1733a36a444ab65d0a6faaba404f0583ce0ca4dad92d"),
|
||||
common.Hex2Bytes("f7a00cbe7d4b30b11faea3ae61b7f1f2b315b61d9f6bd68bfe587ad0eeceb721"),
|
||||
common.Hex2Bytes("7117ef9fc932f1a88e908eaead8565c19b5645dc9e5b1b6e841c5edbdfd71681"),
|
||||
common.Hex2Bytes("69eb2de283f32c11f859d7bcf93da23990d3e662935ed4d6b39ce3673ec84472"),
|
||||
common.Hex2Bytes("203d26456312bbc4da5cd293b75b840fc5045e493d6f904d180823ec22bfed8e"),
|
||||
common.Hex2Bytes("9287b5c21f2254af4e64fca76acc5cd87399c7f1ede818db4326c98ce2dc2208"),
|
||||
common.Hex2Bytes("6fc2d754e304c48ce6a517753c62b1a9c1d5925b89707486d7fc08919e0a94ec"),
|
||||
common.Hex2Bytes("7b1c54f15e299bd58bdfef9741538c7828b5d7d11a489f9c20d052b3471df475"),
|
||||
common.Hex2Bytes("51f9dd3739a927c89e357580a4c97b40234aa01ed3d5e0390dc982a7975880a0"),
|
||||
common.Hex2Bytes("89d613f26159af43616fd9455bb461f4869bfede26f2130835ed067a8b967bfb"),
|
||||
[]byte{},
|
||||
})
|
||||
|
||||
// block3 data
|
||||
// path 060e0f
|
||||
blcok3CoinbaseBalance, _ = new(big.Int).SetString("5156250000000000000", 10)
|
||||
block3CoinbaseAccount = &types.StateAccount{
|
||||
Nonce: 0,
|
||||
Balance: blcok3CoinbaseBalance,
|
||||
CodeHash: test_helpers.NullCodeHash.Bytes(),
|
||||
Root: test_helpers.EmptyContractRoot,
|
||||
}
|
||||
block3CoinbaseAccountRLP, _ = rlp.EncodeToBytes(block3CoinbaseAccount)
|
||||
block3CoinbaseLeafNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("3a174f00e64521a535f35e67c1aa241951c791639b2f3d060f49c5d9fa8b9e"),
|
||||
block3CoinbaseAccountRLP,
|
||||
})
|
||||
block3CoinbaseLeafNodeHash = crypto.Keccak256(block3CoinbaseLeafNode)
|
||||
// path 0c0e050703
|
||||
block3MovedPremineBalance1, _ = new(big.Int).SetString("3750000000000000000", 10)
|
||||
block3MovedPremineAccount1 = &types.StateAccount{
|
||||
Nonce: 0,
|
||||
Balance: block3MovedPremineBalance1,
|
||||
CodeHash: test_helpers.NullCodeHash.Bytes(),
|
||||
Root: test_helpers.EmptyContractRoot,
|
||||
}
|
||||
block3MovedPremineAccount1RLP, _ = rlp.EncodeToBytes(block3MovedPremineAccount1)
|
||||
block3MovedPremineLeafNode1, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("3ced93917e658d10e2d9009470dad72b63c898d173721194a12f2ae5e190"), // ce573ced93917e658d10e2d9009470dad72b63c898d173721194a12f2ae5e190
|
||||
block3MovedPremineAccount1RLP,
|
||||
})
|
||||
block3MovedPremineLeafNodeHash1 = crypto.Keccak256(block3MovedPremineLeafNode1)
|
||||
// path 0c0e050708
|
||||
block3MovedPremineBalance2, _ = new(big.Int).SetString("1999944000000000000000", 10)
|
||||
block3MovedPremineAccount2 = &types.StateAccount{
|
||||
Nonce: 0,
|
||||
Balance: block3MovedPremineBalance2,
|
||||
CodeHash: test_helpers.NullCodeHash.Bytes(),
|
||||
Root: test_helpers.EmptyContractRoot,
|
||||
}
|
||||
block3MovedPremineAccount2RLP, _ = rlp.EncodeToBytes(block3MovedPremineAccount2)
|
||||
block3MovedPremineLeafNode2, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("33bc1e69eedf90f402e11f6862da14ed8e50156635a04d6393bbae154012"), // ce5783bc1e69eedf90f402e11f6862da14ed8e50156635a04d6393bbae154012
|
||||
block3MovedPremineAccount2RLP,
|
||||
})
|
||||
block3MovedPremineLeafNodeHash2 = crypto.Keccak256(block3MovedPremineLeafNode2)
|
||||
|
||||
block3x0c0e0507BranchNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
block3MovedPremineLeafNodeHash1,
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
block3MovedPremineLeafNodeHash2,
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
})
|
||||
block3x0c0e0507BranchNodeHash = crypto.Keccak256(block3x0c0e0507BranchNode)
