Mercurial > hg4j
view src/org/tmatesoft/hg/internal/DiffHelper.java @ 576:3c4db86e8c1f
Issue 43: poor performance with InflaterDataAccess. Phase 2: inflate into buffer, effective skip and readByte/readBytes()
author | Artem Tikhomirov <tikhomirov.artem@gmail.com> |
---|---|
date | Tue, 16 Apr 2013 19:31:57 +0200 |
parents | e55f17a7a195 |
children | 47dfa0ec7e35 |
line wrap: on
line source
/* * Copyright (c) 2013 TMate Software Ltd * * 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; version 2 of the License. * * 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. * * For information on how to redistribute this software under * the terms of a license other than GNU General Public License * contact TMate Software at support@hg4j.com */ package org.tmatesoft.hg.internal; import java.util.ArrayList; import java.util.HashMap; import java.util.Map; import org.tmatesoft.hg.repo.HgInvalidStateException; /** * Mercurial cares about changes only up to the line level, e.g. a simple file version dump in manifest looks like (RevlogDump output): * * 522: 233748 0 103 17438 433 522 521 -1 756073cf2321df44d3ed0585f2a5754bc8a1b2f6 * <PATCH>: * 3487..3578, 91:src/org/tmatesoft/hg/core/HgIterateDirection.java\00add61a8a665c5d8f092210767f812fe0d335ac8 * * I.e. for the {fname}{revision} entry format of manifest, not only {revision} is changed, but the whole line, with unchanged {fname} is recorded * in the patch. * * Mercurial paper describes reasons for choosing this approach to delta generation, too. * * * @author Artem Tikhomirov * @author TMate Software Ltd. */ public class DiffHelper<T extends DiffHelper.ChunkSequence<?>> { private Map<Object, IntVector> chunk2UseIndex; private T seq1, seq2; // get filled by #longestMatch, track start of common sequence in seq1 and seq2, respectively private int matchStartS1, matchStartS2; private MatchInspector<T> matchInspector; public void init(T s1, T s2) { seq1 = s1; seq2 = s2; prepare(s2); } public void init(T s1) { if (seq2 == null) { throw new IllegalStateException("Use this #init() only when target sequence shall be matched against different origin"); } seq1 = s1; } private void prepare(T s2) { chunk2UseIndex = new HashMap<Object, IntVector>(); for (int i = 0, len = s2.chunkCount(); i < len; i++) { Object bc = s2.chunk(i); IntVector loc = chunk2UseIndex.get(bc); if (loc == null) { chunk2UseIndex.put(bc, loc = new IntVector()); } loc.add(i); // bc.registerUseIn(i) - BEWARE, use of bc here is incorrect // in this case need to find the only ByteChain to keep indexes // i.e. when there are few equal ByteChain instances, notion of "usedIn" shall be either shared (reference same vector) // or kept within only one of them } } public void findMatchingBlocks(MatchInspector<T> insp) { insp.begin(seq1, seq2); matchInspector = insp; findMatchingBlocks(0, seq1.chunkCount(), 0, seq2.chunkCount()); insp.end(); } /** * implementation based on Python's difflib.py and SequenceMatcher */ public int longestMatch(int startS1, int endS1, int startS2, int endS2) { matchStartS1 = matchStartS2 = 0; int maxLength = 0; IntMap<Integer> chunkIndex2MatchCount = new IntMap<Integer>(8); for (int i = startS1; i < endS1; i++) { Object bc = seq1.chunk(i); IntVector occurencesInS2 = chunk2UseIndex.get(bc); if (occurencesInS2 == null) { chunkIndex2MatchCount.clear(); continue; } IntMap<Integer> newChunkIndex2MatchCount = new IntMap<Integer>(8); for (int j : occurencesInS2.toArray()) { // s1[i] == s2[j] if (j < startS2) { continue; } if (j >= endS2) { break; } int prevChunkMatches = chunkIndex2MatchCount.containsKey(j-1) ? chunkIndex2MatchCount.get(j-1) : 