Mercurial > jhg
view src/org/tmatesoft/hg/internal/Patch.java @ 448:2e402c12ebc6 smartgit3
Issue 31: Revlog#walk() fails with AIOOBE when start > 0
author | Artem Tikhomirov <tikhomirov.artem@gmail.com> |
---|---|
date | Wed, 06 Jun 2012 21:23:57 +0200 |
parents | 9747a786a34d |
children | 7f136a3fa671 |
line wrap: on
line source
/* * Copyright (c) 2011 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.io.IOException; import java.util.ArrayList; import java.util.Formatter; /** * @see http://mercurial.selenic.com/wiki/BundleFormat, in Changelog group description * * range [start..end] in original source gets replaced with data of length (do not keep, use data.length instead) * range [end(i)..start(i+1)] is copied from the source * * @author Artem Tikhomirov * @author TMate Software Ltd. */ public final class Patch { private final IntVector starts, ends; private final ArrayList<byte[]> data; private static byte[] generate(int c) { byte[] rv = new byte[c]; for (int i = 0; i < c; i++) { byte x = (byte) ('a' + i); rv[i] = x; } return rv; } public static void main(String[] args) { Patch p1 = new Patch(), p2 = new Patch(); // simple cases (one element in either patch) // III: (1,10 20) & (5,15,15) p2End from [p1End..p1AppliedEnd] (i.e. within p1 range but index is past p2 end index) // II: (1,10,7) & (3,15,15) insideP2 = true and no more p1 entries // II: (1,1,10) & (3,11,15) // independent: (1,10,10) & (15,25,10); (15, 25, 10) & (1, 10, 10) // I: (15, 25, 10) & (10, 20, 10). result: [10, 20, 10] [20, 25, 5] // IV: (15, 25, 10) & (10, 30, 20) // // cycle with insideP2 // // cycle with insideP1 // // multiple elements in patches (offsets) p1.add(15, 25, generate(10)); p2.add(10, 30, generate(20)); System.out.println("p1: " + p1); System.out.println("p2: " + p2); Patch r = p1.apply(p2); System.out.println("r: " + r); } public Patch() { starts = new IntVector(); ends = new IntVector(); data = new ArrayList<byte[]>(); } public String toString() { StringBuilder sb = new StringBuilder(); Formatter f = new Formatter(sb); for (int i = 0; i < count(); i++) { f.format("[%d, %d, %d] ", starts.get(i), ends.get(i), data.get(i).length); } return sb.toString(); } public int count() { return data.size(); } // number of bytes this patch will add (or remove, if negative) from the base revision private int patchSizeDelta() { int rv = 0; int prevEnd = 0; for (int i = 0, x = data.size(); i < x; i++) { final int start = starts.get(i); final int len = data.get(i).length; rv += start - prevEnd; // would copy from original rv += len; // and add new prevEnd = ends.get(i); } rv -= prevEnd; return rv; } public byte[] apply(DataAccess baseRevisionContent, int outcomeLen) throws IOException { if (outcomeLen == -1) { outcomeLen = baseRevisionContent.length() + patchSizeDelta(); } int prevEnd = 0, destIndex = 0; byte[] rv = new byte[outcomeLen]; for (int i = 0, x = data.size(); i < x; i++) { final int start = starts.get(i); baseRevisionContent.seek(prevEnd); // copy source bytes that were not modified (up to start of the record) baseRevisionContent.readBytes(rv, destIndex, start - prevEnd); destIndex += start - prevEnd; // insert new data from the patch, if any byte[] d = data.get(i); System.arraycopy(d, 0, rv, destIndex, d.length); destIndex += d.length; prevEnd = ends.get(i); } baseRevisionContent.seek(prevEnd); // copy everything in the source past last record's end baseRevisionContent.readBytes(rv, destIndex, (int) (baseRevisionContent.length() - prevEnd)); return rv; } public void clear() { starts.clear(); ends.clear(); data.clear(); } /** * Initialize instance from stream. Any previous patch information (i.e. if instance if reused) is cleared first. * Read up to the end of DataAccess and interpret data as patch records. */ public void read(DataAccess da) throws IOException { clear(); while (!da.isEmpty()) { readOne(da); } } /** * Caller is