Mercurial > jhg
view src/org/tmatesoft/hg/internal/BlameHelper.java @ 655:bcbcc318f250
Performance: reuse unzip output buffer
author | Artem Tikhomirov <tikhomirov.artem@gmail.com> |
---|---|
date | Thu, 04 Jul 2013 18:36:38 +0200 |
parents | 5f52074707b2 |
children | cce0387c6041 |
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 static org.tmatesoft.hg.core.HgIterateDirection.OldToNew; import static org.tmatesoft.hg.repo.HgRepository.NO_REVISION; import java.util.LinkedList; import java.util.ListIterator; import org.tmatesoft.hg.core.HgCallbackTargetException; import org.tmatesoft.hg.internal.DiffHelper.LineSequence; import org.tmatesoft.hg.internal.DiffHelper.LineSequence.ByteChain; import org.tmatesoft.hg.core.HgBlameInspector; import org.tmatesoft.hg.core.HgBlameInspector.*; import org.tmatesoft.hg.repo.HgDataFile; import org.tmatesoft.hg.repo.HgInvalidStateException; import org.tmatesoft.hg.repo.HgRuntimeException; import org.tmatesoft.hg.util.Adaptable; import org.tmatesoft.hg.util.CancelledException; import org.tmatesoft.hg.util.Pair; /** * Blame implementation * @see HgBlameInspector * @author Artem Tikhomirov * @author TMate Software Ltd. */ public class BlameHelper { private final HgBlameInspector insp; private FileLinesCache linesCache; public BlameHelper(HgBlameInspector inspector) { insp = inspector; } /** * Build history of the file for the specified range (follow renames if necessary). This history * is used to access various file revision data during subsequent {@link #diff(int, int, int, int)} and * {@link #annotateChange(int, int, int[], int[])} calls. Callers can use returned history for own approaches * to iteration over file history. * <p>NOTE, clogRevIndexEnd has to list name of the supplied file in the corresponding manifest, * as it's not possible to trace rename history otherwise. */ public FileHistory prepare(HgDataFile df, int clogRevIndexStart, int clogRevIndexEnd) throws HgRuntimeException { assert clogRevIndexStart <= clogRevIndexEnd; FileHistory fileHistory = new FileHistory(df, clogRevIndexStart, clogRevIndexEnd); fileHistory.build(); int cacheHint = 5; // cache comes useful when we follow merge branches and don't want to // parse base revision twice. There's no easy way to determine max(distance(all(base,merge))), // hence the heuristics to use the longest history chunk: for (FileRevisionHistoryChunk c : fileHistory.iterate(OldToNew)) { // iteration order is not important here if (c.revisionCount() > cacheHint) { cacheHint = c.revisionCount(); } } linesCache = new FileLinesCache(cacheHint); for (FileRevisionHistoryChunk fhc : fileHistory.iterate(OldToNew)) { // iteration order is not important here linesCache.useFileUpTo(fhc.getFile(), fhc.getEndChangeset()); } return fileHistory; } // NO_REVISION is not allowed as any argument public void diff(int fileRevIndex1, int clogRevIndex1, int fileRevIndex2, int clogRevIndex2) throws HgCallbackTargetException, HgRuntimeException { HgDataFile targetFile = linesCache.getFile(clogRevIndex2); LineSequence c1 = linesCache.lines(clogRevIndex1, fileRevIndex1); LineSequence c2 = linesCache.lines(clogRevIndex2, fileRevIndex2); DiffHelper<LineSequence> pg = new DiffHelper<LineSequence>(); pg.init(c1, c2); BlameBlockInspector bbi = new BlameBlockInspector(targetFile, fileRevIndex2, insp, clogRevIndex1, clogRevIndex2); pg.findMatchingBlocks(bbi); bbi.checkErrors(); } public void annotateChange(int fileRevIndex, int csetRevIndex, int[] fileParentRevs, int[] fileParentClogRevs) throws HgCallbackTargetException, HgRuntimeException { HgDataFile targetFile = linesCache.getFile(csetRevIndex); final LineSequence fileRevLines = linesCache.lines(csetRevIndex, fileRevIndex); if (fileParentClogRevs[0] != NO_REVISION && fileParentClogRevs[1] != NO_REVISION) { int p1ClogIndex = fileParentClogRevs[0]; int p2ClogIndex = fileParentClogRevs[1]; LineSequence