Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Implement trie.put #3473

Merged
merged 14 commits into from
Jun 28, 2024
7 changes: 7 additions & 0 deletions packages/verkle/src/node/util.ts
Original file line number Diff line number Diff line change
Expand Up @@ -57,3 +57,10 @@ export const createCValues = (values: Uint8Array[], deletedValues = new Array(12
}
return expandedValues
}
export function isLeafNode(node: VerkleNode): node is LeafNode {
return node.type === VerkleNodeType.Leaf
}

export function isInternalNode(node: VerkleNode): node is InternalNode {
return node.type === VerkleNodeType.Internal
}
21 changes: 2 additions & 19 deletions packages/verkle/src/util/bytes.ts
Original file line number Diff line number Diff line change
Expand Up @@ -10,30 +10,13 @@ export function matchingBytesLength(bytes1: Uint8Array, bytes2: Uint8Array): num
let count = 0
const minLength = Math.min(bytes1.length, bytes2.length)

// Unroll the loop for better performance
for (let i = 0; i < minLength - 3; i += 4) {
// Compare 4 bytes at a time
if (
bytes1[i] === bytes2[i] &&
bytes1[i + 1] === bytes2[i + 1] &&
bytes1[i + 2] === bytes2[i + 2] &&
bytes1[i + 3] === bytes2[i + 3]
) {
count += 4
} else {
// Break early if a mismatch is found
break
}
}

// Handle any remaining elements
for (let i = minLength - (minLength % 4); i < minLength; i++) {
for (let i = 0; i < minLength; i++) {
if (bytes1[i] === bytes2[i]) {
count++
} else {
// Break early if a mismatch is found
break
}
}

return count
}
190 changes: 175 additions & 15 deletions packages/verkle/src/verkleTree.ts
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@ import {
ValueEncoding,
bytesToHex,
equalsBytes,
intToHex,
zeros,
} from '@ethereumjs/util'
import debug from 'debug'
Expand All @@ -14,7 +15,7 @@ import { CheckpointDB } from './db/checkpoint.js'
import { InternalNode } from './node/internalNode.js'
import { LeafNode } from './node/leafNode.js'
import { type VerkleNode } from './node/types.js'
import { decodeNode } from './node/util.js'
import { decodeNode, isLeafNode } from './node/util.js'
import {
type Proof,
ROOT_DB_KEY,
Expand Down Expand Up @@ -133,7 +134,9 @@ export class VerkleTree {
}
}

return new VerkleTree(opts)
const trie = new VerkleTree(opts)
await trie._createRootNode()
return trie
}

database(db?: DB<Uint8Array, Uint8Array>) {
Expand Down Expand Up @@ -194,7 +197,6 @@ export class VerkleTree {
const suffix = key[key.length - 1]
this.DEBUG && this.debug(`Stem: ${bytesToHex(stem)}; Suffix: ${suffix}`, ['GET'])
const res = await this.findPath(stem)

if (res.node instanceof LeafNode) {
// The retrieved leaf node contains an array of 256 possible values.
// The index of the value we want is at the key's last byte
Expand All @@ -213,11 +215,169 @@ export class VerkleTree {
* @param value - the value to store
* @returns A Promise that resolves once value is stored.
*/
// TODO: Rewrite following logic in verkle.spec.ts "findPath validation" test
async put(_key: Uint8Array, _value: Uint8Array): Promise<void> {
throw new Error('not implemented')
async put(key: Uint8Array, value: Uint8Array): Promise<void> {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This function is very long. I find functions more difficult to read and debug when I have to scroll in order to see different parts of it.

