kitaev@213: /*
kitaev@213:  * Copyright (c) 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 java.nio.ByteBuffer;
kitaev@213: import java.util.ArrayList;
kitaev@213: import java.util.Map;
kitaev@213: import java.util.TreeMap;
kitaev@213: 
kitaev@213: import org.tmatesoft.hg.repo.HgChangelog.RawChangeset;
kitaev@213: import org.tmatesoft.hg.repo.HgRepository;
kitaev@213: import org.tmatesoft.hg.util.Path;
kitaev@213: 
kitaev@213: /**
kitaev@213:  *
kitaev@213:  * @author Artem Tikhomirov
kitaev@213:  * @author TMate Software Ltd.
kitaev@213:  */
kitaev@213: public class KeywordFilter implements Filter {
kitaev@213: 	// present implementation is stateless, however, filter use pattern shall not assume that. In fact, Factory may us that 
kitaev@213: 	private final HgRepository repo;
kitaev@213: 	private final boolean isExpanding;
kitaev@213: 	private final TreeMap<String,String> keywords;
kitaev@213: 	private final int minBufferLen;
kitaev@213: 	private final Path path;
kitaev@213: 	private RawChangeset latestFileCset;
kitaev@213: 
kitaev@213: 	/**
kitaev@213: 	 * 
kitaev@213: 	 * @param hgRepo 
kitaev@213: 	 * @param path 
kitaev@213: 	 * @param expand <code>true</code> to expand keywords, <code>false</code> to shrink
kitaev@213: 	 */
kitaev@213: 	private KeywordFilter(HgRepository hgRepo, Path p, boolean expand) {
kitaev@213: 		repo = hgRepo;
kitaev@213: 		path = p;
kitaev@213: 		isExpanding = expand;
kitaev@213: 		keywords = new TreeMap<String,String>();
kitaev@213: 		keywords.put("Id", "Id");
kitaev@213: 		keywords.put("Revision", "Revision");
kitaev@213: 		keywords.put("Author", "Author");
kitaev@213: 		keywords.put("Date", "Date");
kitaev@213: 		keywords.put("LastChangedRevision", "LastChangedRevision");
kitaev@213: 		keywords.put("LastChangedBy", "LastChangedBy");
kitaev@213: 		keywords.put("LastChangedDate", "LastChangedDate");
kitaev@213: 		keywords.put("Source", "Source");
kitaev@213: 		keywords.put("Header", "Header");
kitaev@213: 
kitaev@213: 		int l = 0;
kitaev@213: 		for (String s : keywords.keySet()) {
kitaev@213: 			if (s.length() > l) {
kitaev@213: 				l = s.length();
kitaev@213: 			}
kitaev@213: 		}
kitaev@213: 		// FIXME later may implement #filter() not to read full kw value (just "$kw:"). However, limit of maxLen + 2 would keep valid.
kitaev@213: 		// for buffers less then minBufferLen, there are chances #filter() implementation would never end
kitaev@213: 		// (i.e. for input "$LongestKey"$
kitaev@213: 		minBufferLen = l + 2 + (isExpanding ? 0 : 120 /*any reasonable constant for max possible kw value length*/);
kitaev@213: 	}
kitaev@213: 
kitaev@213: 	/**
kitaev@213: 	 * @param src buffer ready to be read
kitaev@213: 	 * @return buffer ready to be read and original buffer's position modified to reflect consumed bytes. IOW, if source buffer
kitaev@213: 	 * on return has remaining bytes, they are assumed not-read (not processed) and next chunk passed to filter is supposed to 
kitaev@213: 	 * start with them  
kitaev@213: 	 */
kitaev@213: 	public ByteBuffer filter(ByteBuffer src) {
kitaev@213: 		if (src.capacity() < minBufferLen) {
kitaev@213: 			throw new IllegalStateException(String.format("Need buffer of at least %d bytes to ensure filter won't hang", minBufferLen));
kitaev@213: 		}
kitaev@213: 		ByteBuffer rv = null;
kitaev@213: 		int keywordStart = -1;
kitaev@213: 		int x = src.position();