|
||||
|
||||
block3x0c0e05BranchNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("452e3beb503b1d87ae7c672b98a8e3fd043a671405502562ae1043dc97151a50"),
|
||||
[]byte{},
|
||||
common.Hex2Bytes("2f5bb16f77086f67ce8c4258cb9061cb299e597b2ad4ad6d7ccc474d6d88e85e"),
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
block3x0c0e0507BranchNodeHash,
|
||||
[]byte{},
|
||||
common.Hex2Bytes("44623e5a9319f83870db0ea4611a25fca1e1da3eeea2be4a091dfc15ab45689e"),
|
||||
common.Hex2Bytes("b41e047a97f44fa4cb8146467b88c8f4705811029d9e170abb0aba7d0af9f0da"),
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
[]byte{},
|
||||
})
|
||||
block3x0c0e05BranchNodeHash = crypto.Keccak256(block3x0c0e05BranchNode)
|
||||
|
||||
block3x060eBranchNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("94d77c7c30b88829c9989948b206cda5e532b38b49534261c517aebf4a3e6fdb"),
|
||||
common.Hex2Bytes("a5cf57a50da8204964e834a12a53f9bed7afc9b700a4a81b440122d60c7603a7"),
|
||||
[]byte{},
|
||||
common.Hex2Bytes("3730ec0571f34b6c3b178dc26ccb31a3f50c29da9b1921e41b9477ddab41b0fe"),
|
||||
[]byte{},
|
||||
common.Hex2Bytes("543952bb9566c2018cf8d7b90d6a7903cdfff3d79ac36189be5322de42fc3fc0"),
|
||||
[]byte{},
|
||||
common.Hex2Bytes("c4a49b66f0bcc08531e50cdea5577a281d111fa542eaefd9a9aead8febb0735e"),
|
||||
common.Hex2Bytes("362ad58916c71463b98c079649fc486c5f082c4f548bd4ab501515f0c5641cb4"),
|
||||
common.Hex2Bytes("36aae109f6f55f0bd05eb05bb365af2332dfe5f06d3d17903e88534c319eb709"),
|
||||
common.Hex2Bytes("430dcfc5cc49a6b490dd54138920e8f94e427239c2bccc14705cfd4ff6cc4383"),
|
||||
common.Hex2Bytes("73ed77563dfed2fdb38900b474db88b2270f449167e0d877fda9e2229f119fe8"),
|
||||
common.Hex2Bytes("5dfe06013f2a41f1779194ceb07769d019f518b2a694a82fa1661e60fd973eaa"),
|
||||
common.Hex2Bytes("80bdfd85fbb6b45850bad6e34136aaa1b04711e47469fa2f0d19eca52089efb5"),
|
||||
[]byte{},
|
||||
block3CoinbaseLeafNodeHash,
|
||||
[]byte{},
|
||||
})
|
||||
block3x060eBranchNodeHash = crypto.Keccak256(block3x060eBranchNode)
|
||||
|
||||
block3x0c0eBranchNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("70647f11b2b995d718f9e8aceb44c8839e0055641930d216fa6090280a9d63d5"),
|
||||
common.Hex2Bytes("fdfb17cd2fba2a14219981cb7886a1977cd85dbef5c767c562f4a5f547febff0"),
|
||||
common.Hex2Bytes("ff87313253ec6f860142b7bf62efb4cb07ea668c57aa90cbe9ef22b72fee15c7"),
|
||||
common.Hex2Bytes("3a77b3c26a54ad37bdf4e19c1bce93493ec0f79d9ad90190b70bc840b54918e1"),
|
||||
common.Hex2Bytes("af1b3b14324561b68f2e24dbcc28673ab35ce3fd0230fe2bc86b3d1931745195"),
|
||||
block3x0c0e05BranchNodeHash,
|
||||
common.Hex2Bytes("647dcbfe6aabcd9d219ff40422af4326bfc1ec66703195a78eb48618ddef248d"),
|
||||
common.Hex2Bytes("2d2bf06159cc8928283c3419a03f08ea34c493a9d002a0ec76d5c429508ccaf4"),
|
||||
common.Hex2Bytes("d7147251b3f48f25e1e4c6d8f83a00b1eca66e99a4ea0d238942ce72d0ba6414"),
|
||||
common.Hex2Bytes("cb859370869967594fb29f4e2904413310146733d7fcbd11407f3e47626e0e34"),
|
||||
common.Hex2Bytes("b93ab9b0bd83963860fbe0b7d543879cfde756ea1618d2a40d85483058cc5a26"),
|
||||
common.Hex2Bytes("45aee096499d209931457ce251c5c7e5543f22524f67785ff8f0f3f02588b0ed"),
|
||||
[]byte{},
|
||||
common.Hex2Bytes("aa2ae9379797c5066bba646108074ae8677e82c923d584b6d1c1268ca3708c5c"),
|
||||
common.Hex2Bytes("e6eb055f0d8e194c083471479a3de87fa0f90c0f4aaa518416ec1e469ec32e3a"),
|
||||
common.Hex2Bytes("0cc9c50fc7eba162fb17f2e04e3599c13abbf210d9781864d0edec401ecaebba"),
|
||||
[]byte{},
|
||||
})
|
||||
block3x0c0eBranchNodeHash = crypto.Keccak256(block3x0c0eBranchNode)
