plugeth-statediff/indexer/database/file/csv_writer.go

416 lines
12 KiB
Go
Raw Normal View History

2023-06-14 12:43:34 +00:00
// 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/ethereum/go-ethereum/common"
2023-06-14 12:43:34 +00:00
"github.com/thoas/go-funk"
"github.com/cerc-io/plugeth-statediff/indexer/database/metrics"
"github.com/cerc-io/plugeth-statediff/indexer/ipld"
"github.com/cerc-io/plugeth-statediff/indexer/models"
nodeinfo "github.com/cerc-io/plugeth-statediff/indexer/node"
"github.com/cerc-io/plugeth-statediff/indexer/shared/schema"
sdtypes "github.com/cerc-io/plugeth-statediff/types"
2023-06-14 12:43:34 +00:00
)
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) {
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
}