kitaev@213: /*
kitaev@213:  * Copyright (c) 2010-2011 TMate Software Ltd
kitaev@213:  *  
kitaev@213:  * This program is free software; you can redistribute it and/or modify
kitaev@213:  * it under the terms of the GNU General Public License as published by
kitaev@213:  * the Free Software Foundation; version 2 of the License.
kitaev@213:  *
kitaev@213:  * This program is distributed in the hope that it will be useful,
kitaev@213:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
kitaev@213:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
kitaev@213:  * GNU General Public License for more details.
kitaev@213:  *
kitaev@213:  * For information on how to redistribute this software under
kitaev@213:  * the terms of a license other than GNU General Public License
kitaev@213:  * contact TMate Software at support@hg4j.com
kitaev@213:  */
kitaev@213: package org.tmatesoft.hg.internal;
kitaev@213: 
kitaev@213: import static org.tmatesoft.hg.repo.HgRepository.BAD_REVISION;
kitaev@213: import static org.tmatesoft.hg.repo.HgRepository.TIP;
kitaev@213: 
kitaev@213: import java.io.File;
kitaev@213: import java.io.IOException;
kitaev@213: import java.util.ArrayList;
kitaev@213: import java.util.LinkedList;
kitaev@213: import java.util.List;
kitaev@213: 
kitaev@213: import org.tmatesoft.hg.core.HgBadStateException;
kitaev@213: import org.tmatesoft.hg.core.Nodeid;
kitaev@213: import org.tmatesoft.hg.repo.HgRepository;
kitaev@213: 
kitaev@213: 
kitaev@213: /**
kitaev@213:  * ? Single RevlogStream per file per repository with accessor to record access session (e.g. with back/forward operations), 
kitaev@213:  * or numerous RevlogStream with separate representation of the underlying data (cached, lazy ChunkStream)?
kitaev@213:  * 
kitaev@213:  * @see http://mercurial.selenic.com/wiki/Revlog
kitaev@213:  * @see http://mercurial.selenic.com/wiki/RevlogNG
kitaev@213:  *
kitaev@213:  * @author Artem Tikhomirov
kitaev@213:  * @author TMate Software Ltd.
kitaev@213:  */
kitaev@213: public class RevlogStream {
kitaev@213: 
kitaev@213: 	/*
kitaev@213: 	 * makes sense for index with inline data only - actual offset of the record in the .i file (record entry + revision * record size))
kitaev@213: 	 * 
kitaev@213: 	 * long[] in fact (there are 8-bytes field in the revlog)
kitaev@213: 	 * However, (a) DataAccess currently doesn't operate with long seek/length
kitaev@213: 	 * and, of greater significance, (b) files with inlined data are designated for smaller files,  
kitaev@213: 	 * guess, about 130 Kb, and offset there won't ever break int capacity
kitaev@213: 	 */
kitaev@213: 	private int[] indexRecordOffset;  
kitaev@213: 	private int[] baseRevisions;
kitaev@213: 	private boolean inline = false;
kitaev@213: 	private final File indexFile;
kitaev@213: 	private final DataAccessProvider dataAccess;
kitaev@213: 
kitaev@213: 	// if we need anything else from HgRepo, might replace DAP parameter with HgRepo and query it for DAP.
kitaev@213: 	public RevlogStream(DataAccessProvider dap, File indexFile) {
kitaev@213: 		this.dataAccess = dap;
kitaev@213: 		this.indexFile = indexFile;
kitaev@213: 	}
kitaev@213: 
kitaev@213: 	/*package*/ DataAccess getIndexStream() {
kitaev@213: 		return dataAccess.create(indexFile);
kitaev@213: 	}
kitaev@213: 
kitaev@213: 	/*package*/ DataAccess getDataStream() {
kitaev@213: 		final String indexName = indexFile.getName();
kitaev@213: 		File dataFile = new File(indexFile.getParentFile(), indexName.substring(0, indexName.length() - 1) + "d");
kitaev@213: 		return dataAccess.create(dataFile);
kitaev@213: 	}
kitaev@213: 	
kitaev@213: 	public int revisionCount() {
kitaev@213: 		initOutline();
kitaev@213: 		return baseRevisions.length;
kitaev@213: 	}
kitaev@213: 	
kitaev@213: 	public int dataLength(int revision) {
kitaev@213: 		// XXX in fact, use of iterate() instead of this implementation may be quite reasonable.
