// 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 . package postgres import ( "fmt" "github.com/cerc-io/plugeth-statediff/indexer/database/sql" "github.com/cerc-io/plugeth-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 } func (db *DB) ExistsHeaderStm() string { return fmt.Sprintf("SELECT EXISTS(SELECT 1 from %s WHERE block_number = $1 AND block_hash = $2 LIMIT 1)", schema.TableHeader.Name) } // 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"} }