|
||||
|
||||
block3x06BranchNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("68f7ff8c074d6e4cccd55b5b1c2116a6dd7047d4332090e6db8839362991b0ae"),
|
||||
common.Hex2Bytes("c446eb4377c750701374c56e50759e6ba68b7adf4d543e718c8b28a99ae3b6ad"),
|
||||
common.Hex2Bytes("ef2c49ec64cb65eae0d99684e74c8af2bd0206c9a0214d9d3eddf0881dd8412a"),
|
||||
common.Hex2Bytes("7096c4cc7e8125f0b142d8644ad681f8a8142e210c806f33f3f7004f0e9d6002"),
|
||||
common.Hex2Bytes("bc9a8ae647b234cd6607b6b0245e3b3d5ec4f7ea006e7eda1f92d02f0ea91116"),
|
||||
common.Hex2Bytes("a87720deb92ff2f899e809befab9970a61c86148c4fa09d04b77505ee4a5bda5"),
|
||||
common.Hex2Bytes("2460e5b6ded7c0001de29c15db124614432fef6486370cc9970f63b0d95fd5e2"),
|
||||
common.Hex2Bytes("ed1c447d4a32bc31e9e32259dc63da10df91231e786332e3df122b301b1f8fc3"),
|
||||
common.Hex2Bytes("0d27dfc201d995c2323b792860dbca087da7cc56d1698c39b7c4b9277729c5ca"),
|
||||
common.Hex2Bytes("f6d2be168d9c17643c9ea80c29322b364604cdfd36eef40123d83fad364e43fa"),
|
||||
common.Hex2Bytes("004bf1c30a5730f464de1a0ba4ac5b5618df66d6106073d08742166e33a7eeb5"),
|
||||
common.Hex2Bytes("7298d019a57a1b04ac31ed874d654ba0d3c249704c5d9efa1d08959fc89e0779"),
|
||||
common.Hex2Bytes("fb3d50b7af6f839e371ff8ebd0322e94e6b6fb7888416737f88cf55bcf5859ec"),
|
||||
common.Hex2Bytes("4e7a2618fa1fc560a73c24839657adf7e48d600ecfb12333678115936597a913"),
|
||||
block3x060eBranchNodeHash,
|
||||
common.Hex2Bytes("1909706c5db040f54c19f4050659ad484982145b02474653917de379f15ebb36"),
|
||||
[]byte{},
|
||||
})
|
||||
block3x06BranchNodeHash = crypto.Keccak256(block3x06BranchNode)
|
||||
|
||||
block3x0cBranchNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("dae48f5b47930c28bb116fbd55e52cd47242c71bf55373b55eb2805ee2e4a929"),
|
||||
common.Hex2Bytes("0f1f37f337ec800e2e5974e2e7355f10f1a4832b39b846d916c3597a460e0676"),
|
||||
common.Hex2Bytes("da8f627bb8fbeead17b318e0a8e4f528db310f591bb6ab2deda4a9f7ca902ab5"),
|
||||
common.Hex2Bytes("971c662648d58295d0d0aa4b8055588da0037619951217c22052802549d94a2f"),
|
||||
common.Hex2Bytes("ccc701efe4b3413fd6a61a6c9f40e955af774649a8d9fd212d046a5a39ddbb67"),
|
||||
common.Hex2Bytes("d607cdb32e2bd635ee7f2f9e07bc94ddbd09b10ec0901b66628e15667aec570b"),
|
||||
common.Hex2Bytes("5b89203dc940e6fa70ec19ad4e01d01849d3a5baa0a8f9c0525256ed490b159f"),
|
||||
common.Hex2Bytes("b84227d48df68aecc772939a59afa9e1a4ab578f7b698bdb1289e29b6044668e"),
|
||||
common.Hex2Bytes("fd1c992070b94ace57e48cbf6511a16aa770c645f9f5efba87bbe59d0a042913"),
|
||||
common.Hex2Bytes("e16a7ccea6748ae90de92f8aef3b3dc248a557b9ac4e296934313f24f7fced5f"),
|
||||
common.Hex2Bytes("42373cf4a00630d94de90d0a23b8f38ced6b0f7cb818b8925fee8f0c2a28a25a"),
|
||||
common.Hex2Bytes("5f89d2161c1741ff428864f7889866484cef622de5023a46e795dfdec336319f"),
|
||||
common.Hex2Bytes("7597a017664526c8c795ce1da27b8b72455c49657113e0455552dbc068c5ba31"),
|
||||
common.Hex2Bytes("d5be9089012fda2c585a1b961e988ea5efcd3a06988e150a8682091f694b37c5"),
|
||||
block3x0c0eBranchNodeHash,
|
||||
common.Hex2Bytes("49bf6e8df0acafd0eff86defeeb305568e44d52d2235cf340ae15c6034e2b241"),
|
||||
[]byte{},
|
||||
})
|
||||
block3x0cBranchNodeHash = crypto.Keccak256(block3x0cBranchNode)
|
||||
|
||||
block3RootBranchNode, _ = rlp.EncodeToBytes(&[]interface{}{
|
||||
common.Hex2Bytes("f646da473c426e79f1c796b00d4873f47de1dbe1c9d19d63993a05eeb8b4041d"),
|
||||
common.Hex2Bytes("babe369f6b12092f49181ae04ca173fb68d1a5456f18d20fa32cba73954052bd"),
|
||||
common.Hex2Bytes("473ecf8a7e36a829e75039a3b055e51b8332cbf03324ab4af2066bbd6fbf0021"),
