Mercurial > hg4j
view src/org/tmatesoft/hg/repo/HgDataFile.java @ 609:e4a71afd3c71
Test TODOs: test for ConfigFile (covering %include and %unset directives)
author | Artem Tikhomirov <tikhomirov.artem@gmail.com> |
---|---|
date | Wed, 08 May 2013 17:11:45 +0200 |
parents | e1b29756f901 |
children | 868b2ffdcd5c |
line wrap: on
line source
/* * Copyright (c) 2010-2013 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.repo; import static org.tmatesoft.hg.core.HgIterateDirection.OldToNew; import static org.tmatesoft.hg.repo.HgInternals.wrongRevisionIndex; import static org.tmatesoft.hg.repo.HgRepository.*; import static org.tmatesoft.hg.util.LogFacility.Severity.Info; import static org.tmatesoft.hg.util.LogFacility.Severity.Warn; import java.io.File; import java.io.FileInputStream; import java.io.IOException; import java.nio.ByteBuffer; import java.nio.channels.FileChannel; import java.util.Arrays; import org.tmatesoft.hg.core.HgCallbackTargetException; import org.tmatesoft.hg.core.HgChangesetFileSneaker; import org.tmatesoft.hg.core.HgIterateDirection; import org.tmatesoft.hg.core.Nodeid; import org.tmatesoft.hg.internal.BlameHelper; import org.tmatesoft.hg.internal.DataAccess; import org.tmatesoft.hg.internal.FileHistory; import org.tmatesoft.hg.internal.FileRevisionHistoryChunk; import org.tmatesoft.hg.internal.FilterByteChannel; import org.tmatesoft.hg.internal.FilterDataAccess; import org.tmatesoft.hg.internal.Internals; import org.tmatesoft.hg.internal.Metadata; import org.tmatesoft.hg.internal.RevlogStream; import org.tmatesoft.hg.repo.HgBlameInspector; import org.tmatesoft.hg.util.ByteChannel; import org.tmatesoft.hg.util.CancelSupport; import org.tmatesoft.hg.util.CancelledException; import org.tmatesoft.hg.util.LogFacility; import org.tmatesoft.hg.util.Pair; import org.tmatesoft.hg.util.Path; import org.tmatesoft.hg.util.ProgressSupport; /** * Regular user data file stored in the repository. * * <p> Note, most methods accept index in the file's revision history, not that of changelog. Easy way to obtain * changeset revision index from file's is to use {@link #getChangesetRevisionIndex(int)}. To obtain file's revision * index for a given changeset, {@link HgManifest#getFileRevision(int, Path)} or {@link HgChangesetFileSneaker} may * come handy. * * @author Artem Tikhomirov * @author TMate Software Ltd. */ public final class HgDataFile extends Revlog { // absolute from repo root? // slashes, unix-style? // repo location agnostic, just to give info to user, not to access real storage private final Path path; private Metadata metadata; // get initialized on first access to file content. /*package-local*/HgDataFile(HgRepository hgRepo, Path filePath, RevlogStream content) { super(hgRepo, content, false); path = filePath; } /*package-local*/HgDataFile(HgRepository hgRepo, Path filePath) { super(hgRepo); path = filePath; } // exists is not the best name possible. now it means no file with such name was ever known to the repo. // it might be confused with files existed before but lately removed. TODO HgFileNode.exists makes more sense. // or HgDataFile.known() public boolean exists() { return content != null; // XXX need better impl } /** * Human-readable file name, i.e. "COPYING", not "store/data/_c_o_p_y_i_n_g.i" */ public Path getPath() { return path; // hgRepo.backresolve(this) -> name? In this case, what about hashed long names? } /** * Handy shorthand for {@link #getLength(int) length(getRevisionIndex(nodeid))} * * @param nodeid revision of the file * * @return size of the file content at the given revision * @throws HgRuntimeException subclass thereof to indicate issues with the library. <em>Runtime exception</em> */ public int getLength(Nodeid nodeid) throws HgRuntimeException { try { return getLength(getRevisionIndex(nodeid)); } catch (HgInvalidControlFileException ex) { throw ex.isRevisionSet() ? ex : ex.setRevision(nodeid); } catch (HgInvalidRevisionException ex) { throw ex.isRevisionSet() ? ex : ex.setRevision(nodeid); } } /** * @param fileRevisionIndex - revision local index, non-negative. From predefined constants, only {@link HgRepository#TIP} makes sense. * @return size of the file content at the revision identified by local revision number. * @throws HgRuntimeException subclass thereof to indicate issues with the library. <em>Runtime exception</em> */ public int getLength(int fileRevisionIndex) throws HgRuntimeException { if (wrongRevisionIndex(fileRevisionIndex) || fileRevisionIndex == BAD_REVISION) { throw new HgInvalidRevisionException(fileRevisionIndex); } if (fileRevisionIndex == TIP) { fileRevisionIndex = getLastRevision(); } else if (fileRevisionIndex == WORKING_COPY) { File f = getRepo().getFile(this); if (f.exists()) { // single revision can't be greater than 2^32, shall be safe to cast to int return Internals.ltoi(f.length()); } Nodeid fileRev = getWorkingCopyRevision(); if (fileRev == null) { throw new HgInvalidRevisionException(String.format("File %s is not part of working copy", getPath()), null, fileRevisionIndex); } fileRevisionIndex = getRevisionIndex(fileRev); } if (metadata == null || !metadata.checked(fileRevisionIndex)) { checkAndRecordMetadata(fileRevisionIndex); } final int dataLen = content.dataLength(fileRevisionIndex); if (metadata.known(fileRevisionIndex)) { return dataLen - metadata.dataOffset(fileRevisionIndex); } return dataLen; } /** * Reads content of the file from working directory. If file present in the working directory, its actual content without * any filters is supplied through the sink. If file does not exist in the working dir, this method provides content of a file * as if it would be refreshed in the working copy, i.e. its corresponding revision (according to dirstate) is read from the * repository, and filters repo -> working copy get applied. * * NOTE, if file is missing from the working directory and is not part of the dirstate (but otherwise legal repository file, * e.g. from another branch), no content would be supplied. * * @param sink content consumer * @throws CancelledException if execution of the operation was cancelled * @throws HgRuntimeException subclass thereof to indicate issues with the library. <em>Runtime exception</em> */ public void workingCopy(ByteChannel sink) throws CancelledException, HgRuntimeException { File f = getRepo().getFile(this); if (f.exists()) { final CancelSupport cs = CancelSupport.Factory.get(sink); final ProgressSupport progress = ProgressSupport.Factory.get(sink); final long flength = f.length(); final int bsize = (int) Math.min(flength, 32*1024); progress.start((int) (flength > Integer.MAX_VALUE ? flength >>> 15 /*32 kb buf size*/ : flength)); ByteBuffer buf = ByteBuffer.allocate(bsize); FileChannel fc = null; try { fc = new FileInputStream(f).getChannel(); while (fc.read(buf) != -1) { cs.checkCancelled(); buf.flip(); int consumed = sink.write(buf); progress.worked(flength > Integer.MAX_VALUE ? 1 : consumed); buf.compact(); } } catch (IOException ex) { throw new HgInvalidFileException("Working copy read failed", ex, f); } finally { progress.done(); if (fc != null) { try { fc.close(); } catch (IOException ex) { getRepo().getSessionContext().getLog().dump(getClass(), Warn, ex, null); } } } } else { Nodeid fileRev = getWorkingCopyRevision(); if (fileRev == null) { // no content for this data file in the working copy - it is not part of the actual working state. // XXX perhaps, shall report this to caller somehow, not silently pass no data? return; } final int fileRevIndex = getRevisionIndex(fileRev); contentWithFilters(fileRevIndex, sink); } } /** * @return file revision as recorded in repository manifest for dirstate parent, or <code>null</code> if no file revision can be found */ private Nodeid getWorkingCopyRevision() throws HgInvalidControlFileException { final Pair<Nodeid, Nodeid> wcParents = getRepo().getWorkingCopyParents(); Nodeid p = wcParents.first().isNull() ? wcParents.second() : wcParents.first(); final HgChangelog clog = getRepo().getChangelog(); final int csetRevIndex; if (p.isNull()) { // no dirstate parents getRepo().getSessionContext().getLog().dump(getClass(), Info, "No dirstate parents, resort to TIP", getPath()); // if it's a repository with no dirstate, use TIP then csetRevIndex = clog.getLastRevision(); if (csetRevIndex == -1) { // shall not happen provided there's .i for this data file (hence at least one cset) // and perhaps exception is better here. However, null as "can't find" indication seems reasonable. return null; } } else { // common case to avoid searching complete changelog for nodeid match final Nodeid tipRev = clog.getRevision(TIP); if (tipRev.equals(p)) { csetRevIndex = clog.getLastRevision(); } else { // bad luck, need to search honestly csetRevIndex = getRepo().getChangelog().getRevisionIndex(p); } } Nodeid fileRev = getRepo().getManifest().getFileRevision(csetRevIndex, getPath()); // it's possible for a file to be in working dir and have store/.i but to belong e.g. to a different // branch than the one from dirstate. Thus it's possible to get null fileRev // which would serve as an indication this data file is not part of working copy return fileRev; } /** * Access content of a file revision * XXX not sure distinct method contentWithFilters() is the best way to do, perhaps, callers shall add filters themselves? * * @param fileRevisionIndex - revision local index, non-negative. From predefined constants, {@link HgRepository#TIP} and {@link HgRepository#WORKING_COPY} make sense. * @param sink content consumer * * @throws CancelledException if execution of the operation was cancelled * @throws HgRuntimeException subclass thereof to indicate issues with the library. <em>Runtime exception</em> */ public void contentWithFilters(int fileRevisionIndex, ByteChannel sink) throws CancelledException, HgRuntimeException { if (fileRevisionIndex == WORKING_COPY) { workingCopy(sink); // pass un-mangled sink } else { content(fileRevisionIndex, new FilterByteChannel(sink, getRepo().getFiltersFromRepoToWorkingDir(getPath()))); } } /** * Retrieve content of specific revision. Content is provided as is, without any filters (e.g. keywords, eol, etc.) applied. * For filtered content, use {@link #contentWithFilters(int, ByteChannel)}. * * @param fileRevisionIndex - revision local index, non-negative. From predefined constants, {@link HgRepository#TIP} and {@link HgRepository#WORKING_COPY} make sense. * @param sink content consumer * * @throws CancelledException if execution of the operation was cancelled * @throws HgRuntimeException subclass thereof to indicate issues with the library. <em>Runtime exception</em> */ public void content(int fileRevisionIndex, ByteChannel sink) throws CancelledException, HgRuntimeException { // for data files need to check heading of the file content for possible metadata // @see http://mercurial.selenic.com/wiki/FileFormats#data.2BAC8- if (fileRevisionIndex == TIP) { fileRevisionIndex = getLastRevision(); } if (fileRevisionIndex == WORKING_COPY) { // sink is supposed to come into workingCopy without filters // thus we shall not get here (into #content) from #contentWithFilters(WC) workingCopy(sink); return; } if (wrongRevisionIndex(fileRevisionIndex) || fileRevisionIndex == BAD_REVISION) { throw new HgInvalidRevisionException(fileRevisionIndex); } if (sink == null) { throw new IllegalArgumentException(); } if (metadata == null) { metadata = new Metadata(getRepo()); } ErrorHandlingInspector insp; final LogFacility lf = getRepo().getSessionContext().getLog(); if (metadata.none(fileRevisionIndex)) { insp = new ContentPipe(sink, 0, lf); } else if (metadata.known(fileRevisionIndex)) { insp = new