p1Lines = linesCache.lines(p1ClogIndex, fileParentRevs[0]); LineSequence p2Lines = linesCache.lines(p2ClogIndex, fileParentRevs[1]); DiffHelper<LineSequence> pg = new DiffHelper<LineSequence>(); pg.init(p2Lines, fileRevLines); EqualBlocksCollector p2MergeCommon = new EqualBlocksCollector(); pg.findMatchingBlocks(p2MergeCommon); // pg.init(p1Lines); BlameBlockInspector bbi = new BlameBlockInspector(targetFile, fileRevIndex, insp, p1ClogIndex, csetRevIndex); bbi.setMergeParent2(p2MergeCommon, p2ClogIndex); pg.findMatchingBlocks(bbi); bbi.checkErrors(); } else if (fileParentClogRevs[0] == fileParentClogRevs[1]) { // may be equal iff both are unset assert fileParentClogRevs[0] == NO_REVISION; // everything added BlameBlockInspector bbi = new BlameBlockInspector(targetFile, fileRevIndex, insp, NO_REVISION, csetRevIndex); bbi.begin(LineSequence.newlines(new byte[0]), fileRevLines); bbi.match(0, fileRevLines.chunkCount()-1, 0); bbi.end(); bbi.checkErrors(); } else { int soleParentIndex = fileParentClogRevs[0] == NO_REVISION ? 1 : 0; assert fileParentClogRevs[soleParentIndex] != NO_REVISION; LineSequence parentLines = linesCache.lines(fileParentClogRevs[soleParentIndex], fileParentRevs[soleParentIndex]); DiffHelper<LineSequence> pg = new DiffHelper<LineSequence>(); pg.init(parentLines, fileRevLines); BlameBlockInspector bbi = new BlameBlockInspector(targetFile, fileRevIndex, insp, fileParentClogRevs[soleParentIndex], csetRevIndex); pg.findMatchingBlocks(bbi); bbi.checkErrors(); } } private static class FileLinesCache { private final LinkedList<Pair<Integer, LineSequence>> lruCache; private final int limit; private final LinkedList<Pair<Integer, HgDataFile>> files; // TODO in fact, need sparse array /** * @param lruLimit how many parsed file revisions to keep */ public FileLinesCache(int lruLimit) { limit = lruLimit; lruCache = new LinkedList<Pair<Integer, LineSequence>>(); files = new LinkedList<Pair<Integer,HgDataFile>>(); } public void useFileUpTo(HgDataFile df, int clogRevIndex) { Pair<Integer, HgDataFile> newEntry = new Pair<Integer, HgDataFile>(clogRevIndex, df); for (ListIterator<Pair<Integer, HgDataFile>> it = files.listIterator(); it.hasNext();) { Pair<Integer, HgDataFile> e = it.next(); if (e.first() == clogRevIndex) { assert e.second().getPath().equals(df.getPath()); return; } if (e.first() > clogRevIndex) { // insert new entry before current it.previous(); it.add(newEntry); return; } } files.add(newEntry); } public HgDataFile getFile(int clogRevIndex) { for (Pair<Integer, HgDataFile> e : files) { if (e.first() >= clogRevIndex) { return e.second(); } } throw new HgInvalidStateException(String.format("Got %d file-changelog mappings, but no luck for revision %d.", files.size(), clogRevIndex)); } public LineSequence lines(int clogRevIndex, int fileRevIndex) throws HgRuntimeException { Pair<Integer, LineSequence> cached = checkCache(clogRevIndex); if (cached != null) { return cached.second(); } HgDataFile df = getFile(clogRevIndex); try { ByteArrayChannel c; df.content(fileRevIndex, c = new ByteArrayChannel()); LineSequence rv = LineSequence.newlines(c.toArray()); lruCache.addFirst(new Pair<Integer, LineSequence>(clogRevIndex, rv)); if (lruCache.size() > limit) { lruCache.removeLast(); } return rv; } catch (CancelledException ex) { // TODO likely it was bad idea to throw cancelled exception from content() // deprecate and provide alternative? HgInvalidStateException ise = new HgInvalidStateException("ByteArrayChannel never throws CancelledException"); ise.initCause(ex); throw ise; } } private Pair<Integer,LineSequence> checkCache(int fileRevIndex) { Pair<Integer, LineSequence> rv = null; for (ListIterator<Pair<Integer, LineSequence>> it = lruCache.listIterator(); it.hasNext(); ) { Pair<Integer, LineSequence> p = it.next(); if (p.first() == fileRevIndex) { rv = p; it.remove(); break; } } if (rv != null) { lruCache.addFirst(rv); } return rv; } } private static class BlameBlockInspector extends DiffHelper.DeltaInspector<LineSequence> { private final HgBlameInspector insp; private final int csetOrigin; private final int csetTarget; private EqualBlocksCollector p2MergeCommon; private int csetMergeParent; private IntVector mergeRanges; private final AnnotateRev annotatedRevision; private HgCallbackTargetException error; public BlameBlockInspector(HgDataFile df, int fileRevIndex, HgBlameInspector inspector, int originCset, int targetCset) { assert inspector != null; insp = inspector; annotatedRevision = new AnnotateRev(); annotatedRevision.set(df, fileRevIndex); csetOrigin = originCset; csetTarget = targetCset; } public void setMergeParent2(EqualBlocksCollector p2Merge, int parentCset2) { p2MergeCommon = p2Merge; csetMergeParent = parentCset2; mergeRanges = new IntVector(3*10, 3*10); } @Override public void begin(LineSequence s1, LineSequence s2) { super.begin(s1, s2); if (shallStop()) { return; } ContentBlock originContent = new ContentBlock(s1); ContentBlock targetContent = new ContentBlock(s2); annotatedRevision.set(originContent, targetContent); annotatedRevision.set(csetOrigin, csetTarget, p2MergeCommon != null ? csetMergeParent : NO_REVISION); RevisionDescriptor.Recipient curious = Adaptable.Factory.getAdapter(insp, RevisionDescriptor.Recipient.class, null); if (curious != null) { try { curious.start(annotatedRevision); } catch (HgCallbackTargetException ex) { error = ex; } } } @Override public void end() { super.end(); if (shallStop()) { return; } RevisionDescriptor.Recipient curious = Adaptable.Factory.getAdapter(insp, RevisionDescriptor.Recipient.class, null); if (curious != null) { try { curious.done(annotatedRevision); } catch (HgCallbackTargetException ex) { error = ex; } } p2MergeCommon = null; } @Override protected void changed(int s1From, int s1To, int s2From, int s2To) { if (shallStop()) { return; } try { if (p2MergeCommon != null) { mergeRanges.clear(); p2MergeCommon.combineAndMarkRangesWithTarget(s2From, s2To - s2From, csetOrigin, csetMergeParent, mergeRanges); /* * Usecases, how it USED TO BE initially: * 3 lines changed to 10 lines. range of 10 lines breaks down to 2 from p2, 3 from p1, and 5 from p2. * We report: 2 lines changed to 2(p2), then 1 line changed with 3(p1) and 5 lines added from p2. * * 10 lines changed to 3 lines, range of 3 lines breaks down to 2 line from p1 and 1 line from p2. * We report: 2 lines changed to 2(p1) and 8 lines changed to 1(p2) * * NOW, lines from p2 are always reported as pure add (since we need their insertion point to be in p2, not in p1) * and we try to consume p1 changes as soon as we see first p1's range */ int s1TotalLines = s1To - s1From, s1ConsumedLines = 0, s1Start = s1From; for (int i = 0; i < mergeRanges.size(); i += 3) { final int rangeOrigin = mergeRanges.get(i); final int rangeStart = mergeRanges.get(i+1); final int rangeLen = mergeRanges.get(i+2); final boolean lastRange = i+3 >= mergeRanges.size(); final int s1LinesLeft = s1TotalLines - s1ConsumedLines; // how many lines we may report as changed (don't use more than in range unless it's the very last range) final int s1LinesToBorrow = lastRange ? s1LinesLeft : Math.min(s1LinesLeft, rangeLen); if (rangeOrigin != csetMergeParent && s1LinesToBorrow > 0) { ChangeBlockImpl block = getChangeBlock(s1Start, s1LinesToBorrow, rangeStart, rangeLen); block.setOriginAndTarget(rangeOrigin, csetTarget); insp.changed(block); s1ConsumedLines += s1LinesToBorrow; s1Start += s1LinesToBorrow; } else { int blockInsPoint = rangeOrigin != csetMergeParent ? s1Start : p2MergeCommon.reverseMapLine(rangeStart); ChangeBlockImpl block = getAddBlock(rangeStart, rangeLen, blockInsPoint); block.setOriginAndTarget(rangeOrigin, csetTarget); insp.added(block); } } if (s1ConsumedLines != s1TotalLines) { assert