hntrie.js (28193B) - View raw
/******************************************************************************* uBlock Origin - a comprehensive, efficient content blocker Copyright (C) 2017-present Raymond Hill This program is free software: you can redistribute it and/or modify it under the terms of the GNU 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 General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see {http://www.gnu.org/licenses/}. Home: https://github.com/gorhill/uBlock */ /******************************************************************************* The original prototype was to develop an idea I had about using jump indices in a TypedArray for quickly matching hostnames (or more generally strings)[1]. Once I had a working, un-optimized prototype, I realized I had ended up with something formally named a "trie": <https://en.wikipedia.org/wiki/Trie>, hence the name. I have no idea whether the implementation here or one resembling it has been done elsewhere. "HN" in HNTrieContainer stands for "HostName", because the trie is specialized to deal with matching hostnames -- which is a bit more complicated than matching plain strings. For example, `www.abc.com` is deemed matching `abc.com`, because the former is a subdomain of the latter. The opposite is of course not true. The resulting read-only tries created as a result of using HNTrieContainer are simply just typed arrays filled with integers. The matching algorithm is just a matter of reading/comparing these integers, and further using them as indices in the array as a way to move around in the trie. [1] To solve <https://github.com/gorhill/uBlock/issues/3193> Since this trie is specialized for matching hostnames, the stored strings are reversed internally, because of hostname comparison logic: Correct matching: index 0123456 abc.com | www.abc.com index 01234567890 Incorrect matching (typically used for plain strings): index 0123456 abc.com | www.abc.com index 01234567890 ------------------------------------------------------------------------------ 1st iteration: - https://github.com/gorhill/uBlock/blob/ff58107dac3a32607f8113e39ed5015584506813/src/js/hntrie.js - Suitable for small to medium set of hostnames - One buffer per trie 2nd iteration: goal was to make matches() method wasm-able - https://github.com/gorhill/uBlock/blob/c3b0fd31f64bd7ffecdd282fb1208fe07aac3eb0/src/js/hntrie.js - Suitable for small to medium set of hostnames - Distinct tries all share same buffer: - Reduced memory footprint - https://stackoverflow.com/questions/45803829/memory-overhead-of-typed-arrays-vs-strings/45808835#45808835 - Reusing needle character lookups for all tries - This significantly reduce the number of String.charCodeAt() calls - Slightly improved creation time This is the 3rd iteration: goal was to make add() method wasm-able and further improve memory/CPU efficiency. This 3rd iteration has the following new traits: - Suitable for small to large set of hostnames - Support multiple trie containers (instanciable) - Designed to hold large number of hostnames - Hostnames can be added at any time (instead of all at once) - This means pre-sorting is no longer a requirement - The trie is always compact - There is no longer a need for a `vacuum` method - This makes the add() method wasm-able - It can return the exact hostname which caused the match - serializable/unserializable available for fast loading - Distinct trie reference support the iteration protocol, thus