kitaev@213: 		//
kitaev@213: 		final int indexSize = revisionCount();
kitaev@213: 		DataAccess daIndex = getIndexStream(); // XXX may supply a hint that I'll need really few bytes of data (although at some offset)
kitaev@213: 		if (revision == TIP) {
kitaev@213: 			revision = indexSize - 1;
kitaev@213: 		}
kitaev@213: 		try {
kitaev@213: 			int recordOffset = getIndexOffsetInt(revision);
kitaev@213: 			daIndex.seek(recordOffset + 12); // 6+2+4
kitaev@213: 			int actualLen = daIndex.readInt();
kitaev@213: 			return actualLen; 
kitaev@213: 		} catch (IOException ex) {
kitaev@213: 			ex.printStackTrace(); // log error. FIXME better handling
kitaev@213: 			throw new IllegalStateException(ex);
kitaev@213: 		} finally {
kitaev@213: 			daIndex.done();
kitaev@213: 		}
kitaev@213: 	}
kitaev@213: 	
kitaev@213: 	public byte[] nodeid(int revision) {
kitaev@213: 		final int indexSize = revisionCount();
kitaev@213: 		if (revision == TIP) {
kitaev@213: 			revision = indexSize - 1;
kitaev@213: 		}
kitaev@213: 		if (revision < 0 || revision >= indexSize) {
kitaev@213: 			throw new IllegalArgumentException(Integer.toString(revision));
kitaev@213: 		}
kitaev@213: 		DataAccess daIndex = getIndexStream();
kitaev@213: 		try {
kitaev@213: 			int recordOffset = getIndexOffsetInt(revision);
kitaev@213: 			daIndex.seek(recordOffset + 32);
kitaev@213: 			byte[] rv = new byte[20];
kitaev@213: 			daIndex.readBytes(rv, 0, 20);
kitaev@213: 			return rv;
kitaev@213: 		} catch (IOException ex) {
kitaev@213: 			ex.printStackTrace();
kitaev@213: 			throw new IllegalStateException();
kitaev@213: 		} finally {
kitaev@213: 			daIndex.done();
kitaev@213: 		}
kitaev@213: 	}
kitaev@213: 	
kitaev@213: 	public int linkRevision(int revision) {
kitaev@213: 		final int last = revisionCount() - 1;
kitaev@213: 		if (revision == TIP) {
kitaev@213: 			revision = last;
kitaev@213: 		}
kitaev@213: 		if (revision < 0 || revision > last) {
kitaev@213: 			throw new IllegalArgumentException(Integer.toString(revision));
kitaev@213: 		}
kitaev@213: 		DataAccess daIndex = getIndexStream();
kitaev@213: 		try {
kitaev@213: 			int recordOffset = getIndexOffsetInt(revision);
kitaev@213: 			daIndex.seek(recordOffset + 20);
kitaev@213: 			int linkRev = daIndex.readInt();
kitaev@213: 			return linkRev;
kitaev@213: 		} catch (IOException ex) {
kitaev@213: 			ex.printStackTrace();
kitaev@213: 			throw new IllegalStateException();
kitaev@213: 		} finally {
kitaev@213: 			daIndex.done();
kitaev@213: 		}
kitaev@213: 	}
kitaev@213: 	
kitaev@213: 	// Perhaps, RevlogStream should be limited to use of plain int revisions for access,
kitaev@213: 	// while Nodeids should be kept on the level up, in Revlog. Guess, Revlog better keep
kitaev@213: 	// map of nodeids, and once this comes true, we may get rid of this method.