s1ConsumedLines < s1TotalLines : String.format("Expected to process %d lines, but actually was %d", s1TotalLines, s1ConsumedLines); // either there were no ranges from p1, whole s2From..s2To range came from p2, shall report as deleted // or the ranges found were not enough to consume whole s2From..s2To // The "deletion point" is shifted to the end of last csetOrigin->csetTarget change int s2DeletePoint = s2From + s1ConsumedLines; ChangeBlockImpl block = new ChangeBlockImpl(annotatedRevision.origin, null, s1Start, s1To - s1Start, -1, -1, -1, s2DeletePoint); block.setOriginAndTarget(csetOrigin, csetTarget); insp.deleted(block); } } else { ChangeBlockImpl block = getChangeBlock(s1From, s1To - s1From, s2From, s2To - s2From); block.setOriginAndTarget(csetOrigin, csetTarget); insp.changed(block); } } catch (HgCallbackTargetException ex) { error = ex; } } @Override protected void added(int s1InsertPoint, int s2From, int s2To) { if (shallStop()) { return; } try { if (p2MergeCommon != null) { mergeRanges.clear(); p2MergeCommon.combineAndMarkRangesWithTarget(s2From, s2To - s2From, csetOrigin, csetMergeParent, mergeRanges); int insPoint = s1InsertPoint; // track changes to insertion point for (int i = 0; i < mergeRanges.size(); i += 3) { int rangeOrigin = mergeRanges.get(i); int rangeStart = mergeRanges.get(i+1); int rangeLen = mergeRanges.get(i+2); ChangeBlockImpl block = getAddBlock(rangeStart, rangeLen, insPoint); block.setOriginAndTarget(rangeOrigin, csetTarget); insp.added(block); // indicate insPoint moved down number of lines we just reported insPoint += rangeLen; } } else { ChangeBlockImpl block = getAddBlock(s2From, s2To - s2From, s1InsertPoint); block.setOriginAndTarget(csetOrigin, csetTarget); insp.added(block); } } catch (HgCallbackTargetException ex) { error = ex; } } @Override protected void deleted(int s2DeletePoint, int s1From, int s1To) { if (shallStop()) { return; } try { ChangeBlockImpl block = new ChangeBlockImpl(annotatedRevision.origin, null, s1From, s1To - s1From, -1, -1, -1, s2DeletePoint); block.setOriginAndTarget(csetOrigin, csetTarget); insp.deleted(block); } catch (HgCallbackTargetException ex) { error = ex; } } @Override protected void unchanged(int s1From, int s2From, int length) { if (shallStop()) { return; } try { EqualBlockImpl block = new EqualBlockImpl(s1From, s2From, length, annotatedRevision.target); block.setOriginAndTarget(csetOrigin, csetTarget); insp.same(block); } catch (HgCallbackTargetException ex) { error = ex; } } void checkErrors() throws HgCallbackTargetException { if (error != null) { throw error; } } private boolean shallStop() { return error != null; } private ChangeBlockImpl getAddBlock(int start, int len, int insPoint) { return new ChangeBlockImpl(null, annotatedRevision.target, -1, -1, start, len, insPoint, -1); } private ChangeBlockImpl getChangeBlock(int start1, int len1, int start2, int len2) { return new ChangeBlockImpl(annotatedRevision.origin, annotatedRevision.target, start1, len1, start2, len2, start1, start2); } } private static class BlockImpl implements Block { private int originCset; private int targetCset; void setOriginAndTarget(int originChangesetIndex, int targetChangesetIndex) { // XXX perhaps, shall be part of Inspector API, rather than Block's // as they don't change between blocks (although the moment about merged revisions) // is not yet clear to me originCset = originChangesetIndex; targetCset = targetChangesetIndex; } public int originChangesetIndex() { return originCset; } public int targetChangesetIndex() { return targetCset; } } private static class EqualBlockImpl extends BlockImpl implements EqualBlock { private final int start1, start2; private final int length; private final ContentBlock fullContent; private FilterBlock myContent; EqualBlockImpl(int blockStartSeq1, int blockStartSeq2, int blockLength, ContentBlock targetContent) { start1 = blockStartSeq1; start2 = blockStartSeq2; length = blockLength; fullContent = targetContent; } public int originStart() { return start1; } public int targetStart() { return start2; } public int length() { return length; } public BlockData content() { if (myContent == null) { myContent = new FilterBlock(fullContent, start2, length); } return myContent; } @Override public String toString() { return String.format("@@ [%d..