|
||||
common.Hex2Bytes("bbda34753d7aa6c38e603f360244e8f59611921d9e1f128372fec0d586d4f9e0"),
|
||||
common.Hex2Bytes("d9cff5d5f2418afd16a4da5c221fdc8bd47520c5927922f69a68177b64da6ac0"),
|
||||
common.Hex2Bytes("a5f3f2f7542148c973977c8a1e154c4300fec92f755f7846f1b734d3ab1d90e7"),
|
||||
block3x06BranchNodeHash,
|
||||
common.Hex2Bytes("f7a00cbe7d4b30b11faea3ae61b7f1f2b315b61d9f6bd68bfe587ad0eeceb721"),
|
||||
common.Hex2Bytes("7117ef9fc932f1a88e908eaead8565c19b5645dc9e5b1b6e841c5edbdfd71681"),
|
||||
common.Hex2Bytes("69eb2de283f32c11f859d7bcf93da23990d3e662935ed4d6b39ce3673ec84472"),
|
||||
common.Hex2Bytes("203d26456312bbc4da5cd293b75b840fc5045e493d6f904d180823ec22bfed8e"),
|
||||
common.Hex2Bytes("9287b5c21f2254af4e64fca76acc5cd87399c7f1ede818db4326c98ce2dc2208"),
|
||||
block3x0cBranchNodeHash,
|
||||
common.Hex2Bytes("7b1c54f15e299bd58bdfef9741538c7828b5d7d11a489f9c20d052b3471df475"),
|
||||
common.Hex2Bytes("51f9dd3739a927c89e357580a4c97b40234aa01ed3d5e0390dc982a7975880a0"),
|
||||
common.Hex2Bytes("89d613f26159af43616fd9455bb461f4869bfede26f2130835ed067a8b967bfb"),
|
||||
[]byte{},
|
||||
})
|
||||
)
|
||||
|
||||
func init() {
|
||||
if os.Getenv("MODE") != "statediff" {
|
||||
fmt.Println("Skipping statediff test")
|
||||
os.Exit(0)
|
||||
}
|
||||
db = rawdb.NewMemoryDatabase()
|
||||
genesisBlock = core.DefaultGenesisBlock().MustCommit(db)
|
||||
genBy, err := rlp.EncodeToBytes(genesisBlock)
|
||||
if err != nil {
|
||||
log.Fatal(err)
|
||||
}
|
||||
var block0RLP []byte
|
||||
block0, block0RLP, err = loadBlockFromRLPFile("./block0_rlp")
|
||||
if err != nil {
|
||||
log.Fatal(err)
|
||||
}
|
||||
if !bytes.Equal(genBy, block0RLP) {
|
||||
log.Fatal("mainnet genesis blocks do not match")
|
||||
}
|
||||
block1, _, err = loadBlockFromRLPFile("./block1_rlp")
|
||||
if err != nil {
|
||||
log.Fatal(err)
|
||||
}
|
||||
block1CoinbaseAddr = block1.Coinbase()
|
||||
block1CoinbaseHash = crypto.Keccak256Hash(block1CoinbaseAddr.Bytes())
|
||||
block2, _, err = loadBlockFromRLPFile("./block2_rlp")
|
||||
if err != nil {
|
||||
log.Fatal(err)
|
||||
}
|
||||
block2CoinbaseAddr = block2.Coinbase()
|
||||
block2CoinbaseHash = crypto.Keccak256Hash(block2CoinbaseAddr.Bytes()) // 0x08d4679cbcf198c1741a6f4e4473845659a30caa8b26f8d37a0be2e2bc0d8892
|
||||
block3, _, err = loadBlockFromRLPFile("./block3_rlp")
|
||||
if err != nil {
|
||||
log.Fatal(err)
|
||||
}
|
||||
block3CoinbaseAddr = block3.Coinbase()
|
||||
block3CoinbaseHash = crypto.Keccak256Hash(block3CoinbaseAddr.Bytes())
|
||||
}
|
||||
|
||||
func loadBlockFromRLPFile(filename string) (*types.Block, []byte, error) {
|
||||
f, err := os.Open(filename)
|
||||
if err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
defer f.Close()
|
||||
blockRLP, err := io.ReadAll(f)
|
||||
if err != nil {
|
||||
return nil, nil, err
|
||||
}
|
||||
block := new(types.Block)
|
||||
return block, blockRLP, rlp.DecodeBytes(blockRLP, block)
|
||||
}
|
||||
|
||||
func TestBuilderOnMainnetBlocks(t *testing.T) {
|
||||
chain, _ := core.NewBlockChain(db, nil, nil, nil, ethash.NewFaker(), vm.Config{}, nil, nil)
|
||||
_, err := chain.InsertChain([]*types.Block{block1, block2, block3})
|
||||
if err != nil {
|
||||
t.Error(err)
|
||||
}
|
||||
params := statediff.Params{}
|
||||
builder = statediff.NewBuilder(chain.StateCache())
|
||||
|
||||
var tests = []struct {
|
||||
name string
|
||||
startingArguments statediff.Args
|
||||
expected *sdtypes.StateObject
|
||||
}{
|
||||
