tikhomirov@13: /* tikhomirov@537: * Copyright (c) 2010-2013 TMate Software Ltd tikhomirov@74: * tikhomirov@74: * This program is free software; you can redistribute it and/or modify tikhomirov@74: * it under the terms of the GNU General Public License as published by tikhomirov@74: * the Free Software Foundation; version 2 of the License. tikhomirov@74: * tikhomirov@74: * This program is distributed in the hope that it will be useful, tikhomirov@74: * but WITHOUT ANY WARRANTY; without even the implied warranty of tikhomirov@74: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the tikhomirov@74: * GNU General Public License for more details. tikhomirov@74: * tikhomirov@74: * For information on how to redistribute this software under tikhomirov@74: * the terms of a license other than GNU General Public License tikhomirov@102: * contact TMate Software at support@hg4j.com tikhomirov@2: */ tikhomirov@74: package org.tmatesoft.hg.repo; tikhomirov@74: tikhomirov@390: import static org.tmatesoft.hg.core.Nodeid.NULL; tikhomirov@423: import static org.tmatesoft.hg.repo.HgRepository.*; tikhomirov@456: import static org.tmatesoft.hg.util.LogFacility.Severity.Info; tikhomirov@218: tikhomirov@232: import java.io.ByteArrayOutputStream; tikhomirov@157: import java.io.IOException; tikhomirov@218: import java.util.ArrayList; tikhomirov@218: import java.util.Arrays; tikhomirov@157: tikhomirov@417: import org.tmatesoft.hg.core.HgChangesetFileSneaker; tikhomirov@74: import org.tmatesoft.hg.core.Nodeid; tikhomirov@423: import org.tmatesoft.hg.internal.Callback; tikhomirov@157: import org.tmatesoft.hg.internal.DataAccess; tikhomirov@266: import org.tmatesoft.hg.internal.DigestHelper; tikhomirov@320: import org.tmatesoft.hg.internal.EncodingHelper; tikhomirov@415: import org.tmatesoft.hg.internal.IntMap; tikhomirov@312: import org.tmatesoft.hg.internal.IterateControlMediator; tikhomirov@218: import org.tmatesoft.hg.internal.Lifecycle; tikhomirov@431: import org.tmatesoft.hg.internal.IdentityPool; tikhomirov@77: import org.tmatesoft.hg.internal.RevlogStream; tikhomirov@312: import org.tmatesoft.hg.util.CancelSupport; tikhomirov@232: import org.tmatesoft.hg.util.Path; tikhomirov@312: import org.tmatesoft.hg.util.ProgressSupport; tikhomirov@456: import org.tmatesoft.hg.util.LogFacility.Severity; tikhomirov@74: tikhomirov@2: tikhomirov@2: /** tikhomirov@425: * Representation of Mercurial manifest file (list of file names and their revisions in a particular changeset) tikhomirov@425: * tikhomirov@425: * @see http://mercurial.selenic.com/wiki/Manifest tikhomirov@74: * @author Artem Tikhomirov tikhomirov@74: * @author TMate Software Ltd. tikhomirov@2: */ tikhomirov@426: public final class HgManifest extends Revlog { tikhomirov@218: private RevisionMapper revisionMap; tikhomirov@412: private EncodingHelper encodingHelper; tikhomirov@285: tikhomirov@415: /** tikhomirov@415: * File flags recorded in manifest tikhomirov@415: */ tikhomirov@285: public enum Flags { tikhomirov@415: /** tikhomirov@415: * Executable bit set tikhomirov@415: */ tikhomirov@415: Exec, tikhomirov@415: /** tikhomirov@415: * Symbolic link tikhomirov@415: */ tikhomirov@415: Link, tikhomirov@415: /** tikhomirov@415: * Regular file tikhomirov@415: */ tikhomirov@415: RegularFile; tikhomirov@285: tikhomirov@285: static Flags parse(String flags) { tikhomirov@285: if ("x".equalsIgnoreCase(flags)) { tikhomirov@285: return Exec; tikhomirov@285: } tikhomirov@285: if ("l".equalsIgnoreCase(flags)) { tikhomirov@285: return Link; tikhomirov@285: } tikhomirov@285: if (flags == null) { tikhomirov@415: return RegularFile; tikhomirov@285: } tikhomirov@285: throw new IllegalStateException(flags); tikhomirov@285: } tikhomirov@285: tikhomirov@285: static Flags parse(byte[] data, int start, int length) { tikhomirov@285: if (length == 0) { tikhomirov@415: return RegularFile; tikhomirov@285: } tikhomirov@285: if (length == 1) { tikhomirov@285: if (data[start] == 'x') { tikhomirov@285: return Exec; tikhomirov@285: } tikhomirov@285: if (data[start] == 'l') { tikhomirov@285: return Link; tikhomirov@285: } tikhomirov@285: // FALL THROUGH tikhomirov@285: } tikhomirov@285: throw new IllegalStateException(new String(data, start, length)); tikhomirov@285: } tikhomirov@285: tikhomirov@285: String nativeString() { tikhomirov@285: if (this == Exec) { tikhomirov@285: return "x"; tikhomirov@285: } tikhomirov@285: if (this == Link) { tikhomirov@285: return "l"; tikhomirov@285: } tikhomirov@415: if (this == RegularFile) { tikhomirov@415: return ""; tikhomirov@415: } tikhomirov@285: throw new IllegalStateException(toString()); tikhomirov@285: } tikhomirov@285: } tikhomirov@2: tikhomirov@412: /*package-local*/ HgManifest(HgRepository hgRepo, RevlogStream content, EncodingHelper eh) { tikhomirov@21: super(hgRepo, content); tikhomirov@412: encodingHelper = eh; tikhomirov@13: } tikhomirov@13: tikhomirov@218: /** tikhomirov@390: * Walks manifest revisions that correspond to specified range of changesets. The order in which manifest versions get reported tikhomirov@390: * to the inspector corresponds to physical order of manifest revisions, not that of changesets (with few exceptions as noted below). tikhomirov@390: * That is, for cset-manifest revision pairs: tikhomirov@390: *
tikhomirov@390: * 3 8 tikhomirov@390: * 4 7 tikhomirov@390: * 5 9 tikhomirov@390: *tikhomirov@390: * call
walk(3,5, insp) would yield (4,7), (3,8) and (5,9) to the inspector; 
tikhomirov@390: 	 * different order of arguments, walk(5, 3, insp), makes no difference.
tikhomirov@390: 	 * 
tikhomirov@390: 	 * Physical layout of mercurial files (revlog) doesn't impose any restriction on whether manifest and changeset revisions shall go 
tikhomirov@390: 	 * incrementally, nor it mandates presence of manifest version for a changeset. Thus, there might be changesets that record {@link Nodeid#NULL}
tikhomirov@390: 	 * as corresponding manifest revision. This situation is deemed exceptional now and what would inspector get depends on whether 
tikhomirov@390: 	 * start or end arguments point to such changeset, or such changeset happen to be somewhere inside the range 
tikhomirov@423: 	 * [start..end]. Implementation does it best to report empty manifests 
tikhomirov@423: 	 * (Inspector.begin(HgRepository.NO_REVISION, NULL, csetRevIndex);
tikhomirov@423: 	 * followed immediately by Inspector.end(HgRepository.NO_REVISION) 
tikhomirov@423: 	 * when start and/or end point to changeset with no associated 
tikhomirov@390: 	 * manifest revision. However, if changeset-manifest revision pairs look like:
tikhomirov@390: 	 * 
tikhomirov@390: * 3 8 tikhomirov@390: * 4 -1 (cset records null revision for manifest) tikhomirov@390: * 5 9 tikhomirov@390: *tikhomirov@390: * call
walk(3,5, insp) would yield only (3,8) and (5,9) to the inspector, without additional empty 
tikhomirov@390: 	 * Inspector.begin(); Inspector.end() call pair.   
tikhomirov@218: 	 * 
tikhomirov@423: 	 * @see HgRepository#NO_REVISION
tikhomirov@218: 	 * @param start changelog (not manifest!) revision to begin with
tikhomirov@218: 	 * @param end changelog (not manifest!) revision to stop, inclusive.