I appreciate the way you use the comments to outline how things are happening step by step. I wonder if these different steps could be externalized into separate functions, so that the put function itself can fit on a screen?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is a 100% valid criticism. Lemme look at the logic and see if there's a way to break it down. I think you're right that it can be done in smaller chunks.

if (key.length !== 32) throw new Error(`expected key with length 32; got ${key.length}`)
const stem = key.slice(0, 31)
const suffix = key[key.length - 1]
this.DEBUG && this.debug(`Stem: ${bytesToHex(stem)}; Suffix: ${suffix}`, ['PUT'])

const putStack: [Uint8Array, VerkleNode][] = []
// Find path to nearest node
const foundPath = await this.findPath(stem)

// Sanity check - we should at least get the root node back
if (foundPath.stack.length === 0) {
throw new Error(`Root node not found in trie`)
}

// Step 1) Create or update the leaf node
let leafNode: LeafNode
// First see if leaf node already exists
if (foundPath.node !== null) {
// Sanity check to verify we have the right node type
if (!isLeafNode(foundPath.node)) {
throw new Error(
`expected leaf node found at ${bytesToHex(stem)}. Got internal node instead`
)
}
leafNode = foundPath.node
// Sanity check to verify we have the right leaf node
if (!equalsBytes(leafNode.stem, stem)) {
throw new Error(
`invalid leaf node found. Expected stem: ${bytesToHex(stem)}; got ${bytesToHex(
foundPath.node.stem
)}`
)
}
} else {
// Leaf node doesn't exist, create a new one
leafNode = await LeafNode.create(
stem,
new Array(256).fill(new Uint8Array(32)),
this.verkleCrypto
)
this.DEBUG && this.debug(`Creating new leaf node at stem: ${bytesToHex(stem)}`, ['PUT'])
}
// Update value in leaf node and push to putStack
leafNode.setValue(suffix, value)
this.DEBUG &&
this.debug(
`Updating value for suffix: ${suffix} at leaf node with stem: ${bytesToHex(stem)}`,
['PUT']
)
putStack.push([leafNode.hash(), leafNode])

// `path` is the path to the last node pushed to the `putStack`
let lastPath = leafNode.stem

// Step 2) Determine if a new internal node is needed
if (foundPath.stack.length > 1) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just to make sure I am getting this correctly, could you confirm that this scenario is correct:

  • Let's say we start with a root node
  • We put a key/value pair where the key starts with 0x01.
  • findPath returns the rootNode.
  • We create a leafNode, and add a pointer to the newly created leaf node at index 2 of the root node.
  • We put another key/value pair that also starts with 0x01.
  • findPath now returns the leafNode that was created in the previous put operation, along with the remaining values where the paths branch off.
  • We then create an internal node where the paths branch off, commit to the two children to it, and then commit the internal node to the root node (at index 2).
  • rootHash gets updated.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That should be how it works. Not guaranteeing I got it exactly right but the tests seem to indicate so

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Cool, I didn't infer the above from how the function/tests were structured, but wrote it based on my understanding of verkle trees, good to hear it matches

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looking at your comment again, let me confirm something. What you call "index 2" here would be rootNode.children[1] right? These are zero indexed arrays and so inserting a node that starts with 0x01 means you check rootNode.children[1] to see if there's an existing reference to some child node and if not, create a new leaf node with the value for the key that starts with 0x01 and then put the commitment from that new leaf node in rootNode.children[1]

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah yes that's what I meant, sorry for the confusing indexing

// Only insert new internal node if we have more than 1 node in the path
// since a single node indicates only the root node is in the path
const nearestNodeTuple = foundPath.stack.pop()!
const nearestNode = nearestNodeTuple[0]
lastPath = nearestNodeTuple[1]
const updatedParentTuple = this.updateParent(leafNode, nearestNode, lastPath)
putStack.push([updatedParentTuple.node.hash(), updatedParentTuple.node])
lastPath = updatedParentTuple.lastPath

// Step 3) Walk up trie and update child references in parent internal nodes
while (foundPath.stack.length > 1) {
const [nextNode, nextPath] = foundPath.stack.pop()! as [InternalNode, Uint8Array]
// Compute the child index to be updated on `nextNode`
const childIndex = lastPath[matchingBytesLength(lastPath, nextPath)]
// Update child reference
nextNode.setChild(childIndex, {
commitment: putStack[putStack.length - 1][1].commitment,
path: lastPath,
})
this.DEBUG &&
this.debug(
`Updating child reference for node with path: ${bytesToHex(
lastPath
)} at index ${childIndex} in internal node at path ${bytesToHex(nextPath)}`,
['PUT']
)
// Hold onto `path` to current node for updating next parent node child index
lastPath = nextPath
putStack.push([nextNode.hash(), nextNode])
}
}