allowing to extract all the hostnames in the trie Its primary purpose is to replace the use of Set() as a mean to hold large number of hostnames (ex. FilterHostnameDict in static filtering engine). A HNTrieContainer is mostly a large buffer in which distinct but related tries are stored. The memory layout of the buffer is as follow: 0-254: needle being processed 255: length of needle 256-259: offset to start of trie data section (=> trie0) 260-263: offset to end of trie data section (=> trie1) 264-267: offset to start of character data section (=> char0) 268-271: offset to end of character data section (=> char1) 272: start of trie data section */ const PAGE_SIZE = 65536; // i32 / i8 const TRIE0_SLOT = 256 >>> 2; // 64 / 256 const TRIE1_SLOT = TRIE0_SLOT + 1; // 65 / 260 const CHAR0_SLOT = TRIE0_SLOT + 2; // 66 / 264 const CHAR1_SLOT = TRIE0_SLOT + 3; // 67 / 268 const TRIE0_START = TRIE0_SLOT + 4 << 2; // 272 const roundToPageSize = v => (v + PAGE_SIZE-1) & ~(PAGE_SIZE-1); // http://www.cse.yorku.ca/~oz/hash.html#djb2 const i32Checksum = (buf32) => { const n = buf32.length; let hash = 177573 ^ n; for ( let i = 0; i < n; i++ ) { hash = (hash << 5) + hash ^ buf32[i]; } return hash; }; class HNTrieContainer { constructor() { const len = PAGE_SIZE * 2; this.buf = new Uint8Array(len); this.buf32 = new Uint32Array(this.buf.buffer); this.needle = ''; this.buf32[TRIE0_SLOT] = TRIE0_START; this.buf32[TRIE1_SLOT] = this.buf32[TRIE0_SLOT]; this.buf32[CHAR0_SLOT] = len >>> 1; this.buf32[CHAR1_SLOT] = this.buf32[CHAR0_SLOT]; this.wasmMemory = null; this.lastStored = ''; this.lastStoredLen = this.lastStoredIndex = 0; } //-------------------------------------------------------------------------- // Public methods //-------------------------------------------------------------------------- reset(details) { if ( details instanceof Object && typeof details.byteLength === 'number' && typeof details.char0 === 'number' ) { if ( details.byteLength > this.buf.byteLength ) { this.reallocateBuf(details.byteLength); } this.buf32[CHAR0_SLOT] = details.char0; } this.buf32[TRIE1_SLOT] = this.buf32[TRIE0_SLOT]; this.buf32[CHAR1_SLOT] = this.buf32[CHAR0_SLOT]; this.lastStored = ''; this.lastStoredLen = this.lastStoredIndex = 0; } setNeedle(needle) { if ( needle !== this.needle ) { const buf = this.buf; let i = needle.length; if ( i > 255 ) { i = 255; } buf[255] = i; while ( i-- ) { buf[i] = needle.charCodeAt(i); } this.needle = needle; } return this; } matchesJS(iroot) { const buf32 = this.buf32; const buf8 = this.buf; const char0 = buf32[CHAR0_SLOT]; let ineedle = buf8[255]; let icell = buf32[iroot+0]; if ( icell === 0 ) { return -1; } let c = 0, v = 0, i0 = 0, n = 0; for (;;) { if ( ineedle === 0 ) { return -1; } ineedle -= 1; c = buf8[ineedle]; // find first segment with a first-character match for (;;) { v = buf32[icell+2]; i0 = char0 + (v >>> 8); if ( buf8[i0] === c ) { break; } icell = buf32[icell+0]; if ( icell === 0 ) { return -1; } } // all characters in segment must match n = v & 0x7F; if ( n > 1 ) { n -= 1; if ( n > ineedle ) { return -1; } i0 += 1; const i1 = i0 + n; do { ineedle -= 1; if ( buf8[i0] !== buf8[ineedle] ) { return -1; } i0 += 1; } while ( i0 < i1 ); } // boundary at end of segment? if ( (v & 0x80) !