kitaev@213: 		int copyFrom = x; // needs to be updated each time we copy a slice, but not each time we modify source index (x)
kitaev@213: 		while (x < src.limit()) {
kitaev@213: 			if (keywordStart == -1) {
kitaev@213: 				int i = indexOf(src, '$', x, false);
kitaev@213: 				if (i == -1) {
kitaev@213: 					if (rv == null) {
kitaev@213: 						return src;
kitaev@213: 					} else {
kitaev@213: 						copySlice(src, copyFrom, src.limit(), rv);
kitaev@213: 						rv.flip();
kitaev@213: 						src.position(src.limit());
kitaev@213: 						return rv;
kitaev@213: 					}
kitaev@213: 				}
kitaev@213: 				keywordStart = i;
kitaev@213: 				// fall-through
kitaev@213: 			}
kitaev@213: 			if (keywordStart >= 0) {
kitaev@213: 				int i = indexOf(src, '$', keywordStart+1, true);
kitaev@213: 				if (i == -1) {
kitaev@213: 					// end of buffer reached
kitaev@213: 					if (rv == null) {
kitaev@213: 						if (keywordStart == x) {
kitaev@213: 							// FIXME in fact, x might be equal to keywordStart and to src.position() here ('$' is first character in the buffer, 
kitaev@213: 							// and there are no other '$' not eols till the end of the buffer). This would lead to deadlock (filter won't consume any
kitaev@213: 							// bytes). To prevent this, either shall copy bytes [keywordStart..buffer.limit()) to local buffer and use it on the next invocation,
kitaev@213: 							// or add lookup of the keywords right after first '$' is found (do not wait for closing '$'). For now, large enough src buffer would be sufficient
kitaev@213: 							// not to run into such situation
kitaev@213: 							throw new IllegalStateException("Try src buffer of a greater size");
kitaev@213: 						}
kitaev@213: 						rv = ByteBuffer.allocate(keywordStart - copyFrom);
kitaev@213: 					}
kitaev@213: 					// copy all from source till latest possible kw start 
kitaev@213: 					copySlice(src, copyFrom, keywordStart, rv);
kitaev@213: 					rv.flip();
kitaev@213: 					// and tell caller we've consumed only to the potential kw start
kitaev@213: 					src.position(keywordStart);
kitaev@213: 					return rv;
kitaev@213: 				} else if (src.get(i) == '$') {
kitaev@213: 					// end of keyword, or start of a new one.
kitaev@213: 					String keyword;
kitaev@213: 					if ((keyword = matchKeyword(src, keywordStart, i)) != null) {
kitaev@213: 						if (rv == null) {
kitaev@213: 							// src.remaining(), not .capacity because src is not read, and remaining represents 
kitaev@213: 							// actual bytes count, while capacity - potential.
kitaev@213: 							// Factor of 4 is pure guess and a HACK, need to be fixed with re-expanding buffer on demand
kitaev@213: 							rv = ByteBuffer.allocate(isExpanding ? src.remaining() * 4 : src.remaining());
kitaev@213: 						}
kitaev@213: 						copySlice(src, copyFrom, keywordStart+1, rv);
kitaev@213: 						rv.put(keyword.getBytes());
kitaev@213: 						if (isExpanding) {
kitaev@213: 							rv.put((byte) ':');
kitaev@213: 							rv.put((byte) ' ');
kitaev@213: 							expandKeywordValue(keyword, rv);
kitaev@213: 							rv.put((byte) ' ');
kitaev@213: 						}
kitaev@213: 						rv.put((byte) '$');
kitaev@213: 						keywordStart = -1;
kitaev@213: 						x = i+1;
kitaev@213: 						copyFrom = x;
kitaev@213: 						continue;
kitaev@213: 					} else {
kitaev@213: 						if (rv != null) {
kitaev@213: 							// we've already did some substitution, thus need to copy bytes we've scanned. 