ContentPipe(sink, metadata.dataOffset(fileRevisionIndex), lf); } else { // do not know if there's metadata insp = new MetadataInspector(metadata, new ContentPipe(sink, 0, lf)); } insp.checkCancelled(); super.content.iterate(fileRevisionIndex, fileRevisionIndex, true, insp); try { insp.checkFailed(); } catch (HgInvalidControlFileException ex) { ex = ex.setFileName(getPath()); throw ex.isRevisionIndexSet() ? ex : ex.setRevisionIndex(fileRevisionIndex); } catch (IOException ex) { HgInvalidControlFileException e = new HgInvalidControlFileException("Revision content access failed", ex, null); throw content.initWithIndexFile(e).setFileName(getPath()).setRevisionIndex(fileRevisionIndex); } } /** * Walk complete change history of the file. * @param inspector callback to visit changesets * @throws HgRuntimeException subclass thereof to indicate issues with the library. <em>Runtime exception</em> */ public void history(HgChangelog.Inspector inspector) throws HgRuntimeException { history(0, getLastRevision(), inspector); } /** * Walk subset of the file's change history. * @param start revision local index, inclusive; non-negative or {@link HgRepository#TIP} * @param end revision local index, inclusive; non-negative or {@link HgRepository#TIP} * @param inspector callback to visit changesets * @throws HgRuntimeException subclass thereof to indicate issues with the library. <em>Runtime exception</em> */ public void history(int start, int end, HgChangelog.Inspector inspector) throws HgRuntimeException { if (!exists()) { throw new IllegalStateException("Can't get history of invalid repository file node"); } final int last = getLastRevision(); if (end == TIP) { end = last; } if (start == TIP) { start = last; } HgInternals.checkRevlogRange(start, end, last); final int[] commitRevisions = new int[end - start + 1]; final boolean[] needsSorting = { false }; RevlogStream.Inspector insp = new RevlogStream.Inspector() { int count = 0; public void next(int revisionNumber, int actualLen, int baseRevision, int linkRevision, int parent1Revision, int parent2Revision, byte[] nodeid, DataAccess data) { if (count > 0) { if (commitRevisions[count -1] > linkRevision) { needsSorting[0] = true; } } commitRevisions[count++] = linkRevision; } }; content.iterate(start, end, false, insp); final HgChangelog changelog = getRepo().getChangelog(); if (needsSorting[0]) { // automatic tools (svnmerge?) produce unnatural file history // (e.g. cpython/Lib/doctest.py, revision 164 points to cset 63509, 165 - to 38453) Arrays.sort(commitRevisions); } changelog.rangeInternal(inspector, commitRevisions); } /** * For a given revision of the file (identified with revision index), find out index of the corresponding changeset. * * @return changeset revision index * @throws HgRuntimeException subclass thereof to indicate issues with the library. <em>Runtime exception</em> */ public int getChangesetRevisionIndex(int fileRevisionIndex) throws HgRuntimeException { return content.linkRevision(fileRevisionIndex); } /** * Complements {@link #getChangesetRevisionIndex(int)} to get changeset revision that corresponds to supplied file revision * * @param nid revision of the file * @return changeset revision * @throws HgRuntimeException subclass thereof to indicate issues with the library. <em>Runtime exception</em> */ public Nodeid getChangesetRevision(Nodeid nid) throws HgRuntimeException { int changelogRevision = getChangesetRevisionIndex(getRevisionIndex(nid)); return getRepo().getChangelog().getRevision(changelogRevision); } /** * Tells whether this file originates from another repository file * @return <code>true</code> if this file is a copy of another from the repository * @throws HgRuntimeException subclass thereof to indicate issues with the library. <em>Runtime exception</em> */ public boolean isCopy() throws HgRuntimeException { if (metadata == null || !metadata.checked(0)) { checkAndRecordMetadata(0); } if (!metadata.known(0)) { return false; } return metadata.find(0, "copy") != null; } /** * Get name of the file this one was copied from. * * @return name of the file origin * @throws HgRuntimeException subclass thereof to indicate issues with the library. <em>Runtime exception</em> */ public Path getCopySourceName() throws HgRuntimeException { if (isCopy()) { Path.Source ps = getRepo().getSessionContext().getPathFactory(); return ps.path(metadata.find(0, "copy")); } throw new UnsupportedOperationException(); // XXX REVISIT, think over if Exception is good (clients would check isCopy() anyway, perhaps null is sufficient?) } /** * * @return revision this file was copied from * @throws HgRuntimeException subclass thereof to indicate issues with the library. <em>Runtime exception</em> */ public Nodeid getCopySourceRevision() throws HgRuntimeException { if (isCopy()) { return Nodeid.fromAscii(metadata.find(0, "copyrev")); // XXX reuse/cache Nodeid } throw new UnsupportedOperationException(); } /** * Get file flags recorded in the manifest * @param fileRevisionIndex - revision local index, non-negative, or {@link HgRepository#TIP}. * @see HgManifest#getFileFlags(int, Path) * @throws HgRuntimeException subclass thereof to indicate issues with the library. <em>Runtime exception</em> */ public HgManifest.Flags getFlags(int fileRevisionIndex) throws HgRuntimeException { int changesetRevIndex = getChangesetRevisionIndex(fileRevisionIndex); return getRepo().getManifest().getFileFlags(changesetRevIndex, getPath()); } /** * mimic 'hg diff -r clogRevIndex1 -r clogRevIndex2' */ public void diff(int clogRevIndex1, int clogRevIndex2, HgBlameInspector insp) throws HgCallbackTargetException { // FIXME clogRevIndex1 and clogRevIndex2 may point to different files, need to decide whether to throw an exception // or to attempt to look up correct file node (tricky) int fileRevIndex1 = fileRevIndex(this, clogRevIndex1); int fileRevIndex2 = fileRevIndex(this, clogRevIndex2); BlameHelper bh = new BlameHelper(insp, 5); bh.useFileUpTo(this, clogRevIndex2); bh.diff(fileRevIndex1, clogRevIndex1, fileRevIndex2, clogRevIndex2); } /** * Walk file history up/down to revision at given changeset and report changes for each revision */ public void annotate(int changelogRevisionIndex, HgBlameInspector insp, HgIterateDirection iterateOrder) throws HgCallbackTargetException { annotate(0, changelogRevisionIndex, insp, iterateOrder); } /** * Walk file history range and report changes for each revision */ public void annotate(int changelogRevIndexStart, int changelogRevIndexEnd, HgBlameInspector insp, HgIterateDirection iterateOrder) throws HgCallbackTargetException { if (wrongRevisionIndex(changelogRevIndexStart) || wrongRevisionIndex(changelogRevIndexEnd)) { throw new IllegalArgumentException(); } // Note, changelogRevIndexEnd may be TIP, while the code below doesn't tolerate constants // int lastRevision = getRepo().getChangelog().getLastRevision(); if (changelogRevIndexEnd == TIP) { changelogRevIndexEnd = lastRevision; } HgInternals.checkRevlogRange(changelogRevIndexStart, changelogRevIndexEnd, lastRevision); if (!exists()) { return; } FileHistory fileHistory = new FileHistory(this, changelogRevIndexStart, changelogRevIndexEnd); fileHistory.build(); BlameHelper bh = new BlameHelper(insp, 10); for (FileRevisionHistoryChunk fhc : fileHistory.iterate(OldToNew)) { // iteration order is not important here bh.useFileUpTo(fhc.getFile(), fhc.getEndChangeset()); } int[] fileClogParentRevs = new int[2]; int[] fileParentRevs = new int[2]; for (FileRevisionHistoryChunk fhc : fileHistory.iterate(iterateOrder)) { for (int fri : fhc.fileRevisions(iterateOrder)) { int clogRevIndex = fhc.changeset(fri); // the way we built fileHistory ensures we won't walk past [changelogRevIndexStart..changelogRevIndexEnd] assert clogRevIndex >= changelogRevIndexStart; assert clogRevIndex <= changelogRevIndexEnd; fhc.fillFileParents(fri, fileParentRevs); fhc.fillCsetParents(fri, fileClogParentRevs); bh.annotateChange(fri, clogRevIndex, fileParentRevs, fileClogParentRevs); } } } /** * Annotates changes of the file against its parent(s). * Unlike {@link #annotate(HgDataFile, int, Inspector, HgIterateDirection)}, doesn't * walk file history, looks at the specified revision only. Handles both parents (if merge revision). */ public void annotateSingleRevision(int changelogRevisionIndex, HgBlameInspector insp) throws HgCallbackTargetException { // TODO detect if file is text/binary (e.g. looking for chars < ' ' and not \t\r\n\f int fileRevIndex = fileRevIndex(this, changelogRevisionIndex); int[] fileRevParents = new int[2]; parents(fileRevIndex, fileRevParents, null, null); if (changelogRevisionIndex == TIP) { changelogRevisionIndex = getChangesetRevisionIndex(fileRevIndex); } BlameHelper bh = new BlameHelper(insp, 5); bh.useFileUpTo(this, changelogRevisionIndex); int[] fileClogParentRevs = new int[2]; fileClogParentRevs[0] = fileRevParents[0] == NO_REVISION ? NO_REVISION : getChangesetRevisionIndex(fileRevParents[0]); fileClogParentRevs[1] = fileRevParents[1] == NO_REVISION ? NO_REVISION : getChangesetRevisionIndex(fileRevParents[1]); bh.annotateChange(fileRevIndex, changelogRevisionIndex, fileRevParents, fileClogParentRevs); } @Override public String toString() { StringBuilder sb = new StringBuilder(getClass().getSimpleName()); sb.append('('); sb.append(getPath()); sb.append(')'); return sb.toString(); } private void checkAndRecordMetadata(int localRev) throws HgInvalidControlFileException { if (metadata == null) { metadata = new Metadata(getRepo()); } // use MetadataInspector without delegate to process metadata only RevlogStream.Inspector insp = new MetadataInspector(metadata, null); super.content.iterate(localRev, localRev, true, insp); } private static int fileRevIndex(HgDataFile df, int csetRevIndex) { Nodeid fileRev = df.getRepo().getManifest().getFileRevision(csetRevIndex, df.getPath()); return df.getRevisionIndex(fileRev); } private static class MetadataInspector extends ErrorHandlingInspector implements RevlogStream.Inspector { private final Metadata metadata; private final RevlogStream.Inspector delegate; /** * @param _metadata never <code>null</code> * @param chain <code>null</code> if no further data processing other than metadata is desired */ public MetadataInspector(Metadata _metadata, RevlogStream.Inspector chain) { metadata = _metadata; delegate = chain; setCancelSupport(CancelSupport.Factory.get(chain)); } public void next(int revisionNumber, int actualLen, int baseRevision, int linkRevision, int parent1Revision, int parent2Revision, byte[] nodeid, DataAccess data) { try { if (metadata.tryRead(revisionNumber, data)) { // da is in prepared state (i.e. we consumed all bytes up to metadata end). // However, it's not safe to assume delegate won't call da.reset() for some reason, // and we need to ensure predictable result. data.reset(); int offset = metadata.dataOffset(revisionNumber); data = new FilterDataAccess(data, offset, data.length() - offset); } else { data.reset(); } if (delegate != null) { delegate.next(revisionNumber, actualLen, baseRevision, linkRevision, parent1Revision, parent2Revision, nodeid, data); } } catch (IOException ex) { recordFailure(ex); } catch (HgInvalidControlFileException ex) { recordFailure(ex.isRevisionIndexSet() ? ex : ex.setRevisionIndex(revisionNumber)); } } @Override public void checkFailed() throws HgRuntimeException, IOException, CancelledException { super.checkFailed(); if (delegate instanceof ErrorHandlingInspector) { // TODO need to add ErrorDestination (ErrorTarget/Acceptor?) and pass it around (much like CancelSupport get passed) // so that delegate would be able report its failures directly to caller without this hack ((ErrorHandlingInspector) delegate).checkFailed(); } } } }