== 0 ) { if ( ineedle === 0 || buf8[ineedle-1] === 0x2E /* '.' */ ) { return ineedle; } } // next segment icell = buf32[icell+1]; if ( icell === 0 ) { break; } } return -1; } createTrie() { // grow buffer if needed if ( (this.buf32[CHAR0_SLOT] - this.buf32[TRIE1_SLOT]) < 12 ) { this.growBuf(12, 0); } const iroot = this.buf32[TRIE1_SLOT] >>> 2; this.buf32[TRIE1_SLOT] += 12; this.buf32[iroot+0] = 0; this.buf32[iroot+1] = 0; this.buf32[iroot+2] = 0; return iroot; } createTrieFromIterable(hostnames) { const itrie = this.createTrie(); for ( const hn of hostnames ) { if ( hn === '' ) { continue; } this.setNeedle(hn).add(itrie); } return itrie; } createTrieFromStoredDomainOpt(i, n) { const itrie = this.createTrie(); const jend = i + n; let j = i, offset = 0, k = 0, c = 0; while ( j !== jend ) { offset = this.buf32[CHAR0_SLOT]; // Important k = 0; for (;;) { if ( j === jend ) { break; } c = this.buf[offset+j]; j += 1; if ( c === 0x7C /* '|' */ ) { break; } if ( k === 255 ) { continue; } this.buf[k] = c; k += 1; } if ( k !== 0 ) { this.buf[255] = k; this.add(itrie); } } this.needle = ''; // Important this.buf[255] = 0; // Important return itrie; } dumpTrie(iroot) { let hostnames = Array.from(this.trieIterator(iroot)); if ( String.prototype.padStart instanceof Function ) { const maxlen = Math.min( hostnames.reduce((maxlen, hn) => Math.max(maxlen, hn.length), 0), 64 ); hostnames = hostnames.map(hn => hn.padStart(maxlen)); } for ( const hn of hostnames ) { console.log(hn); } } trieIterator(iroot) { return { value: undefined, done: false, next() { if ( this.icell === 0 ) { if ( this.forks.length === 0 ) { this.value = undefined; this.done = true; return this; } this.charPtr = this.forks.pop(); this.icell = this.forks.pop(); } for (;;) { const idown = this.container.buf32[this.icell+0]; if ( idown !== 0 ) { this.forks.push(idown, this.charPtr); } const v = this.container.buf32[this.icell+2]; let i0 = this.container.buf32[CHAR0_SLOT] + (v >>> 8); const i1 = i0 + (v & 0x7F); while ( i0 < i1 ) { this.charPtr -= 1; this.charBuf[this.charPtr] = this.container.buf[i0]; i0 += 1; } this.icell = this.container.buf32[this.icell+1]; if ( (v & 0x80) !== 0 ) { return this.toHostname(); } } }, toHostname() { this.value = this.textDecoder.decode( new Uint8Array(this.charBuf.buffer, this.charPtr) ); return this; }, container: this, icell: this.buf32[iroot], charBuf: new Uint8Array(256), charPtr: 256, forks: [], textDecoder: new TextDecoder(), [Symbol.iterator]() { return this; }, }; } // TODO: // Rework code to add from a string already present in the character // buffer, i.e. not having to go through setNeedle() when adding a new // hostname to a trie. This will require much work though, and probably // changing the order in which string segments are stored in the // character buffer. addJS(iroot) { let lhnchar = this.buf[255]; if ( lhnchar === 0 ) { return 0; } // grow buffer if needed if ( (this.buf32[CHAR0_SLOT] - this.buf32[TRIE1_SLOT]) < 24 || (this.buf.length - this.buf32[CHAR1_SLOT]) < 256 ) { this.growBuf(24, 256); } let icell = this.buf32[iroot+0]; // special case: first node in trie if ( icell === 0 ) { this.buf32[iroot+0] = this.addLeafCell(lhnchar); return 1; } // const char0 = this.buf32[CHAR0_SLOT]; let isegchar, lsegchar, boundaryBit, inext; // find a matching cell: move down for (;;) { const v = this.buf32[icell+2]; let isegchar0 = char0 + (v >>> 8); // if first character is no match, move to next descendant if ( this.buf[isegchar0] !== this.buf[lhnchar-1] ) { inext = this.buf32[icell+0]; if ( inext === 0 ) { this.buf32[icell+0] = this.addLeafCell(lhnchar); return 1; } icell = inext; continue; } // 1st character was tested isegchar = 1; lhnchar -= 1; // find 1st mismatch in rest of segment lsegchar = v & 0x7F; if ( lsegchar !