0; int k = prevChunkMatches + 1; newChunkIndex2MatchCount.put(j, k); if (k > maxLength) { matchStartS1 = i-k+1; matchStartS2 = j-k+1; maxLength = k; } } chunkIndex2MatchCount = newChunkIndex2MatchCount; } return maxLength; } private void findMatchingBlocks(int startS1, int endS1, int startS2, int endS2) { int matchLength = longestMatch(startS1, endS1, startS2, endS2); if (matchLength > 0) { final int saveStartS1 = matchStartS1, saveStartS2 = matchStartS2; if (startS1 < matchStartS1 && startS2 < matchStartS2) { findMatchingBlocks(startS1, matchStartS1, startS2, matchStartS2); } matchInspector.match(saveStartS1, saveStartS2, matchLength); if (saveStartS1+matchLength < endS1 && saveStartS2+matchLength < endS2) { findMatchingBlocks(saveStartS1 + matchLength, endS1, saveStartS2 + matchLength, endS2); } } } public interface MatchInspector<T extends ChunkSequence<?>> { void begin(T s1, T s2); void match(int startSeq1, int startSeq2, int matchLength); void end(); } static class MatchDumpInspector<T extends ChunkSequence<?>> implements MatchInspector<T> { private int matchCount; public void begin(T s1, T s2) { matchCount = 0; } public void match(int startSeq1, int startSeq2, int matchLength) { matchCount++; System.out.printf("match #%d: from line #%d and line #%d of length %d\n", matchCount, startSeq1, startSeq2, matchLength); } public void end() { if (matchCount == 0) { System.out.println("NO MATCHES FOUND!"); } } } /** * Matcher implementation that translates "match/equal" notification to a delta-style "added/removed/changed". */ public static class DeltaInspector<T extends ChunkSequence<?>> implements MatchInspector<T> { protected int changeStartS1, changeStartS2; protected T seq1, seq2; public void begin(T s1, T s2) { seq1 = s1; seq2 = s2; changeStartS1 = changeStartS2 = 0; } public void match(int startSeq1, int startSeq2, int matchLength) { reportDeltaElement(startSeq1, startSeq2, matchLength); changeStartS1 = startSeq1 + matchLength; changeStartS2 = startSeq2 + matchLength; } public void end() { if (changeStartS1 < seq1.chunkCount()-1 || changeStartS2 < seq2.chunkCount()-1) { reportDeltaElement(seq1.chunkCount()-1, seq2.chunkCount()-1, 0); } } protected void reportDeltaElement(int matchStartSeq1, int matchStartSeq2, int matchLength) { if (changeStartS1 < matchStartSeq1) { if (changeStartS2 < matchStartSeq2) { changed(changeStartS1, matchStartSeq1, changeStartS2, matchStartSeq2); } else { assert changeStartS2 == matchStartSeq2; deleted(matchStartSeq2, changeStartS1, matchStartSeq1); } } else { assert changeStartS1 == matchStartSeq1; if(changeStartS2 < matchStartSeq2) { added(changeStartS1, changeStartS2, matchStartSeq2); } else { assert changeStartS2 == matchStartSeq2; if (matchStartSeq1 > 0 || matchStartSeq2 > 0) { // FIXME perhaps, exception is too much for the case // once diff is covered with tests, replace with assert false : msg; throw new HgInvalidStateException(String.format("adjustent equal blocks %d, %d and %d,%d", changeStartS1, matchStartSeq1, changeStartS2, matchStartSeq2)); } } } if (matchLength > 0) { unchanged(matchStartSeq1, matchStartSeq2, matchLength); } } /** * [s1From..s1To) replaced with [s2From..s2To) */ protected void changed(int s1From, int s1To, int s2From, int s2To) { // NO-OP } protected void deleted(int s2DeletePoint, int s1From, int s1To) { // NO-OP } protected void added(int s1InsertPoint, int s2From, int s2To) { // NO-OP } protected void unchanged(int s1From, int s2From, int length) { // NO-OP } } static class DeltaDumpInspector<T extends ChunkSequence<?