trie builds with fake database

minigeth/trie/committer.go

+// 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 trie
+
+import (
+	"errors"
+	"fmt"
+	"sync"
+
+	"github.com/ethereum/go-ethereum/common"
+	"github.com/ethereum/go-ethereum/crypto"
+	"golang.org/x/crypto/sha3"
+)
+
+// leafChanSize is the size of the leafCh. It's a pretty arbitrary number, to allow
+// some parallelism but not incur too much memory overhead.
+const leafChanSize = 200
+
+// leaf represents a trie leaf value
+type leaf struct {
+	size int         // size of the rlp data (estimate)
+	hash common.Hash // hash of rlp data
+	node node        // the node to commit
+}
+
+// committer is a type used for the trie Commit operation. A committer has some
+// internal preallocated temp space, and also a callback that is invoked when
+// leaves are committed. The leafs are passed through the `leafCh`,  to allow
+// some level of parallelism.
+// By 'some level' of parallelism, it's still the case that all leaves will be
+// processed sequentially - onleaf will never be called in parallel or out of order.
+type committer struct {
+	tmp sliceBuffer
+	sha crypto.KeccakState
+
+	onleaf LeafCallback
+	leafCh chan *leaf
+}
+
+// committers live in a global sync.Pool
+var committerPool = sync.Pool{
+	New: func() interface{} {
+		return &committer{
+			tmp: make(sliceBuffer, 0, 550), // cap is as large as a full fullNode.
+			sha: sha3.NewLegacyKeccak256().(crypto.KeccakState),
+		}
+	},
+}
+
+// newCommitter creates a new committer or picks one from the pool.
+func newCommitter() *committer {
+	return committerPool.Get().(*committer)
+}
+
+func returnCommitterToPool(h *committer) {
+	h.onleaf = nil
+	h.leafCh = nil
+	committerPool.Put(h)
+}
+
+// commit collapses a node down into a hash node and inserts it into the database
+func (c *committer) Commit(n node, db *Database) (hashNode, error) {
+	if db == nil {
+		return nil, errors.New("no db provided")
+	}
+	h, err := c.commit(n, db)
+	if err != nil {
+		return nil, err
+	}
+	return h.(hashNode), nil
+}
+
+// commit collapses a node down into a hash node and inserts it into the database
+func (c *committer) commit(n node, db *Database) (node, error) {
+	// if this path is clean, use available cached data
+	hash, dirty := n.cache()
+	if hash != nil && !dirty {
+		return hash, nil
+	}
+	// Commit children, then parent, and remove remove the dirty flag.
+	switch cn := n.(type) {
+	case *shortNode:
+		// Commit child
+		collapsed := cn.copy()
+
+		// If the child is fullnode, recursively commit.
+		// Otherwise it can only be hashNode or valueNode.
+		if _, ok := cn.Val.(*fullNode); ok {
+			childV, err := c.commit(cn.Val, db)
+			if err != nil {
+				return nil, err
+			}
+			collapsed.Val = childV
+		}
+		// The key needs to be copied, since we're delivering it to database
+		collapsed.Key = hexToCompact(cn.Key)
+		hashedNode := c.store(collapsed, db)
+		if hn, ok := hashedNode.(hashNode); ok {
+			return hn, nil
+		}
+		return collapsed, nil
+	case *fullNode:
+		hashedKids, err := c.commitChildren(cn, db)
+		if err != nil {
+			return nil, err
+		}
+		collapsed := cn.copy()
+		collapsed.Children = hashedKids
+
+		hashedNode := c.store(collapsed, db)
+		if hn, ok := hashedNode.(hashNode); ok {
+			return hn, nil
+		}
+		return collapsed, nil
+	case hashNode:
+		return cn, nil
+	default:
+		// nil, valuenode shouldn't be committed
+		panic(fmt.Sprintf("%T: invalid node: %v", n, n))
+	}
+}
+
+// commitChildren commits the children of the given fullnode
+func (c *committer) commitChildren(n *fullNode, db *Database) ([17]node, error) {
+	var children [17]node
+	for i := 0; i < 16; i++ {
+		child := n.Children[i]
+		if child == nil {
+			continue
+		}
+		// If it's the hashed child, save the hash value directly.
+		// Note: it's impossible that the child in range [0, 15]
+		// is a valuenode.
+		if hn, ok := child.(hashNode); ok {
+			children[i] = hn
+			continue
+		}
+		// Commit the child recursively and store the "hashed" value.
+		// Note the returned node can be some embedded nodes, so it's
+		// possible the type is not hashnode.
+		hashed, err := c.commit(child, db)
+		if err != nil {
+			return children, err
+		}
+		children[i] = hashed
+	}
+	// For the 17th child, it's possible the type is valuenode.
+	if n.Children[16] != nil {
+		children[16] = n.Children[16]
+	}
+	return children, nil
+}
+
+// store hashes the node n and if we have a storage layer specified, it writes
+// the key/value pair to it and tracks any node->child references as well as any
+// node->external trie references.
+func (c *committer) store(n node, db *Database) node {
+	// Larger nodes are replaced by their hash and stored in the database.
+	var (
+		hash, _ = n.cache()
+		size    int
+	)
+	if hash == nil {
+		// This was not generated - must be a small node stored in the parent.
+		// In theory we should apply the leafCall here if it's not nil(embedded
+		// node usually contains value). But small value(less than 32bytes) is
+		// not our target.
+		return n
+	} else {
+		// We have the hash already, estimate the RLP encoding-size of the node.
+		// The size is used for mem tracking, does not need to be exact
+		size = estimateSize(n)
+	}
+	// If we're using channel-based leaf-reporting, send to channel.
+	// The leaf channel will be active only when there an active leaf-callback
+	if c.leafCh != nil {
+		c.leafCh <- &leaf{
+			size: size,
+			hash: common.BytesToHash(hash),
+			node: n,
+		}
+	} else if db != nil {
+		// No leaf-callback used, but there's still a database. Do serial
+		// insertion
+		db.lock.Lock()
+		db.insert(common.BytesToHash(hash), size, n)
+		db.lock.Unlock()
+	}
+	return hash
+}
+
+// commitLoop does the actual insert + leaf callback for nodes.
+func (c *committer) commitLoop(db *Database) {
+	for item := range c.leafCh {
+		var (
+			hash = item.hash
+			size = item.size
+			n    = item.node
+		)
+		// We are pooling the trie nodes into an intermediate memory cache
+		db.lock.Lock()
+		db.insert(hash, size, n)
+		db.lock.Unlock()
+
+		if c.onleaf != nil {
+			switch n := n.(type) {
+			case *shortNode:
+				if child, ok := n.Val.(valueNode); ok {
+					c.onleaf(nil, nil, child, hash)
+				}
+			case *fullNode:
+				// For children in range [0, 15], it's impossible
+				// to contain valuenode. Only check the 17th child.
+				if n.Children[16] != nil {
+					c.onleaf(nil, nil, n.Children[16].(valueNode), hash)
+				}
+			}
+		}
+	}
+}
+
+func (c *committer) makeHashNode(data []byte) hashNode {
+	n := make(hashNode, c.sha.Size())
+	c.sha.Reset()
+	c.sha.Write(data)
+	c.sha.Read(n)
+	return n
+}
+
+// estimateSize estimates the size of an rlp-encoded node, without actually
+// rlp-encoding it (zero allocs). This method has been experimentally tried, and with a trie
+// with 1000 leafs, the only errors above 1% are on small shortnodes, where this
+// method overestimates by 2 or 3 bytes (e.g. 37 instead of 35)
+func estimateSize(n node) int {
+	switch n := n.(type) {
+	case *shortNode:
+		// A short node contains a compacted key, and a value.
+		return 3 + len(n.Key) + estimateSize(n.Val)
+	case *fullNode:
+		// A full node contains up to 16 hashes (some nils), and a key
+		s := 3
+		for i := 0; i < 16; i++ {
+			if child := n.Children[i]; child != nil {
+				s += estimateSize(child)
+			} else {
+				s++
+			}
+		}
+		return s
+	case valueNode:
+		return 1 + len(n)
+	case hashNode:
+		return 1 + len(n)
+	default:
+		panic(fmt.Sprintf("node type %T", n))
+	}
+}

