triedb/pathdb: introduce flatStateCodec abstraction

Introduce flatStateCodec, a small interface that captures the
trie-specific aspects of flat-state storage: key derivation from
(address, slot), persistence of account/storage entries, clean-cache
key disambiguation, iterator setup, and progress-marker handling.

Mirrors the existing nodeHasher pattern and complements the Hasher
interface from state-hasher-iface-2 (which abstracts trie-side hashing
and commit). The codec is stored on Database alongside the existing
hasher field, ready to be threaded through the flat-state call sites
(disklayer, flush, generator, reader) in the next commit.

Provides merkleFlatCodec, a thin wrapper over the existing rawdb
snapshot accessors and helpers. This is a pure refactor: behavior is
unchanged. The bintrie-side codec implementation is added in a later
commit, after all call sites have been routed through the abstraction.
This commit is contained in:
CPerezz 2026-04-07 14:17:16 +02:00
parent 2851f7b8c7
commit eaf5523a5a
No known key found for this signature in database
GPG key ID: 62045F34B97177DD
2 changed files with 231 additions and 9 deletions

View file

@ -125,10 +125,11 @@ type Database struct {
// readOnly is the flag whether the mutation is allowed to be applied.
// It will be set automatically when the database is journaled during
// the shutdown to reject all following unexpected mutations.
readOnly bool // Flag if database is opened in read only mode
waitSync bool // Flag if database is deactivated due to initial state sync
isVerkle bool // Flag if database is used for verkle tree
hasher nodeHasher // Trie node hasher
readOnly bool // Flag if database is opened in read only mode
waitSync bool // Flag if database is deactivated due to initial state sync
isVerkle bool // Flag if database is used for verkle tree
hasher nodeHasher // Trie node hasher
flatCodec flatStateCodec // Flat-state key derivation, persistence and iteration
config *Config // Configuration for database
diskdb ethdb.Database // Persistent storage for matured trie nodes
@ -153,11 +154,12 @@ func New(diskdb ethdb.Database, config *Config, isVerkle bool) *Database {
config = config.sanitize()
db := &Database{
readOnly: config.ReadOnly,
isVerkle: isVerkle,
config: config,
diskdb: diskdb,
hasher: merkleNodeHasher,
readOnly: config.ReadOnly,
isVerkle: isVerkle,
config: config,
diskdb: diskdb,
hasher: merkleNodeHasher,
flatCodec: &merkleFlatCodec{},
}
// Establish a dedicated database namespace tailored for verkle-specific
// data, ensuring the isolation of both verkle and merkle tree data. It's
@ -167,6 +169,10 @@ func New(diskdb ethdb.Database, config *Config, isVerkle bool) *Database {
if isVerkle {
db.diskdb = rawdb.NewTable(diskdb, string(rawdb.VerklePrefix))
db.hasher = binaryNodeHasher
// NOTE: bintrieFlatCodec is introduced in a later commit. Until then,
// verkle databases also use the merkle codec for backward compatibility
// (the existing snapshot path is disabled for verkle anyway via the
// noBuild guard at setStateGenerator).
}
// Construct the layer tree by resolving the in-disk singleton state
// and in-memory layer journal.