responsible to ensure stream got some data to read */ public void readOne(DataAccess da) throws IOException { int s = da.readInt(); int e = da.readInt(); int len = da.readInt(); byte[] src = new byte[len]; da.readBytes(src, 0, len); starts.add(s); ends.add(e); data.add(src); } private void add(Patch p, int i) { add(p.starts.get(i), p.ends.get(i), p.data.get(i)); } private void add(int start, int end, byte[] d) { starts.add(start); ends.add(end); data.add(d); } private static byte[] subarray(byte[] d, int start, int end) { byte[] r = new byte[end-start+1]; System.arraycopy(d, start, r, 0, r.length); return r; } /** * Modify this patch with subsequent patch */ private /*SHALL BE PUBLIC ONCE TESTING ENDS*/ Patch apply(Patch another) { Patch r = new Patch(); int p1TotalAppliedDelta = 0; // value to add to start and end indexes of the older patch to get their values as if // in the patched text, iow, directly comparable with respective indexes from the newer patch. int p1EntryStart = 0, p1EntryEnd = 0, p1EntryLen = 0; byte[] p1Data = null; boolean insideP1entry = false; int p2 = 0, p1 = 0; final int p2Max = another.count(), p1Max = this.count(); L0: for (; p2 < p2Max; p2++) { int p2EntryStart = another.starts.get(p2); int p2EntryEnd = another.ends.get(p2); final int p2EntryRange = p2EntryEnd - p2EntryStart; final byte[] p2Data = another.data.get(p2); boolean insideP2entry = false; int p2EntryStartOffset = -1; /// p1EntryStart = p1EntryEnd = p1EntryLen = 0; p1Data = null; L1: while (p1 < p1Max) { if (!insideP1entry) { p1EntryStart = starts.get(p1); p1EntryEnd = ends.get(p1); p1Data = data.get(p1); p1EntryLen = p1Data.length; }// else keep values final int p1EntryDelta = p1EntryLen - (p1EntryEnd - p1EntryStart); // number of actually inserted(+) or deleted(-) chars final int p1EntryAppliedStart = p1TotalAppliedDelta + p1EntryStart; final int p1EntryAppliedEnd = p1EntryAppliedStart + p1EntryLen; // end of j'th patch entry in the text which is source for p2 if (insideP2entry) { if (p2EntryEnd < p1EntryAppliedStart) { r.add(p2EntryStart - p2EntryStartOffset, p2EntryEnd - p1TotalAppliedDelta, p2Data); insideP2entry = false; continue L0; } if (p2EntryEnd >= p1EntryAppliedEnd) { // when p2EntryEnd == p1EntryAppliedEnd, I assume p1TotalAppliedDelta can't be used for p2EntryEnd to get it to p1 range, but rather shall be // augmented with current p1 entry and at the next p1 entry (likely to hit p1EntryAppliedStart > p2EntryEnd above) would do the rest insideP1entry = false; p1++; p1TotalAppliedDelta += p1EntryDelta; continue L1; } // p1EntryAppliedStart <= p2EntryEnd < p1EntryAppliedEnd r.add(p2EntryStart - p2EntryStartOffset, p2EntryEnd - p1TotalAppliedDelta, p2Data); p1EntryStart = p2EntryEnd - p1TotalAppliedDelta; final int p1DataPartShift = p2EntryEnd - p1EntryAppliedStart + 1; if (p1DataPartShift >= p1EntryLen) { p1EntryLen = 0; p1Data = new byte[0]; } else { p1EntryLen -= p1DataPartShift; p1Data = subarray(p1Data, p1DataPartShift, p1Data.length); } insideP1entry = true; insideP2entry = false; continue L0; } if (p1EntryAppliedStart < p2EntryStart) { if (p1EntryAppliedEnd <= p2EntryStart) { // p1EntryAppliedEnd in fact index of the first char *after* patch // completely independent, copy and continue r.add(p1EntryStart, p1EntryEnd, p1Data); insideP1entry = false; p1++; // fall-through to get p1TotalAppliedDelta incremented } else { // SKETCH: II or III // remember, p1EntryDelta may be negative // shall break j'th entry into few // fix p1's end/length // p1EntryAppliedStart < p2EntryStart < p1EntryAppliedEnd int s = p2EntryStart - p1TotalAppliedDelta; // p2EntryStart in p1 scale. Is within p1 range if (s > p1EntryEnd) { s = p1EntryEnd; } int p1DataPartEnd = p2EntryStart - p1EntryAppliedStart; // index, not count. <= (p1EntryEnd-p1EntryStart). // add what left from p1 if (p1DataPartEnd < p1EntryLen) { r.add(p1EntryStart, s, subarray(p1Data, 