// Step 4) Update root node
const rootNode = foundPath.stack.pop()![0] as InternalNode
rootNode.setChild(stem[0], {
commitment: putStack[putStack.length - 1][1].commitment,
path: lastPath,
})
this.root(this.verkleCrypto.serializeCommitment(rootNode.commitment))
this.DEBUG &&
this.debug(
`Updating child reference for node with path: ${bytesToHex(lastPath)} at index ${
lastPath[0]
} in root node`,
['PUT']
)
this.DEBUG && this.debug(`Updating root node hash to ${bytesToHex(this._root)}`, ['PUT'])
putStack.push([this._root, rootNode])
await this.saveStack(putStack)
}

/**
* Helper method for updating or creating the parent internal node for a given leaf node
* @param leafNode the child leaf node that will be referenced by the new/updated internal node
* returned by this method
* @param nearestNode the nearest node to the new leaf node
* @param pathToNode the path to `nearestNode`
* @returns a tuple of the updated parent node and the path to that parent (i.e. the partial stem of the leaf node that leads to the parent)
*/
updateParent(
leafNode: LeafNode,
nearestNode: VerkleNode,
pathToNode: Uint8Array
): { node: InternalNode; lastPath: Uint8Array } {
// Compute the portion of leafNode.stem and nearestNode.path that match (i.e. the partial path closest to leafNode.stem)
const partialMatchingStemIndex = matchingBytesLength(leafNode.stem, pathToNode)
let internalNode: InternalNode
if (isLeafNode(nearestNode)) {
// We need to create a new internal node and set nearestNode and leafNode as child nodes of it
// Create new internal node
internalNode = InternalNode.create(this.verkleCrypto)
// Set leafNode and nextNode as children of the new internal node
internalNode.setChild(leafNode.stem[partialMatchingStemIndex], {
commitment: leafNode.commitment,
path: leafNode.stem,
})
internalNode.setChild(nearestNode.stem[partialMatchingStemIndex], {
commitment: nearestNode.commitment,
path: nearestNode.stem,
})
// Find the path to the new internal node (the matching portion of the leaf node and next node's stems)
pathToNode = leafNode.stem.slice(0, partialMatchingStemIndex)
this.DEBUG &&
this.debug(`Creating new internal node at path ${bytesToHex(pathToNode)}`, ['PUT'])
} else {
// Nearest node is an internal node. We need to update the appropriate child reference
// to the new leaf node
internalNode = nearestNode
internalNode.setChild(leafNode.stem[partialMatchingStemIndex], {
commitment: leafNode.commitment,
path: leafNode.stem,
})
this.DEBUG &&
this.debug(
`Updating child reference for leaf node with stem: ${bytesToHex(
leafNode.stem
)} at index ${
leafNode.stem[partialMatchingStemIndex]
} in internal node at path ${bytesToHex(
leafNode.stem.slice(0, partialMatchingStemIndex)
)}`,
['PUT']
)
}
return { node: internalNode, lastPath: pathToNode }
}
/**
* Tries to find a path to the node for the given key.
* It returns a `stack` of nodes to the closest node.
Expand All @@ -231,12 +391,13 @@ export class VerkleTree {
stack: [],
remaining: key,
}
if (equalsBytes(this.root(), this.EMPTY_TREE_ROOT)) return result

// TODO: Decide if findPath should return an empty stack if we have an empty trie or a path with just the empty root node
// if (equalsBytes(this.root(), this.EMPTY_TREE_ROOT)) return result