tikhomirov@390: 	 * @param inspector manifest revision visitor, can't be null
tikhomirov@423: 	 * @throws HgRuntimeException subclass thereof to indicate issues with the library. Runtime exception
tikhomirov@423: 	 * @throws IllegalArgumentException if inspector callback is null
tikhomirov@218: 	 */
tikhomirov@423: 	public void walk(int start, int end, final Inspector inspector) throws HgRuntimeException, IllegalArgumentException {
tikhomirov@196: 		if (inspector == null) {
tikhomirov@196: 			throw new IllegalArgumentException();
tikhomirov@196: 		}
tikhomirov@390: 		final int csetFirst = start <= end ? start : end, csetLast = start > end ? start : end;
tikhomirov@390: 		int manifestFirst, manifestLast, i = 0;
tikhomirov@390: 		do {
tikhomirov@390: 			manifestFirst = fromChangelog(csetFirst+i);
tikhomirov@403: 			if (manifestFirst == BAD_REVISION) {
tikhomirov@423: 				inspector.begin(NO_REVISION, NULL, csetFirst+i);
tikhomirov@423: 				inspector.end(NO_REVISION);
tikhomirov@390: 			}
tikhomirov@390: 			i++;
tikhomirov@403: 		} while (manifestFirst == BAD_REVISION && csetFirst+i <= csetLast);
tikhomirov@403: 		if (manifestFirst == BAD_REVISION) {
tikhomirov@490: 			getRepo().getSessionContext().getLog().dump(getClass(), Info, "None of changesets [%d..%d] have associated manifest revision", csetFirst, csetLast);
tikhomirov@390: 			// we ran through all revisions in [start..end] and none of them had manifest.
tikhomirov@390: 			// we reported that to inspector and proceeding is done now.
tikhomirov@390: 			return;
tikhomirov@390: 		}
tikhomirov@390: 		i = 0;
tikhomirov@390: 		do {
tikhomirov@390: 			manifestLast = fromChangelog(csetLast-i);
tikhomirov@403: 			if (manifestLast == BAD_REVISION) {
tikhomirov@423: 				inspector.begin(NO_REVISION, NULL, csetLast-i);
tikhomirov@423: 				inspector.end(NO_REVISION);
tikhomirov@390: 			}
tikhomirov@390: 			i++;
tikhomirov@403: 		} while (manifestLast == BAD_REVISION && csetLast-i >= csetFirst);
tikhomirov@403: 		if (manifestLast == BAD_REVISION) {
tikhomirov@423: 			// hmm, manifestFirst != BAD_REVISION here, hence there's i from [csetFirst..csetLast] for which manifest entry exists, 
tikhomirov@423: 			// and thus it's impossible to run into manifestLast == BAD_REVISION. Nevertheless, never hurts to check.
tikhomirov@423: 			throw new HgInvalidStateException(String.format("Manifest %d-%d(!) for cset range [%d..%d] ", manifestFirst, manifestLast, csetFirst, csetLast));
tikhomirov@390: 		}
tikhomirov@390: 		if (manifestLast < manifestFirst) {
tikhomirov@300: 			// there are tool-constructed repositories that got order of changeset revisions completely different from that of manifest
tikhomirov@390: 			int x = manifestLast;
tikhomirov@390: 			manifestLast = manifestFirst;
tikhomirov@390: 			manifestFirst = x;
tikhomirov@300: 		}
tikhomirov@412: 		content.iterate(manifestFirst, manifestLast, true, new ManifestParser(inspector, encodingHelper));
tikhomirov@19: 	}
tikhomirov@218: 	
tikhomirov@254: 	/**
tikhomirov@390: 	 * "Sparse" iteration of the manifest, more effective than accessing revisions one by one.
tikhomirov@390: 	 *  Inspector is invoked for each changeset revision supplied, even when there's no manifest
tikhomirov@390: 	 * revision associated with a changeset (@see {@link #walk(int, int, Inspector)} for more details when it happens). Order inspector
tikhomirov@390: 	 * gets invoked doesn't resemble order of changeset revisions supplied, manifest revisions are reported in the order they appear 
tikhomirov@390: 	 * in manifest revlog (with exception of changesets with missing manifest that may be reported in any order).   