== 1 ) { for (;;) { if ( isegchar === lsegchar ) { break; } if ( lhnchar === 0 ) { break; } if ( this.buf[isegchar0+isegchar] !== this.buf[lhnchar-1] ) { break; } isegchar += 1; lhnchar -= 1; } } boundaryBit = v & 0x80; // all segment characters matched if ( isegchar === lsegchar ) { // needle remainder: no if ( lhnchar === 0 ) { // boundary: yes, already present if ( boundaryBit !== 0 ) { return 0; } // boundary: no, mark as boundary this.buf32[icell+2] = v | 0x80; } // needle remainder: yes else { // remainder is at label boundary? if yes, no need to add // the rest since the shortest match is always reported if ( boundaryBit !== 0 ) { if ( this.buf[lhnchar-1] === 0x2E /* '.' */ ) { return -1; } } inext = this.buf32[icell+1]; if ( inext !== 0 ) { icell = inext; continue; } // add needle remainder this.buf32[icell+1] = this.addLeafCell(lhnchar); } } // some segment characters matched else { // split current cell isegchar0 -= char0; this.buf32[icell+2] = isegchar0 << 8 | isegchar; inext = this.addCell( 0, this.buf32[icell+1], isegchar0 + isegchar << 8 | boundaryBit | lsegchar - isegchar ); this.buf32[icell+1] = inext; // needle remainder: yes, need new cell for remaining characters if ( lhnchar !== 0 ) { this.buf32[inext+0] = this.addLeafCell(lhnchar); } // needle remainder: no, need boundary cell else { this.buf32[icell+2] |= 0x80; } } return 1; } } optimize() { this.shrinkBuf(); return { byteLength: this.buf.byteLength, char0: this.buf32[CHAR0_SLOT], }; } toSelfie() { const buf32 = this.buf32.subarray(0, this.buf32[CHAR1_SLOT] + 3 >>> 2); return { buf32, checksum: i32Checksum(buf32) }; } fromSelfie(selfie) { if ( typeof selfie !== 'object' || selfie === null ) { return false; } if ( selfie.buf32 instanceof Uint32Array === false ) { return false; } if ( selfie.checksum !== i32Checksum(selfie.buf32) ) { return false; } this.needle = ''; let byteLength = selfie.buf32.length << 2; if ( byteLength === 0 ) { return false; } byteLength = roundToPageSize(byteLength); if ( this.wasmMemory !== null ) { const pageCountBefore = this.buf.length >>> 16; const pageCountAfter = byteLength >>> 16; if ( pageCountAfter > pageCountBefore ) { this.wasmMemory.grow(pageCountAfter - pageCountBefore); this.buf = new Uint8Array(this.wasmMemory.buffer); this.buf32 = new Uint32Array(this.buf.buffer); } this.buf32.set(selfie.buf32); } else { this.buf32 = selfie.buf32; this.buf = new Uint8Array(this.buf32.buffer); } // https://github.com/uBlockOrigin/uBlock-issues/issues/2925 this.buf[255] = 0; return true; } // The following *Hostname() methods can be used to store hostname strings // outside the trie. This is useful to store/match hostnames which are // not part of a collection, and yet still benefit from storing the strings // into a trie container's character buffer. // TODO: WASM version of matchesHostname() storeHostname(hn) { let n = hn.length; if ( n > 255 ) { hn = hn.slice(-255); n = 255; } if ( n === this.lastStoredLen && hn === this.lastStored ) { return this.lastStoredIndex; } this.lastStored = hn; this.lastStoredLen = n; if ( (this.buf.length - this.buf32[CHAR1_SLOT]) < n ) { this.growBuf(0, n); } const offset = this.buf32[CHAR1_SLOT]; this.buf32[CHAR1_SLOT] = offset + n; const buf8 = this.buf; for ( let i = 0; i < n; i++ ) { buf8[offset+i] = hn.charCodeAt(i); } return (this.lastStoredIndex = offset - this.buf32[CHAR0_SLOT]); } extractHostname(i, n) { const textDecoder = new