kitaev@213: 							copySlice(src, x, i, rv);
kitaev@213: 							copyFrom = i;
kitaev@213: 						} // no else in attempt to avoid rv creation if no real kw would be found  
kitaev@213: 						keywordStart = i;
kitaev@213: 						x = i; // '$' at i wasn't consumed, hence x points to i, not i+1. This is to avoid problems with case: "sdfsd $ asdfs $Id$ sdf"
kitaev@213: 						continue;
kitaev@213: 					}
kitaev@213: 				} else {
kitaev@213: 					assert src.get(i) == '\n' || src.get(i) == '\r';
kitaev@213: 					// line break
kitaev@213: 					if (rv != null) {
kitaev@213: 						copySlice(src, x, i+1, rv);
kitaev@213: 						copyFrom = i+1;
kitaev@213: 					}
kitaev@213: 					x = i+1;
kitaev@213: 					keywordStart = -1; // Wasn't keyword, really
kitaev@213: 					continue; // try once again
kitaev@213: 				}
kitaev@213: 			}
kitaev@213: 		}
kitaev@213: 		if (keywordStart != -1) {
kitaev@213: 			if (rv == null) {
kitaev@213: 				// no expansion happened yet, and we have potential kw start
kitaev@213: 				rv = ByteBuffer.allocate(keywordStart - src.position());
kitaev@213: 				copySlice(src, src.position(), keywordStart, rv);
kitaev@213: 			}
kitaev@213: 			src.position(keywordStart);
kitaev@213: 		}
kitaev@213: 		if (rv != null) {
kitaev@213: 			rv.flip();
kitaev@213: 			return rv;
kitaev@213: 		}
kitaev@213: 		return src;
kitaev@213: 	}
kitaev@213: 	
kitaev@213: 	/**
kitaev@213: 	 * @param keyword
kitaev@213: 	 * @param rv
kitaev@213: 	 */
kitaev@213: 	private void expandKeywordValue(String keyword, ByteBuffer rv) {
kitaev@213: 		if ("Id".equals(keyword)) {
kitaev@213: 			rv.put(identityString().getBytes());
kitaev@213: 		} else if ("Revision".equals(keyword)) {
kitaev@213: 			rv.put(revision().getBytes());
kitaev@213: 		} else if ("Author".equals(keyword)) {
kitaev@213: 			rv.put(username().getBytes());
kitaev@213: 		} else if ("Date".equals(keyword)) {
kitaev@213: 			rv.put(date().getBytes());
kitaev@213: 		} else {
kitaev@213: 			throw new IllegalStateException(String.format("Keyword %s is not yet supported", keyword));
kitaev@213: 		}
kitaev@213: 	}
kitaev@213: 
kitaev@213: 	private String matchKeyword(ByteBuffer src, int kwStart, int kwEnd) {
kitaev@213: 		assert kwEnd - kwStart - 1 > 0;
kitaev@213: 		assert src.get(kwStart) == src.get(kwEnd) && src.get(kwEnd) == '$';
kitaev@213: 		char[] chars = new char[kwEnd - kwStart - 1];
kitaev@213: 		int i;
kitaev@213: 		for (i = 0; i < chars.length; i++) {
kitaev@213: 			char c = (char) src.get(kwStart + 1 + i);
kitaev@213: 			if (c == ':') {
kitaev@213: 				break;
kitaev@213: 			}
kitaev@213: 			chars[i] = c;
kitaev@213: 		}
kitaev@213: 		String kw = new String(chars, 0, i);
kitaev@213: //		XXX may use subMap to look up keywords based on few available characters (not waiting till closing $)
kitaev@213: //		System.out.println(keywords.subMap("I", "J"));
kitaev@213: //		System.out.println(keywords.subMap("A", "B"));
kitaev@213: //		System.out.println(keywords.subMap("Au", "B"));
kitaev@213: 		return keywords.get(kw);
kitaev@213: 	}
kitaev@213: 	
kitaev@213: 	// copies part of the src buffer, [from..to). doesn't modify src position
kitaev@213: 	static void copySlice(ByteBuffer src, int from, int to, ByteBuffer dst) {
kitaev@213: 		if (to > src.limit()) {
kitaev@213: 			throw new IllegalArgumentException("Bad right boundary");
kitaev@213: 		}
kitaev@213: 		if (dst.remaining() < to - from) {
kitaev@213: 			throw new IllegalArgumentException("Not enough room in the destination buffer");
kitaev@213: 		}
kitaev@213: 		for (int i = from; i < to; i++) {
kitaev@213: 			dst.put(src.get(i));
kitaev@213: 		}
kitaev@213: 	}