tikhomirov@254: 	 * 
tikhomirov@390: 	 * @param inspector manifest revision visitor, can't be null
tikhomirov@390: 	 * @param revisionIndexes local indexes of changesets to visit, non-null
tikhomirov@423: 	 * @throws HgRuntimeException subclass thereof to indicate issues with the library. Runtime exception
tikhomirov@423: 	 * @throws InvalidArgumentException if supplied arguments are nulls
tikhomirov@254: 	 */
tikhomirov@423: 	public void walk(final Inspector inspector, int... revisionIndexes) throws HgRuntimeException, IllegalArgumentException {
tikhomirov@367: 		if (inspector == null || revisionIndexes == null) {
tikhomirov@254: 			throw new IllegalArgumentException();
tikhomirov@254: 		}
tikhomirov@390: 		int[] manifestRevs = toManifestRevisionIndexes(revisionIndexes, inspector);
tikhomirov@412: 		content.iterate(manifestRevs, true, new ManifestParser(inspector, encodingHelper));
tikhomirov@254: 	}
tikhomirov@254: 	
tikhomirov@390: 	// 
tikhomirov@390: 	/**
tikhomirov@403: 	 * Tells manifest revision number that corresponds to the given changeset. May return {@link HgRepository#BAD_REVISION} 
tikhomirov@403: 	 * if changeset has no associated manifest (cset records NULL nodeid for manifest).
tikhomirov@403: 	 * @return manifest revision index, non-negative, or {@link HgRepository#BAD_REVISION}.
tikhomirov@403: 	 * @throws HgInvalidRevisionException if method argument specifies non-existent revision index
tikhomirov@403: 	 * @throws HgInvalidControlFileException if access to revlog index/data entry failed
tikhomirov@390: 	 */
tikhomirov@403: 	/*package-local*/ int fromChangelog(int changesetRevisionIndex) throws HgInvalidRevisionException, HgInvalidControlFileException {
tikhomirov@368: 		if (HgInternals.wrongRevisionIndex(changesetRevisionIndex)) {
tikhomirov@425: 			throw new HgInvalidRevisionException(changesetRevisionIndex);
tikhomirov@218: 		}
tikhomirov@368: 		if (changesetRevisionIndex == HgRepository.WORKING_COPY || changesetRevisionIndex == HgRepository.BAD_REVISION) {
tikhomirov@425: 			throw new HgInvalidRevisionException("Can't use constants like WORKING_COPY or BAD_REVISION", null, changesetRevisionIndex);
tikhomirov@257: 		}
tikhomirov@257: 		// revisionNumber == TIP is processed by RevisionMapper 
tikhomirov@218: 		if (revisionMap == null) {
tikhomirov@218: 			revisionMap = new RevisionMapper(getRepo());
tikhomirov@218: 			content.iterate(0, TIP, false, revisionMap);
tikhomirov@218: 		}
tikhomirov@368: 		return revisionMap.at(changesetRevisionIndex);
tikhomirov@218: 	}
tikhomirov@218: 	
tikhomirov@232: 	/**
tikhomirov@232: 	 * Extracts file revision as it was known at the time of given changeset.
tikhomirov@426: 	 * 
For more thorough details about file at specific changeset, use {@link HgChangesetFileSneaker}. tikhomirov@426: *
To visit few changesets for the same file, use {@link #walkFileRevisions(Path, Inspector, int...)}
tikhomirov@232: 	 * 
tikhomirov@426: 	 * @see #walkFileRevisions(Path, Inspector, int...)
tikhomirov@417: 	 * @see HgChangesetFileSneaker
tikhomirov@367: 	 * @param changelogRevisionIndex local changeset index 
tikhomirov@232: 	 * @param file path to file in question
tikhomirov@232: 	 * @return file revision or  NOTE, this method doesn't respect return values from callback (i.e. to stop iteration), as it's lookup of a single file
tikhomirov@426: 	 * and canceling it seems superfluous. However, this may change in future and it's recommended to return null if manifest at specified revision doesn't list such file
tikhomirov@425: 	 * @throws HgRuntimeException subclass thereof to indicate issues with the library. Runtime exception
tikhomirov@232: 	 */
tikhomirov@403: 	public Nodeid getFileRevision(int changelogRevisionIndex, final Path file) throws HgInvalidRevisionException, HgInvalidControlFileException {
tikhomirov@417: 		// there's no need for HgDataFile to own this method, or get a delegate
tikhomirov@417: 		// as most of HgDataFile API is using file revision indexes, and there's easy step from file revision index to
tikhomirov@417: 		// both file revision and changeset revision index. But there's no easy way to go from changesetRevisionIndex to
tikhomirov@417: 		// file revision (the task this method solves), exept for HgFileInformer
tikhomirov@417: 		// I feel methods dealing with changeset indexes shall be more exposed in HgChangelog and HgManifest API.