%d) == [%d..%d) @@", start1, start1+length, start2, start2+length); } } private static class ChangeBlockImpl extends BlockImpl implements ChangeBlock { private final ContentBlock oldContent; private final ContentBlock newContent; private final int s1Start; private final int s1Len; private final int s2Start; private final int s2Len; private final int s1InsertPoint; private final int s2DeletePoint; private FilterBlock addedBlock, removedBlock; public ChangeBlockImpl(ContentBlock c1, ContentBlock c2, int s1Start, int s1Len, int s2Start, int s2Len, int s1InsertPoint, int s2DeletePoint) { oldContent = c1; newContent = c2; this.s1Start = s1Start; this.s1Len = s1Len; this.s2Start = s2Start; this.s2Len = s2Len; this.s1InsertPoint = s1InsertPoint; this.s2DeletePoint = s2DeletePoint; } public int insertedAt() { return s1InsertPoint; } public int firstAddedLine() { return s2Start; } public int totalAddedLines() { return s2Len; } public BlockData addedLines() { if (addedBlock == null) { addedBlock = new FilterBlock(newContent, firstAddedLine(), totalAddedLines()); } return addedBlock; } public int removedAt() { return s2DeletePoint; } public int firstRemovedLine() { return s1Start; } public int totalRemovedLines() { return s1Len; } public BlockData removedLines() { if (removedBlock == null) { removedBlock = new FilterBlock(oldContent, firstRemovedLine(), totalRemovedLines()); } return removedBlock; } @Override public String toString() { if (s2DeletePoint == -1) { return String.format("@@ -%d,0 +%d,%d @@", insertedAt(), firstAddedLine(), totalAddedLines()); } else if (s1InsertPoint == -1) { // delete only return String.format("@@ -%d,%d +%d,0 @@", firstRemovedLine(), totalRemovedLines(), removedAt()); } return String.format("@@ -%d,%d +%d,%d @@", firstRemovedLine(), totalRemovedLines(), firstAddedLine(), totalAddedLines()); } } private static class SingleLine implements BlockData { private final ByteChain line; public SingleLine(ByteChain lineContent) { line = lineContent; } public BlockData elementAt(int index) { assert false; return null; } public int elementCount() { return 0; } public byte[] asArray() { return line.data(); } } private static class ContentBlock implements BlockData { private final LineSequence seq; public ContentBlock(LineSequence sequence) { seq = sequence; } public BlockData elementAt(int index) { return new SingleLine(seq.chunk(index)); } public int elementCount() { return seq.chunkCount() - 1; } public byte[] asArray() { return seq.data(0, seq.chunkCount() - 1); } } private static class FilterBlock implements BlockData { private final ContentBlock contentBlock; private final int from; private final int length; public FilterBlock(ContentBlock bd, int startFrom, int len) { assert bd != null; assert startFrom + len < bd.seq.chunkCount(); // there's one extra chunk in the end, so strict less is ok contentBlock = bd; from = startFrom; length = len; } public BlockData elementAt(int index) { if (index < 0 || index >= length) { throw new IllegalArgumentException(String.format("Expected value from [0..