minigeth/trie/errors.go

+// Copyright 2015 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 trie
+
+import (
+	"fmt"
+
+	"github.com/ethereum/go-ethereum/common"
+)
+
+// MissingNodeError is returned by the trie functions (TryGet, TryUpdate, TryDelete)
+// in the case where a trie node is not present in the local database. It contains
+// information necessary for retrieving the missing node.
+type MissingNodeError struct {
+	NodeHash common.Hash // hash of the missing node
+	Path     []byte      // hex-encoded path to the missing node
+}
+
+func (err *MissingNodeError) Error() string {
+	return fmt.Sprintf("missing trie node %x (path %x)", err.NodeHash, err.Path)
+}

minigeth/trie/extras.go

@@ -2,9 +2,9 @@ package trie
 
 import (
 	"io"
+	"sync"
 
 	"github.com/ethereum/go-ethereum/common"
-	"github.com/ethereum/go-ethereum/crypto"
 )
 
 // rawNode is a simple binary blob used to differentiate between collapsed trie
@@ -20,10 +20,25 @@ func (n rawNode) EncodeRLP(w io.Writer) error {
 	return err
 }
 
-var (
-	// emptyRoot is the known root hash of an empty trie.
-	emptyRoot = common.HexToHash("56e81f171bcc55a6ff8345e692c0f86e5b48e01b996cadc001622fb5e363b421")
+type Database struct {
+	lock sync.RWMutex
+}
 
-	// emptyState is the known hash of an empty state trie entry.
-	emptyState = crypto.Keccak256Hash(nil)
-)
+// Node retrieves an encoded cached trie node from memory. If it cannot be found
+// cached, the method queries the persistent database for the content.
+func (db *Database) Node(hash common.Hash) ([]byte, error) {
+	return []byte{}, nil
+}
+
+// node retrieves a cached trie node from memory, or returns nil if none can be
+// found in the memory cache.
+func (db *Database) node(hash common.Hash) node {
+	return nilValueNode
+}
+
+// insert inserts a collapsed trie node into the memory database.
+// The blob size must be specified to allow proper size tracking.
+// All nodes inserted by this function will be reference tracked
+// and in theory should only used for **trie nodes** insertion.
+func (db *Database) insert(hash common.Hash, size int, node node) {
+}