kitaev@213: 
kitaev@213: 	private static int indexOf(ByteBuffer b, char ch, int from, boolean newlineBreaks) {
kitaev@213: 		for (int i = from; i < b.limit(); i++) {
kitaev@213: 			byte c = b.get(i);
kitaev@213: 			if (ch == c) {
kitaev@213: 				return i;
kitaev@213: 			}
kitaev@213: 			if (newlineBreaks && (c == '\n' || c == '\r')) {
kitaev@213: 				return i;
kitaev@213: 			}
kitaev@213: 		}
kitaev@213: 		return -1;
kitaev@213: 	}
kitaev@213: 
kitaev@213: 	private String identityString() {
kitaev@213: 		return String.format("%s,v %s %s %s", path, revision(), date(), username());
kitaev@213: 	}
kitaev@213: 
kitaev@213: 	private String revision() {
kitaev@213: 		// FIXME add cset's nodeid into Changeset class 
kitaev@213: 		int csetRev = repo.getFileNode(path).getChangesetLocalRevision(HgRepository.TIP);
kitaev@213: 		return repo.getChangelog().getRevision(csetRev).shortNotation();
kitaev@213: 	}
kitaev@213: 	
kitaev@213: 	private String username() {
kitaev@213: 		return getChangeset().user();
kitaev@213: 	}
kitaev@213: 	
kitaev@213: 	private String date() {
kitaev@213: 		return String.format("%tY/%<tm/%<td %<tH:%<tM:%<tS", getChangeset().date());
kitaev@213: 	}
kitaev@213: 	
kitaev@213: 	private RawChangeset getChangeset() {
kitaev@213: 		if (latestFileCset == null) {
kitaev@213: 			// XXX consider use of ChangelogHelper
kitaev@213: 			int csetRev = repo.getFileNode(path).getChangesetLocalRevision(HgRepository.TIP);
kitaev@213: 			latestFileCset = repo.getChangelog().range(csetRev, csetRev).get(0);
kitaev@213: 		}
kitaev@213: 		return latestFileCset;
kitaev@213: 	}
kitaev@213: 
kitaev@213: 	public static class Factory implements Filter.Factory {
kitaev@213: 		
kitaev@213: 		private HgRepository repo;
kitaev@213: 		private Path.Matcher matcher;
kitaev@213: 
kitaev@213: 		public void initialize(HgRepository hgRepo, ConfigFile cfg) {
kitaev@213: 			repo = hgRepo;
kitaev@213: 			ArrayList<String> patterns = new ArrayList<String>();
kitaev@213: 			for (Map.Entry<String,String> e : cfg.getSection("keyword").entrySet()) {
kitaev@213: 				if (!"ignore".equalsIgnoreCase(e.getValue())) {
kitaev@213: 					patterns.add(e.getKey());
kitaev@213: 				}
kitaev@213: 			}
kitaev@213: 			matcher = new PathGlobMatcher(patterns.toArray(new String[patterns.size()]));
kitaev@213: 			// TODO read and respect keyword patterns from [keywordmaps]
kitaev@213: 		}
kitaev@213: 
kitaev@213: 		public Filter create(Path path, Options opts) {
kitaev@213: 			if (matcher.accept(path)) {
kitaev@213: 				return new KeywordFilter(repo, path, opts.getDirection() == Filter.Direction.FromRepo);
kitaev@213: 			}
kitaev@213: 			return null;
kitaev@213: 		}
kitaev@213: 	}
kitaev@213: 
kitaev@213: //
kitaev@213: //	public static void main(String[] args) throws Exception {
kitaev@213: //		FileInputStream fis = new FileInputStream(new File("/temp/kwoutput.txt"));
kitaev@213: //		FileOutputStream fos = new FileOutputStream(new File("/temp/kwoutput2.txt"));
kitaev@213: //		ByteBuffer b = ByteBuffer.allocate(256);
kitaev@213: //		KeywordFilter kwFilter = new KeywordFilter(false);
kitaev@213: //		while (fis.getChannel().read(b) != -1) {
kitaev@213: //			b.flip(); // get ready to be read
kitaev@213: //			ByteBuffer f = kwFilter.filter(b);
kitaev@213: //			fos.getChannel().write(f); // XXX in fact, f may not be fully consumed
kitaev@213: //			if (b.hasRemaining()) {
kitaev@213: //				b.compact();
kitaev@213: //			} else {
kitaev@213: //				b.clear();
kitaev@213: //			}
kitaev@213: //		}
kitaev@213: //		fis.close();
kitaev@213: //		fos.flush();
kitaev@213: //		fos.close();
kitaev@213: //	}
kitaev@213: }