// note that block0 (genesis) has over 1000 nodes due to the pre-allocation for the crowd-sale
|
||||
// it is not feasible to write a unit test of that size at this time
|
||||
{
|
||||
"testBlock1",
|
||||
//10000 transferred from testBankAddress to account1Addr
|
||||
statediff.Args{
|
||||
OldStateRoot: block0.Root(),
|
||||
NewStateRoot: block1.Root(),
|
||||
BlockNumber: block1.Number(),
|
||||
BlockHash: block1.Hash(),
|
||||
},
|
||||
&sdtypes.StateObject{
|
||||
BlockNumber: block1.Number(),
|
||||
BlockHash: block1.Hash(),
|
||||
Nodes: []sdtypes.StateLeafNode{
|
||||
{
|
||||
Removed: false,
|
||||
AccountWrapper: sdtypes.AccountWrapper{
|
||||
Account: block1CoinbaseAccount,
|
||||
LeafKey: block1CoinbaseHash.Bytes(),
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block1CoinbaseLeafNode)).String(),
|
||||
},
|
||||
StorageDiff: emptyStorage,
|
||||
},
|
||||
},
|
||||
IPLDs: []sdtypes.IPLD{
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block1RootBranchNode)).String(),
|
||||
Content: block1RootBranchNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block1x04BranchNode)).String(),
|
||||
Content: block1x04BranchNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block1x040bBranchNode)).String(),
|
||||
Content: block1x040bBranchNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block1CoinbaseLeafNode)).String(),
|
||||
Content: block1CoinbaseLeafNode,
|
||||
},
|
||||
},
|
||||
},
|
||||
},
|
||||
{
|
||||
"testBlock2",
|
||||
// 1000 transferred from testBankAddress to account1Addr
|
||||
// 1000 transferred from account1Addr to account2Addr
|
||||
// account1addr creates a new contract
|
||||
statediff.Args{
|
||||
OldStateRoot: block1.Root(),
|
||||
NewStateRoot: block2.Root(),
|
||||
BlockNumber: block2.Number(),
|
||||
BlockHash: block2.Hash(),
|
||||
},
|
||||
&sdtypes.StateObject{
|
||||
BlockNumber: block2.Number(),
|
||||
BlockHash: block2.Hash(),
|
||||
Nodes: []sdtypes.StateLeafNode{
|
||||
{
|
||||
Removed: false,
|
||||
AccountWrapper: sdtypes.AccountWrapper{
|
||||
Account: block2CoinbaseAccount,
|
||||
LeafKey: block2CoinbaseHash.Bytes(),
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block2CoinbaseLeafNode)).String(),
|
||||
},
|
||||
StorageDiff: emptyStorage,
|
||||
},
|
||||
},
|
||||
IPLDs: []sdtypes.IPLD{
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block2RootBranchNode)).String(),
|
||||
Content: block2RootBranchNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block2x00BranchNode)).String(),
|
||||
Content: block2x00BranchNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block2x0008BranchNode)).String(),
|
||||
Content: block2x0008BranchNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block2x00080dBranchNode)).String(),
|
||||
Content: block2x00080dBranchNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block2MovedPremineLeafNode)).String(),
|
||||
Content: block2MovedPremineLeafNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block2CoinbaseLeafNode)).String(),
|
||||
Content: block2CoinbaseLeafNode,
|
||||
},
|
||||
},
|
||||
},
|
||||
},
|
||||
{
|
||||
"testBlock3",
|
||||
//the contract's storage is changed
|
||||
//and the block is mined by account 2
|
||||
statediff.Args{
|
||||
OldStateRoot: block2.Root(),
|
||||
NewStateRoot: block3.Root(),
|
||||
BlockNumber: block3.Number(),
|
||||
BlockHash: block3.Hash(),
|
||||
},
|
||||
&sdtypes.StateObject{
|
||||
BlockNumber: block3.Number(),
|
||||
BlockHash: block3.Hash(),
|
||||
Nodes: []sdtypes.StateLeafNode{
|
||||
{ // How was this account created???
