- * Title: - *
- * SequenceIdMatcher - *- * Description: - *
- * Routine which does approximate Sequence Id resolution by name using string + * Routines for approximate Sequence Id resolution by name using string * containment (on word boundaries) rather than equivalence. It also attempts to * resolve ties where no exact match is available by picking the the id closest * to the query. - *- * Copyright: Copyright (c) 2004 - *
- * - *- * Company: Dundee University - *
- * - * @author not attributable - * @version 1.0 */ public class SequenceIdMatcher { @@ -53,7 +39,9 @@ public class SequenceIdMatcher names = new Hashtable(); for (int i = 0; i < seqs.length; i++) { - names.put(new SeqIdName(seqs[i].getName()), seqs[i]); + // TODO: deal with ID collisions - SequenceI should be appended to list + // associated with this key. + names.put(new SeqIdName(seqs[i].getDisplayId(true)), seqs[i]); // add in any interesting identifiers if (seqs[i].getDBRef() != null) { @@ -83,6 +71,24 @@ public class SequenceIdMatcher */ private SequenceI pickbestMatch(SeqIdName candName, Vector matches) { + SequenceI[] st = pickbestMatches(candName, matches); + return st == null || st.length == 0 ? null : st[0]; + } + + /** + * returns the closest SequenceI in matches to SeqIdName and returns all the + * matches to the names hash. + * + * @param candName + * SeqIdName + * @param matches + * Vector of SequenceI objects + * @return Object[] { SequenceI closest SequenceI to SeqIdName, SequenceI[] + * ties } + */ + private SequenceI[] pickbestMatches(SeqIdName candName, Vector matches) + { + ArrayList best = new ArrayList(); SequenceI match = null; if (candName == null || matches == null || matches.size() == 0) { @@ -90,6 +96,7 @@ public class SequenceIdMatcher } match = (SequenceI) matches.elementAt(0); matches.removeElementAt(0); + best.add(match); names.put(new SeqIdName(match.getName()), match); int matchlen = match.getName().length(); int namlen = candName.id.length(); @@ -97,17 +104,31 @@ public class SequenceIdMatcher { // look through for a better one. SequenceI cand = (SequenceI) matches.elementAt(0); + matches.remove(0); names.put(new SeqIdName(cand.getName()), cand); - int candlen = cand.getName().length(); + int q, w, candlen = cand.getName().length(); // keep the one with an id 'closer' to the given seqnam string - if (Math.abs(matchlen - namlen) > Math.abs(candlen - namlen) + if ((q = Math.abs(matchlen - namlen)) > (w = Math.abs(candlen + - namlen)) && candlen > matchlen) { + best.clear(); match = cand; matchlen = candlen; + best.add(match); + } + if (q == w && candlen == matchlen) + { + // record any ties + best.add(cand); } } - return match; + if (best.size() == 0) + { + return null; + } + ; + return (SequenceI[]) best.toArray(new SequenceI[0]); } /** @@ -117,19 +138,32 @@ public class SequenceIdMatcher * SequenceI * @return SequenceI */ - SequenceI findIdMatch(SequenceI seq) + public SequenceI findIdMatch(SequenceI seq) { SeqIdName nam = new SeqIdName(seq.getName()); return findIdMatch(nam); } - SequenceI findIdMatch(String seqnam) + public SequenceI findIdMatch(String seqnam) { SeqIdName nam = new SeqIdName(seqnam); return findIdMatch(nam); } /** + * Find all matches for a given sequence name. + * + * @param seqnam + * string to query Matcher with. + */ + public SequenceI[] findAllIdMatches(String seqnam) + { + + SeqIdName nam = new SeqIdName(seqnam); + return findAllIdMatches(nam); + } + + /** * findIdMatch * * Return pointers to sequences (or sequence object containers) which have @@ -139,7 +173,7 @@ public class SequenceIdMatcher * SequenceI[] * @return SequenceI[] */ - SequenceI[] findIdMatch(SequenceI[] seqs) + public SequenceI[] findIdMatch(SequenceI[] seqs) { SequenceI[] namedseqs = null; int i = 0; @@ -184,6 +218,25 @@ public class SequenceIdMatcher return pickbestMatch(nam, matches); } + /** + * core findIdMatch search method for finding all equivalent matches + * + * @param nam + * SeqIdName + * @return SequenceI[] + */ + private SequenceI[] findAllIdMatches( + jalview.analysis.SequenceIdMatcher.SeqIdName nam) + { + Vector matches = new Vector(); + while (names.containsKey(nam)) + { + matches.addElement(names.remove(nam)); + } + SequenceI[] r = pickbestMatches(nam, matches); + return r; + } + private class SeqIdName { String id; @@ -229,7 +282,8 @@ public class SequenceIdMatcher * arbritrarily extended sequence id's (like portions of an aligned set of * repeats from one sequence) */ - private String WORD_SEP = "~. |#\\/<>!\"£$%^*)}[@',?_"; + private String WORD_SEP = "~. |#\\/<>!\"" + ((char) 0x00A4) + + "$%^*)}[@',?_"; /** * matches if one ID properly contains another at a whitespace boundary. @@ -242,6 +296,8 @@ public class SequenceIdMatcher */ public boolean equals(SeqIdName s) { + // TODO: JAL-732 patch for cases when name includes a list of IDs, and the + // match contains one ID flanked if (id.length() > s.id.length()) { return id.startsWith(s.id) ? (WORD_SEP.indexOf(id.charAt(s.id