kitaev@213: 	// Unlike its counterpart, {@link Revlog#getLocalRevisionNumber()}, doesn't fail with exception if node not found,
kitaev@213: 	/**
kitaev@213: 	 * @return integer in [0..revisionCount()) or {@link HgRepository#BAD_REVISION} if not found
kitaev@213: 	 */
kitaev@213: 	public int findLocalRevisionNumber(Nodeid nodeid) {
kitaev@213: 		// XXX this one may be implemented with iterate() once there's mechanism to stop iterations
kitaev@213: 		final int indexSize = revisionCount();
kitaev@213: 		DataAccess daIndex = getIndexStream();
kitaev@213: 		try {
kitaev@213: 			byte[] nodeidBuf = new byte[20];
kitaev@213: 			for (int i = 0; i < indexSize; i++) {
kitaev@213: 				daIndex.skip(8);
kitaev@213: 				int compressedLen = daIndex.readInt();
kitaev@213: 				daIndex.skip(20);
kitaev@213: 				daIndex.readBytes(nodeidBuf, 0, 20);
kitaev@213: 				if (nodeid.equalsTo(nodeidBuf)) {
kitaev@213: 					return i;
kitaev@213: 				}
kitaev@213: 				daIndex.skip(inline ? 12 + compressedLen : 12);
kitaev@213: 			}
kitaev@213: 		} catch (IOException ex) {
kitaev@213: 			ex.printStackTrace(); // log error. FIXME better handling
kitaev@213: 			throw new IllegalStateException(ex);
kitaev@213: 		} finally {
kitaev@213: 			daIndex.done();
kitaev@213: 		}
kitaev@213: 		return BAD_REVISION;
kitaev@213: 	}
kitaev@213: 
kitaev@213: 
kitaev@213: 	private final int REVLOGV1_RECORD_SIZE = 64;
kitaev@213: 
kitaev@213: 	// should be possible to use TIP, ALL, or -1, -2, -n notation of Hg
kitaev@213: 	// ? boolean needsNodeid
kitaev@213: 	public void iterate(int start, int end, boolean needData, Inspector inspector) {
kitaev@213: 		initOutline();
kitaev@213: 		final int indexSize = revisionCount();
kitaev@213: 		if (indexSize == 0) {
kitaev@213: 			return;
kitaev@213: 		}
kitaev@213: 		if (end == TIP) {
kitaev@213: 			end = indexSize - 1;
kitaev@213: 		}
kitaev@213: 		if (start == TIP) {
kitaev@213: 			start = indexSize - 1;
kitaev@213: 		}
kitaev@213: 		if (start < 0 || start >= indexSize) {
kitaev@213: 			throw new IllegalArgumentException(String.format("Bad left range boundary %d in [0..%d]", start, indexSize-1));
kitaev@213: 		}
kitaev@213: 		if (end < start || end >= indexSize) {
kitaev@213: 			throw new IllegalArgumentException(String.format("Bad right range boundary %d in [0..%d]", end, indexSize-1));
kitaev@213: 		}
kitaev@213: 		// XXX may cache [start .. end] from index with a single read (pre-read)
kitaev@213: 		
kitaev@213: 		DataAccess daIndex = null, daData = null;
kitaev@213: 		daIndex = getIndexStream();
kitaev@213: 		if (needData && !inline) {
kitaev@213: 			daData = getDataStream();
kitaev@213: 		}
kitaev@213: 		try {
kitaev@213: 			byte[] nodeidBuf = new byte[20];
kitaev@213: 			DataAccess lastUserData = null;
kitaev@213: 			int i;
kitaev@213: 			boolean extraReadsToBaseRev = false;
kitaev@213: 			if (needData && getBaseRevision(start) < start) {
kitaev@213: 				i = getBaseRevision(start);
kitaev@213: 				extraReadsToBaseRev = true;
kitaev@213: 			} else {
kitaev@213: 				i = start;
kitaev@213: 			}
kitaev@213: 			
kitaev@213: 			daIndex.seek(getIndexOffsetInt(i));
kitaev@213: 			for (; i <= end; i++ ) {
kitaev@213: 				if (inline && needData) {
kitaev@213: 					// inspector reading data (though FilterDataAccess) may have affected index position
kitaev@213: 					daIndex.seek(getIndexOffsetInt(i));
kitaev@213: 				}
kitaev@213: 				long l = daIndex.readLong(); // 0
kitaev@213: 				long offset = i == 0 ? 0 : (l >>> 16);
kitaev@213: 				@SuppressWarnings("unused")
kitaev@213: 				int flags = (int) (l & 0X0FFFF);
kitaev@213: 				int compressedLen = daIndex.readInt(); // +8
kitaev@213: 				int actualLen = daIndex.readInt(); // +12
kitaev@213: 				int baseRevision = daIndex.readInt(); // +16