|
||||
Removed: false,
|
||||
AccountWrapper: sdtypes.AccountWrapper{
|
||||
Account: block3MovedPremineAccount1,
|
||||
LeafKey: common.HexToHash("ce573ced93917e658d10e2d9009470dad72b63c898d173721194a12f2ae5e190").Bytes(),
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block3MovedPremineLeafNode1)).String(),
|
||||
},
|
||||
StorageDiff: emptyStorage,
|
||||
},
|
||||
{ // this is the new account created due to the coinbase mining a block, it's creation shouldn't affect 0x 0e 05 07
|
||||
Removed: false,
|
||||
AccountWrapper: sdtypes.AccountWrapper{
|
||||
Account: block3CoinbaseAccount,
|
||||
LeafKey: block3CoinbaseHash.Bytes(),
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block3CoinbaseLeafNode)).String(),
|
||||
},
|
||||
StorageDiff: emptyStorage,
|
||||
},
|
||||
},
|
||||
IPLDs: []sdtypes.IPLD{
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block3RootBranchNode)).String(),
|
||||
Content: block3RootBranchNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block3x06BranchNode)).String(),
|
||||
Content: block3x06BranchNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block3x060eBranchNode)).String(),
|
||||
Content: block3x060eBranchNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block3x0cBranchNode)).String(),
|
||||
Content: block3x0cBranchNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block3x0c0eBranchNode)).String(),
|
||||
Content: block3x0c0eBranchNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block3x0c0e05BranchNode)).String(),
|
||||
Content: block3x0c0e05BranchNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block3x0c0e0507BranchNode)).String(),
|
||||
Content: block3x0c0e0507BranchNode,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block3MovedPremineLeafNode1)).String(),
|
||||
Content: block3MovedPremineLeafNode1,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block3MovedPremineLeafNode2)).String(),
|
||||
Content: block3MovedPremineLeafNode2,
|
||||
},
|
||||
{
|
||||
CID: ipld2.Keccak256ToCid(ipld2.MEthStateTrie, crypto.Keccak256(block3CoinbaseLeafNode)).String(),
|
||||
Content: block3CoinbaseLeafNode,
|
||||
},
|
||||
},
|
||||
},
|
||||
},
|
||||
}
|
||||
|
||||
for _, test := range tests {
|
||||
diff, err := builder.BuildStateDiffObject(test.startingArguments, params)
|
||||
if err != nil {
|
||||
t.Error(err)
|
||||
}
|
||||
receivedStateDiffRlp, err := rlp.EncodeToBytes(diff)
|
||||
if err != nil {
|
||||
t.Error(err)
|
||||
}
|
||||
expectedStateDiffRlp, err := rlp.EncodeToBytes(&test.expected)
|
||||
if err != nil {
|
||||
t.Error(err)
|
||||
}
|
||||
sort.Slice(receivedStateDiffRlp, func(i, j int) bool { return receivedStateDiffRlp[i] < receivedStateDiffRlp[j] })
|
||||
sort.Slice(expectedStateDiffRlp, func(i, j int) bool { return expectedStateDiffRlp[i] < expectedStateDiffRlp[j] })
|
||||
if !bytes.Equal(receivedStateDiffRlp, expectedStateDiffRlp) {
|
||||
actual, err := json.Marshal(diff)
|
||||
if err != nil {
|
||||
t.Error(err)
|
||||
}
|
||||
expected, err := json.Marshal(test.expected)
|
||||
if err != nil {
|
||||
t.Error(err)
|
||||
}
|
||||
t.Logf("Test failed: %s", test.name)
|
||||
t.Errorf("actual state diff: %s\r\n\r\n\r\nexpected state diff: %s", actual, expected)
|
||||
}
|
||||
}
|
||||
if !bytes.Equal(crypto.Keccak256(block1RootBranchNode), block1.Root().Bytes()) {
|
||||
t.Errorf("actual state root: %s\r\nexpected state root: %s", crypto.Keccak256(block1RootBranchNode), block1.Root().Bytes())
|
||||
}
|
||||
if !bytes.Equal(crypto.Keccak256(block2RootBranchNode), block2.Root().Bytes()) {
|
||||
t.Errorf("actual state root: %s\r\nexpected state root: %s", crypto.Keccak256(block2RootBranchNode), block2.Root().Bytes())
|
||||
}
|
||||
if !bytes.Equal(crypto.Keccak256(block3RootBranchNode), block3.Root().Bytes()) {
|
||||
t.Errorf("actual state root: %s\r\nexpected state root: %s", crypto.Keccak256(block3RootBranchNode), block3.Root().Bytes())
|
||||
}
|
||||
}
|
108
metrics.go
Normal file
108
metrics.go
Normal file
@ -0,0 +1,108 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package statediff
|
||||
|
||||
import (
|
||||
"strings"
|
||||
|
||||
"github.com/ethereum/go-ethereum/metrics"
|
||||
)
|
||||
|
||||
const (