216
triedb/pathdb/flat_codec.go Normal file
View file

@ -0,0 +1,216 @@
// Copyright 2026 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 pathdb
import (
"bytes"
"github.com/ethereum/go-ethereum/common"
"github.com/ethereum/go-ethereum/core/rawdb"
"github.com/ethereum/go-ethereum/crypto"
"github.com/ethereum/go-ethereum/ethdb"
)
// flatStateCodec abstracts the trie-specific aspects of flat-state storage:
// key derivation from (address, slot), persistence of account/storage entries
// to disk, clean-cache key disambiguation, and iterator construction.
//
// It mirrors the existing nodeHasher pattern (a hot, small interface plugged
// into the Database struct), and complements the Hasher interface from
// state-hasher-iface-2 which abstracts trie-side hashing/commit.
//
// Two implementations are provided:
// - merkleFlatCodec: keccak-keyed flat state, the historical MPT scheme.
// - bintrieFlatCodec: per-stem flat state for the unified binary trie. Added
// in a later commit. Until then, only merkleFlatCodec is wired.
//
// All methods MUST be safe for concurrent use; the codec is shared across
// goroutines (the disk layer's read path, the buffer flush path, and the
// background generator may all call into it simultaneously).
type flatStateCodec interface {
// AccountKey derives the flat-state lookup key for an account.
//
// For Merkle: returns keccak256(addr).
// For Bintrie: returns the stem of the BasicData leaf (a 31-byte stem
// zero-padded into a common.Hash). Subsequent reads of the stem blob
// extract the BasicData and CodeHash leaves at offsets 0 and 1.
AccountKey(addr common.Address) common.Hash
// StorageKey derives the flat-state lookup keys for a storage slot.
//
// The first return value carries the account-side hash (e.g.
// keccak256(addr) for Merkle, or zero for bintrie which has no per-account
// grouping). The second return value carries the slot-side hash
// (keccak256(slot) for Merkle, or the full bintrie key for bintrie).
//
// Read/Write methods receive the same pair, so the codec implementation
// is the only place that has to interpret them.
StorageKey(addr common.Address, slot common.Hash) (accountKey common.Hash, storageKey common.Hash)
// ReadAccount loads an account flat-state entry from persistent storage.
// Returns nil if the entry is not present.
ReadAccount(db ethdb.KeyValueReader, key common.Hash) []byte
// ReadStorage loads a storage flat-state entry from persistent storage.
// Returns nil if the entry is not present.
ReadStorage(db ethdb.KeyValueReader, accountKey common.Hash, storageKey common.Hash) []byte
// WriteAccount persists an account flat-state entry into the supplied batch.
WriteAccount(batch ethdb.Batch, key common.Hash, blob []byte)
// DeleteAccount removes an account flat-state entry via the supplied batch.
DeleteAccount(batch ethdb.Batch, key common.Hash)
// WriteStorage persists a storage flat-state entry into the supplied batch.
WriteStorage(batch ethdb.Batch, accountKey common.Hash, storageKey common.Hash, blob []byte)
// DeleteStorage removes a storage flat-state entry via the supplied batch.
DeleteStorage(batch ethdb.Batch, accountKey common.Hash, storageKey common.Hash)
// AccountCacheKey returns the byte key used in the disk-layer clean state
// cache (fastcache) for an account entry. The cache is shared between
// account and storage lookups, so codecs must ensure their key spaces are
// disjoint to avoid collisions.
AccountCacheKey(key common.Hash) []byte
// StorageCacheKey returns the byte key used in the disk-layer clean state
// cache (fastcache) for a storage entry. See AccountCacheKey for the
// disjointness requirement.
StorageCacheKey(accountKey common.Hash, storageKey common.Hash) []byte
// AccountPrefix returns the rawdb key prefix used by account entries on
// disk. Used by the generator to set up its account-range iterator.
AccountPrefix() []byte
// StoragePrefix returns the rawdb key prefix used by storage entries on
// disk. Used by the generator to set up its storage-range iterator.