kitaev@213: 				int linkRevision = daIndex.readInt(); // +20
kitaev@213: 				int parent1Revision = daIndex.readInt();
kitaev@213: 				int parent2Revision = daIndex.readInt();
kitaev@213: 				// Hg has 32 bytes here, uses 20 for nodeid, and keeps 12 last bytes empty
kitaev@213: 				daIndex.readBytes(nodeidBuf, 0, 20); // +32
kitaev@213: 				daIndex.skip(12);
kitaev@213: 				DataAccess userDataAccess = null;
kitaev@213: 				if (needData) {
kitaev@213: 					final byte firstByte;
kitaev@213: 					int streamOffset;
kitaev@213: 					DataAccess streamDataAccess;
kitaev@213: 					if (inline) {
kitaev@213: 						streamDataAccess = daIndex;
kitaev@213: 						streamOffset = getIndexOffsetInt(i) + REVLOGV1_RECORD_SIZE; // don't need to do seek as it's actual position in the index stream
kitaev@213: 					} else {
kitaev@213: 						streamOffset = (int) offset;
kitaev@213: 						streamDataAccess = daData;
kitaev@213: 						daData.seek(streamOffset);
kitaev@213: 					}
kitaev@213: 					final boolean patchToPrevious = baseRevision != i; // the only way I found to tell if it's a patch
kitaev@213: 					firstByte = streamDataAccess.readByte();
kitaev@213: 					if (firstByte == 0x78 /* 'x' */) {
kitaev@213: 						userDataAccess = new InflaterDataAccess(streamDataAccess, streamOffset, compressedLen, patchToPrevious ? -1 : actualLen);
kitaev@213: 					} else if (firstByte == 0x75 /* 'u' */) {
kitaev@213: 						userDataAccess = new FilterDataAccess(streamDataAccess, streamOffset+1, compressedLen-1);
kitaev@213: 					} else {
kitaev@213: 						// XXX Python impl in fact throws exception when there's not 'x', 'u' or '0'
kitaev@213: 						// but I don't see reason not to return data as is 
kitaev@213: 						userDataAccess = new FilterDataAccess(streamDataAccess, streamOffset, compressedLen);
kitaev@213: 					}
kitaev@213: 					// XXX 
kitaev@213: 					if (patchToPrevious) {
kitaev@213: 						// this is a patch
kitaev@213: 						LinkedList<PatchRecord> patches = new LinkedList<PatchRecord>();
kitaev@213: 						while (!userDataAccess.isEmpty()) {
kitaev@213: 							PatchRecord pr = PatchRecord.read(userDataAccess);
kitaev@213: //							System.out.printf("PatchRecord:%d %d %d\n", pr.start, pr.end, pr.len);
kitaev@213: 							patches.add(pr);
kitaev@213: 						}
kitaev@213: 						userDataAccess.done();
kitaev@213: 						//
kitaev@213: 						byte[] userData = apply(lastUserData, actualLen, patches);
kitaev@213: 						userDataAccess = new ByteArrayDataAccess(userData);
kitaev@213: 					}
kitaev@213: 				} else {
kitaev@213: 					if (inline) {
kitaev@213: 						daIndex.skip(compressedLen);
kitaev@213: 					}
kitaev@213: 				}
kitaev@213: 				if (!extraReadsToBaseRev || i >= start) {
kitaev@213: 					inspector.next(i, actualLen, baseRevision, linkRevision, parent1Revision, parent2Revision, nodeidBuf, userDataAccess);
kitaev@213: 				}
kitaev@213: 				if (userDataAccess != null) {
kitaev@213: 					userDataAccess.reset();
kitaev@213: 					if (lastUserData != null) {
kitaev@213: 						lastUserData.done();
kitaev@213: 					}
kitaev@213: 					lastUserData = userDataAccess;
kitaev@213: 				}
kitaev@213: 			}
kitaev@213: 		} catch (IOException ex) {
kitaev@213: 			throw new HgBadStateException(ex); // FIXME need better handling
kitaev@213: 		} finally {
kitaev@213: 			daIndex.done();
kitaev@213: 			if (daData != null) {
kitaev@213: 				daData.done();
kitaev@213: 			}
kitaev@213: 		}
kitaev@213: 	}
kitaev@213: 
kitaev@213: 	private int getBaseRevision(int revision) {
kitaev@213: 		return baseRevisions[revision];
kitaev@213: 	}
kitaev@213: 
kitaev@213: 	/**
kitaev@213: 	 * @return  offset of the revision's record in the index (.i) stream
kitaev@213: 	 */