|
||||
namespace = "statediff"
|
||||
)
|
||||
|
||||
var defaultStatediffMetrics = RegisterStatediffMetrics(metrics.DefaultRegistry)
|
||||
|
||||
// Build a fully qualified metric name
|
||||
func metricName(subsystem, name string) string {
|
||||
if name == "" {
|
||||
return ""
|
||||
}
|
||||
parts := []string{namespace, name}
|
||||
if subsystem != "" {
|
||||
parts = []string{namespace, subsystem, name}
|
||||
}
|
||||
// Prometheus uses _ but geth metrics uses / and replaces
|
||||
return strings.Join(parts, "/")
|
||||
}
|
||||
|
||||
type statediffMetricsHandles struct {
|
||||
// Height of latest synced by core.BlockChain
|
||||
// FIXME
|
||||
lastSyncHeight metrics.Gauge
|
||||
// Height of the latest block received from chainEvent channel
|
||||
lastEventHeight metrics.Gauge
|
||||
// Height of latest state diff
|
||||
lastStatediffHeight metrics.Gauge
|
||||
// Current length of chainEvent channels
|
||||
serviceLoopChannelLen metrics.Gauge
|
||||
writeLoopChannelLen metrics.Gauge
|
||||
// The start block of the known gap
|
||||
knownGapStart metrics.Gauge
|
||||
// The end block of the known gap
|
||||
knownGapEnd metrics.Gauge
|
||||
// A known gaps start block which had an error being written to the DB
|
||||
knownGapErrorStart metrics.Gauge
|
||||
// A known gaps end block which had an error being written to the DB
|
||||
knownGapErrorEnd metrics.Gauge
|
||||
|
||||
apiRequests metrics.Counter
|
||||
apiRequestsUnderway metrics.Counter
|
||||
|
||||
failed metrics.Counter
|
||||
succeeded metrics.Counter
|
||||
underway metrics.Counter
|
||||
totalProcessingTime metrics.Gauge
|
||||
}
|
||||
|
||||
func RegisterStatediffMetrics(reg metrics.Registry) statediffMetricsHandles {
|
||||
ctx := statediffMetricsHandles{
|
||||
lastSyncHeight: metrics.NewGauge(),
|
||||
lastEventHeight: metrics.NewGauge(),
|
||||
lastStatediffHeight: metrics.NewGauge(),
|
||||
serviceLoopChannelLen: metrics.NewGauge(),
|
||||
writeLoopChannelLen: metrics.NewGauge(),
|
||||
knownGapStart: metrics.NewGauge(),
|
||||
knownGapEnd: metrics.NewGauge(),
|
||||
knownGapErrorStart: metrics.NewGauge(),
|
||||
knownGapErrorEnd: metrics.NewGauge(),
|
||||
apiRequests: metrics.NewCounter(),
|
||||
apiRequestsUnderway: metrics.NewCounter(),
|
||||
failed: metrics.NewCounter(),
|
||||
succeeded: metrics.NewCounter(),
|
||||
underway: metrics.NewCounter(),
|
||||
totalProcessingTime: metrics.NewGauge(),
|
||||
}
|
||||
subsys := "service"
|
||||
reg.Register(metricName(subsys, "last_sync_height"), ctx.lastSyncHeight)
|
||||
reg.Register(metricName(subsys, "last_event_height"), ctx.lastEventHeight)
|
||||
reg.Register(metricName(subsys, "last_statediff_height"), ctx.lastStatediffHeight)
|
||||
reg.Register(metricName(subsys, "service_loop_channel_len"), ctx.serviceLoopChannelLen)
|
||||
reg.Register(metricName(subsys, "write_loop_channel_len"), ctx.writeLoopChannelLen)
|
||||
reg.Register(metricName(subsys, "known_gaps_start"), ctx.knownGapStart)
|
||||
reg.Register(metricName(subsys, "known_gaps_end"), ctx.knownGapEnd)
|
||||
reg.Register(metricName(subsys, "known_gaps_error_start"), ctx.knownGapErrorStart)
|
||||
reg.Register(metricName(subsys, "known_gaps_error_end"), ctx.knownGapErrorEnd)
|
||||
reg.Register(metricName(subsys, "api_requests"), ctx.apiRequests)
|
||||
reg.Register(metricName(subsys, "api_requests_underway"), ctx.apiRequestsUnderway)
|
||||
reg.Register(metricName(subsys, "failed"), ctx.failed)
|
||||
reg.Register(metricName(subsys, "succeeded"), ctx.succeeded)
|
||||
reg.Register(metricName(subsys, "underway"), ctx.underway)
|
||||
reg.Register(metricName(subsys, "total_processing_time"), ctx.totalProcessingTime)
|
||||
return ctx
|
||||
}
|
89
metrics_helpers.go
Normal file
89
metrics_helpers.go
Normal file
@ -0,0 +1,89 @@
|
||||
// Copyright 2019 The go-ethereum Authors
|
||||
// This file is part of the go-ethereum library.
|
||||
//
|
||||
// The go-ethereum library is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Lesser General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
//
|
||||
// The go-ethereum library is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Lesser General Public License for more details.