StoragePrefix() []byte
// AccountKeyLength returns the expected total length (prefix + payload)
// of an on-disk account key. The generator uses this to filter spurious
// matches when iterating with a length-bounded iterator.
AccountKeyLength() int
// StorageKeyLength returns the expected total length (prefix + payload)
// of an on-disk storage key. See AccountKeyLength.
StorageKeyLength() int
// AccountPrefixSize returns the per-entry on-disk overhead used by the
// stateSet to estimate flush sizes. This is just the prefix length for
// merkle codecs; bintrie codecs may use a different convention.
AccountPrefixSize() int
// StoragePrefixSize returns the per-entry on-disk overhead for storage
// entries.
StoragePrefixSize() int
// SplitMarker decomposes a generation progress marker into the account
// portion and the full marker. For Merkle the account part is the first
// 32 bytes; for bintrie both halves are the same single 32-byte stem.
SplitMarker(marker []byte) (accountMarker []byte, fullMarker []byte)
// MarkerCompare compares a flat-state key against a generation progress
// marker. Returns the same semantics as bytes.Compare. Used by the
// disklayer.account/storage gating logic and by writeStates.
MarkerCompare(key []byte, marker []byte) int
}
// merkleFlatCodec implements flatStateCodec for the keccak-keyed MPT flat
// state scheme. All methods are thin wrappers over rawdb accessors and
// existing helpers; this codec preserves the historical behavior bit-for-bit.
type merkleFlatCodec struct{}
// Compile-time interface check.
var _ flatStateCodec = (*merkleFlatCodec)(nil)
func (c *merkleFlatCodec) AccountKey(addr common.Address) common.Hash {
return crypto.Keccak256Hash(addr.Bytes())
}
func (c *merkleFlatCodec) StorageKey(addr common.Address, slot common.Hash) (common.Hash, common.Hash) {
return crypto.Keccak256Hash(addr.Bytes()), crypto.Keccak256Hash(slot.Bytes())
}
func (c *merkleFlatCodec) ReadAccount(db ethdb.KeyValueReader, key common.Hash) []byte {
return rawdb.ReadAccountSnapshot(db, key)
}
func (c *merkleFlatCodec) ReadStorage(db ethdb.KeyValueReader, accountKey, storageKey common.Hash) []byte {
return rawdb.ReadStorageSnapshot(db, accountKey, storageKey)
}
func (c *merkleFlatCodec) WriteAccount(batch ethdb.Batch, key common.Hash, blob []byte) {
rawdb.WriteAccountSnapshot(batch, key, blob)
}
func (c *merkleFlatCodec) DeleteAccount(batch ethdb.Batch, key common.Hash) {
rawdb.DeleteAccountSnapshot(batch, key)
}
func (c *merkleFlatCodec) WriteStorage(batch ethdb.Batch, accountKey, storageKey common.Hash, blob []byte) {
rawdb.WriteStorageSnapshot(batch, accountKey, storageKey, blob)
}
func (c *merkleFlatCodec) DeleteStorage(batch ethdb.Batch, accountKey, storageKey common.Hash) {
rawdb.DeleteStorageSnapshot(batch, accountKey, storageKey)
}
func (c *merkleFlatCodec) AccountCacheKey(key common.Hash) []byte {
// The historical merkle clean cache uses the bare 32-byte account hash.
// This is a slice into the caller's hash; callers must not retain it.
return key[:]
}
func (c *merkleFlatCodec) StorageCacheKey(accountKey, storageKey common.Hash) []byte {
return storageKeySlice(accountKey, storageKey)
}
func (c *merkleFlatCodec) AccountPrefix() []byte {
return rawdb.SnapshotAccountPrefix
}
func (c *merkleFlatCodec) StoragePrefix() []byte {
return rawdb.SnapshotStoragePrefix
}
func (c *merkleFlatCodec) AccountKeyLength() int {
return len(rawdb.SnapshotAccountPrefix) + common.HashLength
}
func (c *merkleFlatCodec) StorageKeyLength() int {
return len(rawdb.SnapshotStoragePrefix) + 2*common.HashLength
}
func (c *merkleFlatCodec) AccountPrefixSize() int {
return len(rawdb.SnapshotAccountPrefix)
}
func (c *merkleFlatCodec) StoragePrefixSize() int {
return len(rawdb.SnapshotStoragePrefix)
}
func (c *merkleFlatCodec) SplitMarker(marker []byte) ([]byte, []byte) {
var accMarker []byte
if len(marker) > 0 {
accMarker = marker[:common.HashLength]
}
return accMarker, marker
}
func (c *merkleFlatCodec) MarkerCompare(key []byte, marker []byte) int {
return bytes.Compare(key, marker)
}