kitaev@213: 	private int getIndexOffsetInt(int revision) {
kitaev@213: 		return inline ? indexRecordOffset[revision] : revision * REVLOGV1_RECORD_SIZE;
kitaev@213: 	}
kitaev@213: 
kitaev@213: 	private void initOutline() {
kitaev@213: 		if (baseRevisions != null && baseRevisions.length > 0) {
kitaev@213: 			return;
kitaev@213: 		}
kitaev@213: 		ArrayList<Integer> resBases = new ArrayList<Integer>();
kitaev@213: 		ArrayList<Integer> resOffsets = new ArrayList<Integer>();
kitaev@213: 		DataAccess da = getIndexStream();
kitaev@213: 		try {
kitaev@213: 			if (da.isEmpty()) {
kitaev@213: 				// do not fail with exception if stream is empty, it's likely intentional
kitaev@213: 				baseRevisions = new int[0];
kitaev@213: 				return;
kitaev@213: 			}
kitaev@213: 			int versionField = da.readInt();
kitaev@213: 			da.readInt(); // just to skip next 4 bytes of offset + flags
kitaev@213: 			final int INLINEDATA = 1 << 16;
kitaev@213: 			inline = (versionField & INLINEDATA) != 0;
kitaev@213: 			long offset = 0; // first offset is always 0, thus Hg uses it for other purposes
kitaev@213: 			while(true) {
kitaev@213: 				int compressedLen = da.readInt();
kitaev@213: 				// 8+4 = 12 bytes total read here
kitaev@213: 				@SuppressWarnings("unused")
kitaev@213: 				int actualLen = da.readInt();
kitaev@213: 				int baseRevision = da.readInt();
kitaev@213: 				// 12 + 8 = 20 bytes read here
kitaev@213: //				int linkRevision = di.readInt();
kitaev@213: //				int parent1Revision = di.readInt();
kitaev@213: //				int parent2Revision = di.readInt();
kitaev@213: //				byte[] nodeid = new byte[32];
kitaev@213: 				resBases.add(baseRevision);
kitaev@213: 				if (inline) {
kitaev@213: 					int o = (int) offset;
kitaev@213: 					if (o != offset) {
kitaev@213: 						// just in case, can't happen, ever, unless HG (or some other bad tool) produces index file 
kitaev@213: 						// with inlined data of size greater than 2 Gb.
kitaev@213: 						throw new HgBadStateException("Data too big, offset didn't fit to sizeof(int)");
kitaev@213: 					}
kitaev@213: 					resOffsets.add(o + REVLOGV1_RECORD_SIZE * resOffsets.size());
kitaev@213: 					da.skip(3*4 + 32 + compressedLen); // Check: 44 (skip) + 20 (read) = 64 (total RevlogNG record size)
kitaev@213: 				} else {
kitaev@213: 					da.skip(3*4 + 32);
kitaev@213: 				}
kitaev@213: 				if (da.isEmpty()) {
kitaev@213: 					// fine, done then
kitaev@213: 					baseRevisions = toArray(resBases);
kitaev@213: 					if (inline) {
kitaev@213: 						indexRecordOffset = toArray(resOffsets);
kitaev@213: 					}
kitaev@213: 					break;
kitaev@213: 				} else {
kitaev@213: 					// start reading next record
kitaev@213: 					long l = da.readLong();
kitaev@213: 					offset = l >>> 16;
kitaev@213: 				}
kitaev@213: 			}
kitaev@213: 		} catch (IOException ex) {
kitaev@213: 			ex.printStackTrace(); // log error
kitaev@213: 			// too bad, no outline then, but don't fail with NPE
kitaev@213: 			baseRevisions = new int[0];
kitaev@213: 		} finally {
kitaev@213: 			da.done();
kitaev@213: 		}
kitaev@213: 	}
kitaev@213: 	
kitaev@213: 	private static int[] toArray(List<Integer> l) {
kitaev@213: 		int[] rv = new int[l.size()];
kitaev@213: 		for (int i = 0; i < rv.length; i++) {
kitaev@213: 			rv[i] = l.get(i);
kitaev@213: 		}
kitaev@213: 		return rv;
kitaev@213: 	}
kitaev@213: 	
kitaev@213: 
kitaev@213: 	// mpatch.c : apply()
kitaev@213: 	// FIXME need to implement patch merge (fold, combine, gather and discard from aforementioned mpatch.[c|py]), also see Revlog and Mercurial PDF
kitaev@213: 	public/*for HgBundle; until moved to better place*/static byte[] apply(DataAccess baseRevisionContent, int outcomeLen, List<PatchRecord> patch) throws IOException {