|
||||
//
|
||||
// You should have received a copy of the GNU Lesser General Public License
|
||||
// along with the go-ethereum library. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package statediff
|
||||
|
||||
import (
|
||||
"fmt"
|
||||
"time"
|
||||
|
||||
"github.com/ethereum/go-ethereum/core/types"
|
||||
"github.com/ethereum/go-ethereum/log"
|
||||
)
|
||||
|
||||
func countStateDiffBegin(block *types.Block) (time.Time, log.Logger) {
|
||||
start := time.Now()
|
||||
logger := log.New("hash", block.Hash().Hex(), "number", block.NumberU64())
|
||||
|
||||
defaultStatediffMetrics.underway.Inc(1)
|
||||
logger.Debug(fmt.Sprintf("writeStateDiff BEGIN [underway=%d, succeeded=%d, failed=%d, total_time=%dms]",
|
||||
defaultStatediffMetrics.underway.Count(),
|
||||
defaultStatediffMetrics.succeeded.Count(),
|
||||
defaultStatediffMetrics.failed.Count(),
|
||||
defaultStatediffMetrics.totalProcessingTime.Value(),
|
||||
))
|
||||
|
||||
return start, logger
|
||||
}
|
||||
|
||||
func countStateDiffEnd(start time.Time, logger log.Logger, err error) time.Duration {
|
||||
duration := time.Since(start)
|
||||
defaultStatediffMetrics.underway.Dec(1)
|
||||
if nil == err {
|
||||
defaultStatediffMetrics.succeeded.Inc(1)
|
||||
} else {
|
||||
defaultStatediffMetrics.failed.Inc(1)
|
||||
}
|
||||
defaultStatediffMetrics.totalProcessingTime.Inc(duration.Milliseconds())
|
||||
|
||||
logger.Debug(fmt.Sprintf("writeStateDiff END (duration=%dms, err=%t) [underway=%d, succeeded=%d, failed=%d, total_time=%dms]",
|
||||
duration.Milliseconds(), nil != err,
|
||||
defaultStatediffMetrics.underway.Count(),
|
||||
defaultStatediffMetrics.succeeded.Count(),
|
||||
defaultStatediffMetrics.failed.Count(),
|
||||
defaultStatediffMetrics.totalProcessingTime.Value(),
|
||||
))
|
||||
|
||||
return duration
|
||||
}
|
||||
|
||||
func countApiRequestBegin(methodName string, blockHashOrNumber interface{}) (time.Time, log.Logger) {
|
||||
start := time.Now()
|
||||
logger := log.New(methodName, blockHashOrNumber)
|
||||
|
||||
defaultStatediffMetrics.apiRequests.Inc(1)
|
||||
defaultStatediffMetrics.apiRequestsUnderway.Inc(1)
|
||||
|
||||
logger.Debug(fmt.Sprintf("statediff API BEGIN [underway=%d, requests=%d])",
|
||||
defaultStatediffMetrics.apiRequestsUnderway.Count(),
|
||||
defaultStatediffMetrics.apiRequests.Count(),
|
||||
))
|
||||
|
||||
return start, logger
|
||||
}
|
||||
|
||||
func countApiRequestEnd(start time.Time, logger log.Logger, err error) time.Duration {
|
||||
duration := time.Since(start)
|
||||
defaultStatediffMetrics.apiRequestsUnderway.Dec(1)
|
||||
|
||||
logger.Debug(fmt.Sprintf("statediff API END (duration=%dms, err=%t) [underway=%d, requests=%d]",
|
||||
duration.Milliseconds(), nil != err,
|
||||
defaultStatediffMetrics.apiRequestsUnderway.Count(),
|
||||
defaultStatediffMetrics.apiRequests.Count(),
|
||||
))
|
||||
|
||||
return duration
|
||||
}
|
57
payload.go
Normal file
57
payload.go
Normal file
@ -0,0 +1,57 @@
|
||||
// VulcanizeDB
|
||||
// Copyright © 2021 Vulcanize
|
||||
|
||||
// This program is free software: you can redistribute it and/or modify
|
||||
// it under the terms of the GNU Affero General Public License as published by
|
||||
// the Free Software Foundation, either version 3 of the License, or
|
||||
// (at your option) any later version.
|
||||
|
||||
// This program is distributed in the hope that it will be useful,
|
||||
// but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
// GNU Affero General Public License for more details.
|
||||
|
||||
// You should have received a copy of the GNU Affero General Public License
|
||||
// along with this program. If not, see <http://www.gnu.org/licenses/>.
|
||||
|
||||
package statediff
|
||||
|
||||
import (
|
||||
"encoding/json"
|
||||
"math/big"
|
||||
)
|
||||
|
||||
// Payload packages the data to send to statediff subscriptions
|
||||
type Payload struct {
|
||||
BlockRlp []byte `json:"blockRlp"`
|
||||
TotalDifficulty *big.Int `json:"totalDifficulty"`
|
||||
ReceiptsRlp []byte `json:"receiptsRlp"`
|
||||
StateObjectRlp []byte `json:"stateObjectRlp" gencodec:"required"`
|
||||
|
||||
encoded []byte
|
||||
err error
|
||||
}
|
||||
|
||||
func (sd *Payload) ensureEncoded() {
|
||||
if sd.encoded == nil && sd.err == nil {
|
||||
sd.encoded, sd.err = json.Marshal(sd)
|
||||
}
|
||||
}
|
||||
|
||||
// Length to implement Encoder interface for Payload
|
||||
func (sd *Payload) Length() int {
|
||||
sd.ensureEncoded()
|
||||
return len(sd.encoded)
|
||||
}
|
||||
|
||||
// Encode to implement Encoder interface for Payload
|
||||
func (sd *Payload) Encode() ([]byte, error) {
|
||||
sd.ensureEncoded()
|
||||
return sd.encoded, sd.err
|
||||
}
|
||||
|
||||
// Subscription struct holds our subscription channels
|
||||
type Subscription struct {
|
||||
PayloadChan chan<- Payload
|
||||
QuitChan chan<- bool
|
||||
}
|
Some files were not shown because too many files have changed in this diff Show More
Loading…
Reference in New Issue
Block a user