%d), got %d", length, index)); } return contentBlock.elementAt(from + index); } public int elementCount() { return length; } public byte[] asArray() { return contentBlock.seq.data(from, from + length); } } private static class EqualBlocksCollector implements DiffHelper.MatchInspector<LineSequence> { private final RangeSeq matches = new RangeSeq(); public void begin(LineSequence s1, LineSequence s2) { } public void match(int startSeq1, int startSeq2, int matchLength) { matches.add(startSeq1, startSeq2, matchLength); } public void end() { } public int reverseMapLine(int ln) { return matches.reverseMapLine(ln); } public void intersectWithTarget(int start, int length, IntVector result) { int s = start; for (int l = start, x = start + length; l < x; l++) { if (!matches.includesTargetLine(l)) { if (l - s > 0) { result.add(s); result.add(l - s); } s = l+1; } } if (s < start+length) { result.add(s); result.add((start + length) - s); } } /* * intersects [start..start+length) with ranges of target lines, and based on the intersection * breaks initial range into smaller ranges and records them into result, with marker to indicate * whether the range is from initial range (markerSource) or is a result of the intersection with target * (markerTarget) */ public void combineAndMarkRangesWithTarget(int start, int length, int markerSource, int markerTarget, IntVector result) { int sourceStart = start, targetStart = start, sourceEnd = start + length; for (int l = sourceStart; l < sourceEnd; l++) { if (matches.includesTargetLine(l)) { // l is from target if (sourceStart < l) { // few lines from source range were not in the target, report them result.add(markerSource); result.add(sourceStart); result.add(l - sourceStart); } // indicate the earliest line from source range to use sourceStart = l + 1; } else { // l is not in target if (targetStart < l) { // report lines from target range result.add(markerTarget); result.add(targetStart); result.add(l - targetStart); } // next line *may* be from target targetStart = l + 1; } } // if source range end with line from target, sourceStart would be == sourceEnd, and we need to add range with markerTarget // if source range doesn't end with target line, targetStart == sourceEnd, while sourceStart < sourceEnd if (sourceStart < sourceEnd) { assert targetStart == sourceEnd; // something left from the source range result.add(markerSource); result.add(sourceStart); result.add(sourceEnd - sourceStart); } else if (targetStart < sourceEnd) { assert sourceStart == sourceEnd; result.add(markerTarget); result.add(targetStart); result.add(sourceEnd - targetStart); } } } private static class AnnotateRev implements RevisionDescriptor { public ContentBlock origin, target; public int originCset, targetCset, mergeCset, fileRevIndex; public HgDataFile df; public void set(HgDataFile file, int fileRev) { df = file; fileRevIndex = fileRev; } public void set(ContentBlock o, ContentBlock t) { origin = o; target = t; } public void set(int o, int t, int m) { originCset = o; targetCset = t; mergeCset = m; } public BlockData origin() { return origin; } public BlockData target() { return target; } public int originChangesetIndex() { return originCset; } public int targetChangesetIndex() { return targetCset; } public boolean isMerge() { return mergeCset != NO_REVISION; } public int mergeChangesetIndex() { return mergeCset; } public int fileRevisionIndex() { return fileRevIndex; } public HgDataFile file() { return df; } @Override public String toString() { if (isMerge()) { return String.format("[%d,%d->%d]", originCset, mergeCset, targetCset); } return String.format("[%d->%d]", originCset, targetCset); } } public static void main(String[] args) { EqualBlocksCollector bc = new EqualBlocksCollector(); bc.match(-1, 5, 3); bc.match(-1, 10, 2); bc.match(-1, 15, 3); bc.match(-1, 20, 3); IntVector r = new IntVector(); bc.intersectWithTarget(7, 10, r); for (int i = 0; i < r.size(); i+=2) { System.out.printf("[%d..%d) ", r.get(i), r.get(i) + r.get(i+1)); } System.out.println(); r.clear(); bc.combineAndMarkRangesWithTarget(0, 16, 508, 514, r); for (int i = 0; i < r.size(); i+=3) { System.out.printf("%d:[%d..%d) ", r.get(i), r.get(i+1), r.get(i+1) + r.get(i+2)); } } }