0, p1DataPartEnd)); } else { p1DataPartEnd = p1EntryLen-1; // record factual number of p1 bytes we consumed. r.add(p1EntryStart, s, p1Data); } p1TotalAppliedDelta += p1DataPartEnd - (s - p1EntryStart); // (s2 - (s1+delta)) - (s2 - delta - s1) = s2-s1-delta-s2+delta+s1 = 0, unless p1DataPartEnd >= p1Data.length p1EntryLen -= (p1DataPartEnd+1); if (p2EntryEnd < p1EntryAppliedEnd) { // SKETCH: III insideP1entry = true; // p2 completely fits into changes of p1 int e = p2EntryEnd - p1TotalAppliedDelta; // p2EntryEnd in p1 scale if (e > p1EntryEnd) { // any index past p1 end shall be calculated with respect to p1 end, thus it's unsafe to go past p1 end (there may be more p1 entries there) e = p1EntryEnd; } r.add(s, e, p2Data); // add p2 // modify p1 leftover p1EntryStart = e; if (p2EntryRange >= p1EntryLen) { p1EntryLen = 0; p1Data = new byte[0]; } else { p1Data = subarray(p1Data, p1DataPartEnd + p2EntryRange, p1Data.length-1 /*up to the last one*/); p1EntryLen -= p2EntryRange; } // p2 is handled, but there are leftovers of p1 continue L0; } else { // p2EntryEnd >= p1EntryAppliedEnd // SKETCH: II insideP1entry = false; p1++; if (p1EntryAppliedStart + p1EntryDelta >= p2EntryEnd) { // here we know next p1 entry would be past p2 entry and thus can put p2 right away r.add(p2EntryStart - p1TotalAppliedDelta, p1EntryEnd, p2Data); p1TotalAppliedDelta += p1EntryDelta; continue L0; } else { // there are chances there are more p1 entries till p2 ends insideP2entry = true; p2EntryStartOffset = p1TotalAppliedDelta; // p2EntryEnd is past delta, no chances for p1Data leftovers to be in use // p2 processing is not over, need to fix end, depending on what else fits into p2 range (if nothing, can put p2.end right away) // fall-through to get p1TotalAppliedDelta incremented; } } } } else { // p1EntryAppliedStart >= p2EntryStart if (p2EntryEnd < p1EntryAppliedStart) { // newer patch completely fits between two older patches r.add(p2EntryStart - p1TotalAppliedDelta, p2EntryEnd - p1TotalAppliedDelta, p2Data); // SHALL NOT increment p1TotalAppliedDelta as we didn't use any of p1 continue L0; // next p2 } else { // p2EntryEnd >= p1EntryAppliedStart // SKETCH: I or IV // p2EntryEnd is either < p1EntryAppliedEnd or past it if (p2EntryEnd <= p1EntryAppliedEnd) { // SKETCH: I: copy p2, strip p1 to start from p2EntryEnd, next i (p2) insideP1entry = true; int e = p2EntryEnd - p1TotalAppliedDelta; if (e > p1EntryEnd) { e = p1EntryEnd; // added by analogy with above. Is needed? } r.add(p2EntryStart - p1TotalAppliedDelta, e, p2Data); p1EntryStart = e; int p1DataShift = p2EntryEnd - p1EntryAppliedStart; if (p1DataShift >= p1EntryLen) { p1EntryLen = 0; p1Data = new byte[0]; } else { p1EntryLen -= p1DataShift; p1Data = subarray(p1Data, p1DataShift, p1Data.length - 1); } // p1TotalAppliedDelta would get incremented once this modified p1 is handled continue L0; // next p2; } else { // p2EntryEnd > p1EntryAppliedEnd // SKETCH IV: skip (rest of) p1 completely, continue the same unless found p1 with start or end past p2EntryEnd. insideP1entry = false; p1++; insideP2entry = true; p2EntryStartOffset = p1TotalAppliedDelta; // fall-through to get p1TotalAppliedDelta incremented } } } p1TotalAppliedDelta += p1EntryDelta; } // while (p1 < p1Max) { // no more p1 entries, shall close p2 (if it's handled, code above jumps directly to L0) // regardless of whether insideP2 is .t int s = p2EntryStart; // p2EntryStartOffset != -1 when we started p2 entry processing, but not completed // if we handled last p1 entry but didn't start with p2 entry processing, it's -1 and regular p1 delta shall be used s -= p2EntryStartOffset == -1 ? p1TotalAppliedDelta : p2EntryStartOffset; r.add(s, p2EntryEnd - p1TotalAppliedDelta, p2Data); } } if (p1 < p1Max && insideP1entry) { r.add(p1EntryStart, p1EntryEnd, p1Data); p1++; } while (p1 < p1Max) { r.add(this, p1); p1++; }; return r; } }