kitaev@213: 		int last = 0, destIndex = 0;
kitaev@213: 		if (outcomeLen == -1) {
kitaev@213: 			outcomeLen = baseRevisionContent.length();
kitaev@213: 			for (PatchRecord pr : patch) {
kitaev@213: 				outcomeLen += pr.start - last + pr.len;
kitaev@213: 				last = pr.end;
kitaev@213: 			}
kitaev@213: 			outcomeLen -= last;
kitaev@213: 			last = 0;
kitaev@213: 		}
kitaev@213: 		byte[] rv = new byte[outcomeLen];
kitaev@213: 		for (PatchRecord pr : patch) {
kitaev@213: 			baseRevisionContent.seek(last);
kitaev@213: 			baseRevisionContent.readBytes(rv, destIndex, pr.start-last);
kitaev@213: 			destIndex += pr.start - last;
kitaev@213: 			System.arraycopy(pr.data, 0, rv, destIndex, pr.data.length);
kitaev@213: 			destIndex += pr.data.length;
kitaev@213: 			last = pr.end;
kitaev@213: 		}
kitaev@213: 		baseRevisionContent.seek(last);
kitaev@213: 		baseRevisionContent.readBytes(rv, destIndex, (int) (baseRevisionContent.length() - last));
kitaev@213: 		return rv;
kitaev@213: 	}
kitaev@213: 
kitaev@213: 	// @see http://mercurial.selenic.com/wiki/BundleFormat, in Changelog group description
kitaev@213: 	public static class PatchRecord {
kitaev@213: 		/*
kitaev@213: 		   Given there are pr1 and pr2:
kitaev@213: 		     pr1.start to pr1.end will be replaced with pr's data (of pr1.len)
kitaev@213: 		     pr1.end to pr2.start gets copied from base
kitaev@213: 		 */
kitaev@213: 		public int start, end, len;
kitaev@213: 		public byte[] data;
kitaev@213: 
kitaev@213: 		// TODO consider PatchRecord that only records data position (absolute in data source), and acquires data as needed 
kitaev@213: 		private PatchRecord(int p1, int p2, int length, byte[] src) {
kitaev@213: 			start = p1;
kitaev@213: 			end = p2;
kitaev@213: 			len = length;
kitaev@213: 			data = src;
kitaev@213: 		}
kitaev@213: 
kitaev@213: 		/*package-local*/ static PatchRecord read(byte[] data, int offset) {
kitaev@213: 			final int x = offset; // shorthand
kitaev@213: 			int p1 =  ((data[x] & 0xFF)<< 24)    | ((data[x+1] & 0xFF) << 16) | ((data[x+2] & 0xFF) << 8)  | (data[x+3] & 0xFF);
kitaev@213: 			int p2 =  ((data[x+4] & 0xFF) << 24) | ((data[x+5] & 0xFF) << 16) | ((data[x+6] & 0xFF) << 8)  | (data[x+7] & 0xFF);
kitaev@213: 			int len = ((data[x+8] & 0xFF) << 24) | ((data[x+9] & 0xFF) << 16) | ((data[x+10] & 0xFF) << 8) | (data[x+11] & 0xFF);
kitaev@213: 			byte[] dataCopy = new byte[len];
kitaev@213: 			System.arraycopy(data, x+12, dataCopy, 0, len);
kitaev@213: 			return new PatchRecord(p1, p2, len, dataCopy);
kitaev@213: 		}
kitaev@213: 
kitaev@213: 		public /*for HgBundle*/ static PatchRecord read(DataAccess da) throws IOException {
kitaev@213: 			int p1 = da.readInt();
kitaev@213: 			int p2 = da.readInt();
kitaev@213: 			int len = da.readInt();
kitaev@213: 			byte[] src = new byte[len];
kitaev@213: 			da.readBytes(src, 0, len);
kitaev@213: 			return new PatchRecord(p1, p2, len, src);
kitaev@213: 		}
kitaev@213: 	}
kitaev@213: 
kitaev@213: 	// FIXME byte[] data might be too expensive, for few usecases it may be better to have intermediate Access object (when we don't need full data 
kitaev@213: 	// instantly - e.g. calculate hash, or comparing two revisions
kitaev@213: 	public interface Inspector {
kitaev@213: 		// XXX boolean retVal to indicate whether to continue?
kitaev@213: 		// TODO specify nodeid and data length, and reuse policy (i.e. if revlog stream doesn't reuse nodeid[] for each call)
kitaev@213: 		// implementers shall not invoke DataAccess.done(), it's accomplished by #iterate at appropraite moment
kitaev@213: 		void next(int revisionNumber, int actualLen, int baseRevision, int linkRevision, int parent1Revision, int parent2Revision, byte[/*20*/] nodeid, DataAccess data);
kitaev@213: 	}
kitaev@213: }