>> extends DeltaInspector<T> { @Override protected void changed(int s1From, int s1To, int s2From, int s2To) { System.out.printf("changed [%d..%d) with [%d..%d)\n", s1From, s1To, s2From, s2To); } @Override protected void deleted(int s2DeletionPoint, int s1From, int s1To) { System.out.printf("deleted [%d..%d)\n", s1From, s1To); } @Override protected void added(int s1InsertPoint, int s2From, int s2To) { System.out.printf("added [%d..%d) at %d\n", s2From, s2To, s1InsertPoint); } @Override protected void unchanged(int s1From, int s2From, int length) { System.out.printf("same [%d..%d) and [%d..%d)\n", s1From, s1From + length, s2From, s2From + length); } } /** * Generic sequence of chunk, where chunk is anything comparable to another chunk, e.g. a string or a single char * Sequence diff algorithm above doesn't care about sequence nature. */ public interface ChunkSequence<T> { public T chunk(int index); public int chunkCount(); } public static final class LineSequence implements ChunkSequence<LineSequence.ByteChain> { private final byte[] input; private ArrayList<ByteChain> lines; public LineSequence(byte[] data) { input = data; } public static LineSequence newlines(byte[] array) { return new LineSequence(array).splitByNewlines(); } // sequence ends with fake, empty line chunk public LineSequence splitByNewlines() { lines = new ArrayList<ByteChain>(); int lastStart = 0; for (int i = 0; i < input.length; i++) { if (input[i] == '\n') { lines.add(new ByteChain(lastStart, i+1)); lastStart = i+1; } else if (input[i] == '\r') { if (i+1 < input.length && input[i+1] == '\n') { i++; } lines.add(new ByteChain(lastStart, i+1)); lastStart = i+1; } } if (lastStart < input.length) { lines.add(new ByteChain(lastStart, input.length)); } // empty chunk to keep offset of input end lines.add(new ByteChain(input.length)); return this; } public ByteChain chunk(int index) { return lines.get(index); } public int chunkCount() { return lines.size(); } public byte[] data(int chunkFrom, int chunkTo) { if (chunkFrom == chunkTo) { return new byte[0]; } int from = chunk(chunkFrom).getOffset(), to = chunk(chunkTo).getOffset(); byte[] rv = new byte[to - from]; System.arraycopy(input, from, rv, 0, rv.length); return rv; } public final class ByteChain { private final int start, end; private final int hash; /** * construct a chunk with a sole purpose to keep * offset of the data end */ ByteChain(int offset) { start = end = offset; // ensure this chunk doesn't match trailing chunk of another sequence hash = System.identityHashCode(this); } ByteChain(int s, int e) { start = s; end = e; hash = calcHash(input, s, e); } /** * byte offset of the this ByteChain inside ChainSequence */ public int getOffset() { return start; } public byte[] data() { byte[] rv = new byte[end - start]; System.arraycopy(input, start, rv, 0, rv.length); return rv; } @Override public boolean equals(Object obj) { if (obj == null || obj.getClass() != ByteChain.class) { return false; } ByteChain other = (ByteChain) obj; if (other.hash != hash || other.end - other.start != end - start) { return false; } return other.match(input, start); } private boolean match(byte[] oi, int from) { for (int i = start, j = from; i < end; i++, j++) { if (LineSequence.this.input[i] != oi[j]) { return false; } } return true; } @Override public int hashCode() { return hash; } @Override public String toString() { return String.format("[@%d\"%s\"]", start, new String(data())); } } // same as Arrays.hashCode(byte[]), just for a slice of a bigger array static int calcHash(byte[] data, int from, int to) { int result = 1; for (int i = from; i < to; i++) { result = 31 * result + data[i]; } return result; } } }