TextDecoder(); const offset = this.buf32[CHAR0_SLOT] + i; return textDecoder.decode(this.buf.subarray(offset, offset + n)); } storeDomainOpt(s) { let n = s.length; if ( n === this.lastStoredLen && s === this.lastStored ) { return this.lastStoredIndex; } this.lastStored = s; this.lastStoredLen = n; if ( (this.buf.length - this.buf32[CHAR1_SLOT]) < n ) { this.growBuf(0, n); } const offset = this.buf32[CHAR1_SLOT]; this.buf32[CHAR1_SLOT] = offset + n; const buf8 = this.buf; for ( let i = 0; i < n; i++ ) { buf8[offset+i] = s.charCodeAt(i); } return (this.lastStoredIndex = offset - this.buf32[CHAR0_SLOT]); } extractDomainOpt(i, n) { const textDecoder = new TextDecoder(); const offset = this.buf32[CHAR0_SLOT] + i; return textDecoder.decode(this.buf.subarray(offset, offset + n)); } matchesHostname(hn, i, n) { this.setNeedle(hn); const buf8 = this.buf; const hr = buf8[255]; if ( n > hr ) { return false; } const hl = hr - n; const nl = this.buf32[CHAR0_SLOT] + i; for ( let j = 0; j < n; j++ ) { if ( buf8[nl+j] !== buf8[hl+j] ) { return false; } } return n === hr || hn.charCodeAt(hl-1) === 0x2E /* '.' */; } async enableWASM(wasmModuleFetcher, path) { if ( typeof WebAssembly === 'undefined' ) { return false; } if ( this.wasmMemory instanceof WebAssembly.Memory ) { return true; } const module = await getWasmModule(wasmModuleFetcher, path); if ( module instanceof WebAssembly.Module === false ) { return false; } const memory = new WebAssembly.Memory({ initial: 2 }); const instance = await WebAssembly.instantiate(module, { imports: { memory, growBuf: this.growBuf.bind(this, 24, 256) } }); if ( instance instanceof WebAssembly.Instance === false ) { return false; } this.wasmMemory = memory; const curPageCount = memory.buffer.byteLength >>> 16; const newPageCount = roundToPageSize(this.buf.byteLength) >>> 16; if ( newPageCount > curPageCount ) { memory.grow(newPageCount - curPageCount); } const buf = new Uint8Array(memory.buffer); buf.set(this.buf); this.buf = buf; this.buf32 = new Uint32Array(this.buf.buffer); this.matches = this.matchesWASM = instance.exports.matches; this.add = this.addWASM = instance.exports.add; return true; } dumpInfo() { return [ `Buffer size (Uint8Array): ${this.buf32[CHAR1_SLOT].toLocaleString('en')}`, ` Char segment size: ${(this.buf32[CHAR1_SLOT] - this.buf32[CHAR0_SLOT]).toLocaleString('en')}`, `WASM: ${this.wasmMemory === null ? 'disabled' : 'enabled'}`, ].join('\n'); } //-------------------------------------------------------------------------- // Private methods //-------------------------------------------------------------------------- addCell(idown, iright, v) { let icell = this.buf32[TRIE1_SLOT]; this.buf32[TRIE1_SLOT] = icell + 12; icell >>>= 2; this.buf32[icell+0] = idown; this.buf32[icell+1] = iright; this.buf32[icell+2] = v; return icell; } addLeafCell(lsegchar) { const r = this.buf32[TRIE1_SLOT] >>> 2; let i = r; while ( lsegchar > 127 ) { this.buf32[i+0] = 0; this.buf32[i+1] = i + 3; this.buf32[i+2] = this.addSegment(lsegchar, lsegchar - 127); lsegchar -= 127; i += 3; } this.buf32[i+0] = 0; this.buf32[i+1] = 0; this.buf32[i+2] = this.addSegment(lsegchar, 0) | 0x80; this.buf32[TRIE1_SLOT] = i + 3 << 2; return r; } addSegment(lsegchar, lsegend) { if ( lsegchar === 0 ) { return 0; } let char1 = this.buf32[CHAR1_SLOT]; const isegchar = char1 - this.buf32[CHAR0_SLOT]; let i = lsegchar; do { this.buf[char1++] = this.buf[--i]; } while ( i !