tikhomirov@426: 		// TODO need tests
tikhomirov@426: 		int manifestRevIndex = fromChangelog(changelogRevisionIndex);
tikhomirov@426: 		if (manifestRevIndex == BAD_REVISION) {
tikhomirov@426: 			return null;
tikhomirov@426: 		}
tikhomirov@426: 		IntMaptrue from
tikhomirov@426: 	 * all {@link Inspector} methods. 
tikhomirov@426: 	 * 
tikhomirov@426: 	 * @see #getFileRevision(int, Path)
tikhomirov@426: 	 * @param file path of interest
tikhomirov@426: 	 * @param inspector callback to receive details about selected file
tikhomirov@426: 	 * @param changelogRevisionIndexes changeset indexes to visit
tikhomirov@426: 	 * @throws HgRuntimeException subclass thereof to indicate issues with the library. Runtime exception
tikhomirov@426: 	 */
tikhomirov@426: 	public void walkFileRevisions(Path file, Inspector inspector, int... changelogRevisionIndexes) throws HgRuntimeException {
tikhomirov@426: 		if (file == null || inspector == null || changelogRevisionIndexes == null) {
tikhomirov@426: 			throw new IllegalArgumentException();
tikhomirov@426: 		}
tikhomirov@426: 		// TODO [post-1.0] need tests. There's Main#checkWalkFileRevisions that may be a starting point
tikhomirov@426: 		int[] manifestRevIndexes = toManifestRevisionIndexes(changelogRevisionIndexes, null);
tikhomirov@426: 		FileLookupInspector parser = new FileLookupInspector(encodingHelper, file, inspector);
tikhomirov@426: 		parser.walk(manifestRevIndexes, content);
tikhomirov@232: 	}
tikhomirov@326: 
tikhomirov@415: 	/**
tikhomirov@417: 	 * Extract file {@link Flags flags} as they were recorded in appropriate manifest version. 
tikhomirov@417: 	 *  
tikhomirov@417: 	 * @see HgDataFile#getFlags(int)
tikhomirov@415: 	 * @param changesetRevIndex changeset revision index
tikhomirov@415: 	 * @param file path to look up
tikhomirov@417: 	 * @return one of predefined enum values, or null if file was not known in the specified revision
tikhomirov@425: 	 * @throws HgRuntimeException subclass thereof to indicate issues with the library. Runtime exception
tikhomirov@415: 	 */
tikhomirov@417: 	public Flags getFileFlags(int changesetRevIndex, Path file) throws HgInvalidRevisionException, HgInvalidControlFileException {
tikhomirov@415: 		int manifestRevIdx = fromChangelog(changesetRevIndex);
tikhomirov@415: 		IntMaptrue to continue iteration, false to stop
tikhomirov@537: 		 */
tikhomirov@537: 		boolean begin(int mainfestRevisionIndex, Nodeid manifestRevision, int changelogRevisionIndex);
tikhomirov@537: 
tikhomirov@537: 		
tikhomirov@537: 		/**
tikhomirov@537: 		 * Reports each manifest entry
tikhomirov@537: 		 *  
tikhomirov@415: 		 * @param nid file revision
tikhomirov@415: 		 * @param fname file name
tikhomirov@415: 		 * @param flags one of {@link HgManifest.Flags} constants, not null
tikhomirov@415: 		 * @return true to continue iteration, false to stop
tikhomirov@415: 		 */
tikhomirov@285: 		boolean next(Nodeid nid, Path fname, Flags flags);
tikhomirov@537: 
tikhomirov@537: 		/**
tikhomirov@537: 		 * Denotes leaving specific manifest revision, after all entries were reported using {@link #next(Nodeid, Path, Flags)}
tikhomirov@537: 		 *   
tikhomirov@537: 		 * @param manifestRevisionIndex indicates manifest revision, corresponds to opening {@link #begin(int, Nodeid, int)}
tikhomirov@537: 		 * @return true to continue iteration, false to stop
tikhomirov@537: 		 */
tikhomirov@537: 		boolean end(int manifestRevisionIndex);
tikhomirov@267: 	}
tikhomirov@424: 	
tikhomirov@268: 	/**
tikhomirov@268: 	 * When Pool uses Strings directly,
tikhomirov@268: 	 * ManifestParser creates new String instance with new char[] value, and does byte->char conversion.