// Get root node
let rawNode = await this._db.get(this.root())
if (rawNode === undefined)
throw new Error('root node should exist when root not empty tree root')
if (rawNode === undefined) throw new Error('root node should exist')

const rootNode = decodeNode(rawNode, this.verkleCrypto) as InternalNode

Expand All @@ -246,7 +407,7 @@ export class VerkleTree {

// Root node doesn't contain a child node's commitment on the first byte of the path so we're done
if (equalsBytes(child.commitment, this.verkleCrypto.zeroCommitment)) {
this.DEBUG && this.debug(`Partial Path ${key[0]} - found no child.`, ['FIND_PATH'])
this.DEBUG && this.debug(`Partial Path ${intToHex(key[0])} - found no child.`, ['FIND_PATH'])
return result
}
let finished = false
Expand All @@ -260,7 +421,7 @@ export class VerkleTree {
// Calculate the index of the last matching byte in the key
const matchingKeyLength = matchingBytesLength(key, child.path)
const foundNode = equalsBytes(key, child.path)
if (foundNode || child.path.length >= key.length || decodedNode instanceof LeafNode) {
if (foundNode || child.path.length >= key.length || isLeafNode(decodedNode)) {
// If the key and child.path are equal, then we found the node
// If the child.path is the same length or longer than the key but doesn't match it
// or the found node is a leaf node, we've found another node where this node should
Expand All @@ -282,16 +443,15 @@ export class VerkleTree {
// We found a different node than the one specified by `key`
// so the sought node doesn't exist
result.remaining = key.slice(matchingKeyLength)
const pathToNearestNode = isLeafNode(decodedNode) ? decodedNode.stem : child.path
this.DEBUG &&
this.debug(
`Path ${bytesToHex(
key.slice(0, matchingKeyLength)
)} - found path to nearest node ${bytesToHex(
`Path ${bytesToHex(pathToNearestNode)} - found path to nearest node ${bytesToHex(
decodedNode.hash()
)} but target node not found.`,
['FIND_PATH']
)
result.stack.push([decodedNode, key.slice(0, matchingKeyLength)])
result.stack.push([decodedNode, pathToNearestNode])
return result
}
// Push internal node to path stack
Expand Down
3 changes: 2 additions & 1 deletion packages/verkle/test/internalNode.spec.ts
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@ import { type VerkleCrypto, equalsBytes, randomBytes } from '@ethereumjs/util'
import { loadVerkleCrypto } from 'verkle-cryptography-wasm'
import { assert, beforeAll, describe, it } from 'vitest'

import { NODE_WIDTH, VerkleNodeType, decodeNode } from '../src/node/index.js'
import { NODE_WIDTH, VerkleNodeType, decodeNode, isInternalNode } from '../src/node/index.js'
import { InternalNode } from '../src/node/internalNode.js'

describe('verkle node - internal', () => {
Expand All @@ -14,6 +14,7 @@ describe('verkle node - internal', () => {
const commitment = randomBytes(32)
const node = new InternalNode({ commitment, verkleCrypto })

assert.ok(isInternalNode(node), 'typeguard should return true')
assert.equal(node.type, VerkleNodeType.Internal, 'type should be set')
assert.ok(equalsBytes(node.commitment, commitment), 'commitment should be set')

Expand Down
3 changes: 2 additions & 1 deletion packages/verkle/test/leafNode.spec.ts
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@ import { type VerkleCrypto, equalsBytes, randomBytes } from '@ethereumjs/util'
import { loadVerkleCrypto } from 'verkle-cryptography-wasm'
import { assert, beforeAll, describe, it } from 'vitest'

import { VerkleNodeType } from '../src/node/index.js'
import { VerkleNodeType, isLeafNode } from '../src/node/index.js'
import { LeafNode } from '../src/node/leafNode.js'

describe('verkle node - leaf', () => {
Expand All @@ -25,6 +25,7 @@ describe('verkle node - leaf', () => {
verkleCrypto,
})

assert.ok(isLeafNode(node), 'typeguard should return true')
assert.equal(node.type, VerkleNodeType.Leaf, 'type should be set')
assert.ok(
equalsBytes(node.commitment as unknown as Uint8Array, commitment),
Expand Down
Loading
Loading