== lsegend ); this.buf32[CHAR1_SLOT] = char1; return isegchar << 8 | lsegchar - lsegend; } growBuf(trieGrow, charGrow) { const char0 = Math.max( roundToPageSize(this.buf32[TRIE1_SLOT] + trieGrow), this.buf32[CHAR0_SLOT] ); const char1 = char0 + this.buf32[CHAR1_SLOT] - this.buf32[CHAR0_SLOT]; const bufLen = Math.max( roundToPageSize(char1 + charGrow), this.buf.length ); this.resizeBuf(bufLen, char0); } shrinkBuf() { // Can't shrink WebAssembly.Memory if ( this.wasmMemory !== null ) { return; } const char0 = this.buf32[TRIE1_SLOT] + 24; const char1 = char0 + this.buf32[CHAR1_SLOT] - this.buf32[CHAR0_SLOT]; const bufLen = char1 + 256; this.resizeBuf(bufLen, char0); } resizeBuf(bufLen, char0) { bufLen = roundToPageSize(bufLen); if ( bufLen === this.buf.length && char0 === this.buf32[CHAR0_SLOT] ) { return; } const charDataLen = this.buf32[CHAR1_SLOT] - this.buf32[CHAR0_SLOT]; if ( this.wasmMemory !== null ) { const pageCount = (bufLen >>> 16) - (this.buf.byteLength >>> 16); if ( pageCount > 0 ) { this.wasmMemory.grow(pageCount); this.buf = new Uint8Array(this.wasmMemory.buffer); this.buf32 = new Uint32Array(this.wasmMemory.buffer); } } else if ( bufLen !== this.buf.length ) { const newBuf = new Uint8Array(bufLen); newBuf.set( new Uint8Array( this.buf.buffer, 0, this.buf32[TRIE1_SLOT] ), 0 ); newBuf.set( new Uint8Array( this.buf.buffer, this.buf32[CHAR0_SLOT], charDataLen ), char0 ); this.buf = newBuf; this.buf32 = new Uint32Array(this.buf.buffer); this.buf32[CHAR0_SLOT] = char0; this.buf32[CHAR1_SLOT] = char0 + charDataLen; } if ( char0 !== this.buf32[CHAR0_SLOT] ) { this.buf.set( new Uint8Array( this.buf.buffer, this.buf32[CHAR0_SLOT], charDataLen ), char0 ); this.buf32[CHAR0_SLOT] = char0; this.buf32[CHAR1_SLOT] = char0 + charDataLen; } } reallocateBuf(newSize) { newSize = roundToPageSize(newSize); if ( newSize === this.buf.length ) { return; } if ( this.wasmMemory === null ) { const newBuf = new Uint8Array(newSize); newBuf.set( newBuf.length < this.buf.length ? this.buf.subarray(0, newBuf.length) : this.buf ); this.buf = newBuf; } else { const growBy = ((newSize + 0xFFFF) >>> 16) - (this.buf.length >>> 16); if ( growBy <= 0 ) { return; } this.wasmMemory.grow(growBy); this.buf = new Uint8Array(this.wasmMemory.buffer); } this.buf32 = new Uint32Array(this.buf.buffer); } } HNTrieContainer.prototype.matches = HNTrieContainer.prototype.matchesJS; HNTrieContainer.prototype.matchesWASM = null; HNTrieContainer.prototype.add = HNTrieContainer.prototype.addJS; HNTrieContainer.prototype.addWASM = null; /******************************************************************************/ // Code below is to attempt to load a WASM module which implements: // // - HNTrieContainer.add() // - HNTrieContainer.matches() // // The WASM module is entirely optional, the JS implementations will be // used should the WASM module be unavailable for whatever reason. const getWasmModule = (( ) => { let wasmModulePromise; return async function(wasmModuleFetcher, path) { if ( wasmModulePromise instanceof Promise ) { return wasmModulePromise; } // The wasm module will work only if CPU is natively little-endian, // as we use native uint32 array in our js code. const uint32s = new Uint32Array(1); const uint8s = new Uint8Array(uint32s.buffer); uint32s[0] = 1; if ( uint8s[0] !== 1 ) { return; } wasmModulePromise = wasmModuleFetcher(`${path}hntrie`).catch(reason => { console.info(reason); }); return wasmModulePromise; }; })(); /******************************************************************************/ export default HNTrieContainer;