tikhomirov@268: 	 * For cpython repo, walk(0..10k), there are over 16 million filenames, of them only 3020 unique.
tikhomirov@268: 	 * This means there are 15.9 million useless char[] instances and byte->char conversions  
tikhomirov@268: 	 * 
tikhomirov@285: 	 * When String (Path) is wrapped into {@link PathProxy}, there's extra overhead of byte[] representation
tikhomirov@285: 	 * of the String, but these are only for unique Strings (Paths) (3020 in the example above). Besides, I save
tikhomirov@268: 	 * useless char[] and byte->char conversions. 
tikhomirov@268: 	 */
tikhomirov@285: 	private static class PathProxy {
tikhomirov@267: 		private byte[] data;
tikhomirov@268: 		private int start; 
tikhomirov@268: 		private final int hash, length;
tikhomirov@285: 		private Path result;
tikhomirov@412: 		private final EncodingHelper encHelper;
tikhomirov@267: 
tikhomirov@412: 		public PathProxy(byte[] data, int start, int length, EncodingHelper eh) {
tikhomirov@267: 			this.data = data;
tikhomirov@267: 			this.start = start;
tikhomirov@267: 			this.length = length;
tikhomirov@412: 			this.encHelper = eh;
tikhomirov@267: 
tikhomirov@268: 			// copy from String.hashCode(). In fact, not necessarily match result of String(data).hashCode
tikhomirov@268: 			// just need some nice algorithm here
tikhomirov@267: 			int h = 0;
tikhomirov@267: 			byte[] d = data;
tikhomirov@267: 			for (int i = 0, off = start, len = length; i < len; i++) {
tikhomirov@267: 				h = 31 * h + d[off++];
tikhomirov@267: 			}
tikhomirov@267: 			hash = h;
tikhomirov@267: 		}
tikhomirov@267: 
tikhomirov@267: 		@Override
tikhomirov@267: 		public boolean equals(Object obj) {
tikhomirov@285: 			if (false == obj instanceof PathProxy) {
tikhomirov@267: 				return false;
tikhomirov@267: 			}
tikhomirov@285: 			PathProxy o = (PathProxy) obj;
tikhomirov@267: 			if (o.result != null && result != null) {
tikhomirov@267: 				return result.equals(o.result);
tikhomirov@267: 			}
tikhomirov@268: 			if (o.length != length || o.hash != hash) {
tikhomirov@267: 				return false;
tikhomirov@267: 			}
tikhomirov@267: 			for (int i = 0, x = o.start, y = start; i < length; i++) {
tikhomirov@267: 				if (o.data[x++] != data[y++]) {
tikhomirov@267: 					return false;
tikhomirov@267: 				}
tikhomirov@267: 			}
tikhomirov@267: 			return true;
tikhomirov@267: 		}
tikhomirov@267: 		@Override
tikhomirov@267: 		public int hashCode() {
tikhomirov@267: 			return hash;
tikhomirov@267: 		}
tikhomirov@267: 		
tikhomirov@285: 		public Path freeze() {
tikhomirov@267: 			if (result == null) {
tikhomirov@412: 				result = Path.create(encHelper.fromManifest(data, start, length));
tikhomirov@268: 				// release reference to bigger data array, make a copy of relevant part only
tikhomirov@285: 				// use original bytes, not those from String above to avoid cache misses due to different encodings 
tikhomirov@268: 				byte[] d = new byte[length];
tikhomirov@268: 				System.arraycopy(data, start, d, 0, length);
tikhomirov@268: 				data = d;
tikhomirov@268: 				start = 0;
tikhomirov@267: 			}
tikhomirov@267: 			return result;
tikhomirov@267: 		}
tikhomirov@267: 	}
tikhomirov@196: 
tikhomirov@312: 	private static class ManifestParser implements RevlogStream.Inspector, Lifecycle {
tikhomirov@196: 		private final Inspector inspector;
tikhomirov@431: 		private IdentityPool