Title:
- * SequenceIdMatcher - *Description:
- * Routine which does approximate Sequence Id resolution by name using string containment rather than equivalence - *Copyright: Copyright (c) 2004
- * - *Company: Dundee University
- * - * @author not attributable - * @version 1.0 + * 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. */ public class SequenceIdMatcher { + private Hashtable names; - private class SeqIdName + public SequenceIdMatcher(SequenceI[] seqs) { - String id; - - SeqIdName(String s) - { - id = new String(s); - } - - public int hashCode() - { - return (id.substring(0, 4).hashCode()); - } - - public boolean equals(Object s) + names = new Hashtable(); + for (int i = 0; i < seqs.length; i++) { - if (s instanceof SeqIdName) + // 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]); + SequenceI dbseq = seqs[i]; + while (dbseq.getDatasetSequence()!=null) { - return this.equals( (SeqIdName) s); + dbseq = dbseq.getDatasetSequence(); } - else + // add in any interesting identifiers + if (dbseq.getDBRef() != null) { - if (s instanceof String) + DBRefEntry dbr[] = dbseq.getDBRef(); + SeqIdName sid = null; + for (int r = 0; r < dbr.length; r++) { - return this.equals( (String) s); + sid = new SeqIdName(dbr[r].getAccessionId()); + if (!names.contains(sid)) + { + names.put(sid, seqs[i]); + } } } - return false; } + } - public boolean equals(SeqIdName s) + /** + * 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 SequenceI closest SequenceI to SeqIdName + */ + 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) { - if (id.startsWith(s.id) || s.id.startsWith(id)) - { - return true; - } - return false; + return null; } - - public boolean equals(String s) + 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(); + while (matches.size() > 0) { - if (id.startsWith(s) || s.startsWith(id)) + // look through for a better one. + SequenceI cand = (SequenceI) matches.elementAt(0); + matches.remove(0); + names.put(new SeqIdName(cand.getName()), cand); + int q, w, candlen = cand.getName().length(); + // keep the one with an id 'closer' to the given seqnam string + if ((q = Math.abs(matchlen - namlen)) > (w = Math.abs(candlen + - namlen)) + && candlen > matchlen) { - return true; + best.clear(); + match = cand; + matchlen = candlen; + best.add(match); + } + if (q == w && candlen == matchlen) + { + // record any ties + best.add(cand); } - return false; } - } - - private Hashtable names; - - public SequenceIdMatcher(SequenceI[] seqs) - { - names = new Hashtable(); - for (int i = 0; i < seqs.length; i++) + if (best.size() == 0) { - names.put(new SeqIdName(seqs[i].getName()), seqs[i]); + return null; } + ; + return (SequenceI[]) best.toArray(new SequenceI[0]); } - SequenceI findIdMatch(SequenceI seq) + /** + * get SequenceI with closest SequenceI.getName() to seq.getName() + * + * @param seq + * SequenceI + * @return SequenceI + */ + public SequenceI findIdMatch(SequenceI seq) { SeqIdName nam = new SeqIdName(seq.getName()); - if (names.containsKey(nam)) - { - return (SequenceI) names.get(nam); - } - return null; + return findIdMatch(nam); } - SequenceI findIdMatch(String seqnam) + public SequenceI findIdMatch(String seqnam) { SeqIdName nam = new SeqIdName(seqnam); - if (names.containsKey(nam)) - { - return (SequenceI) names.get(nam); - } - return null; + 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 same Id as a given set of different sequence objects - * - * @param seqs SequenceI[] + * + * Return pointers to sequences (or sequence object containers) which have + * same Id as a given set of different sequence objects + * + * @param seqs + * SequenceI[] * @return SequenceI[] */ - - SequenceI[] findIdMatch(SequenceI[] seqs) + public SequenceI[] findIdMatch(SequenceI[] seqs) { - SequenceI[] namedseqs = new SequenceI[seqs.length]; - + SequenceI[] namedseqs = null; int i = 0; SeqIdName nam; + if (seqs.length > 0) { + namedseqs = new SequenceI[seqs.length]; do { nam = new SeqIdName(seqs[i].getName()); + if (names.containsKey(nam)) { - namedseqs[i] = (SequenceI) names.get(nam); + namedseqs[i] = findIdMatch(nam); } else { namedseqs[i] = null; } - } - while (i++ < seqs.length); + } while (++i < seqs.length); } + return namedseqs; } + /** + * core findIdMatch search method + * + * @param nam + * SeqIdName + * @return SequenceI + */ + private SequenceI findIdMatch( + jalview.analysis.SequenceIdMatcher.SeqIdName nam) + { + Vector matches = new Vector(); + while (names.containsKey(nam)) + { + matches.addElement(names.remove(nam)); + } + 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; + + SeqIdName(String s) + { + if (s != null) + { + id = new String(s); + } + else + { + id = ""; + } + } + + public int hashCode() + { + return ((id.length() >= 4) ? id.substring(0, 4).hashCode() : id + .hashCode()); + } + + public boolean equals(Object s) + { + if (s instanceof SeqIdName) + { + return this.equals((SeqIdName) s); + } + else + { + if (s instanceof String) + { + return this.equals((String) s); + } + } + + return false; + } + + /** + * Characters that define the end of a unique sequence ID at the beginning + * of an arbitrary ID string JBPNote: This is a heuristic that will fail for + * arbritrarily extended sequence id's (like portions of an aligned set of + * repeats from one sequence) + */ + private String WORD_SEP = "~. |#\\/<>!\"" + ((char) 0x00A4) + + "$%^*)}[@',?_"; + + /** + * matches if one ID properly contains another at a whitespace boundary. + * TODO: (JBPNote) These are not efficient. should use char[] for speed + * todo: (JBPNote) Set separator characters appropriately + * + * @param s + * SeqIdName + * @return boolean + */ + 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 + .length())) > -1) : false; + } + else + { + return s.id.startsWith(id) ? (s.id.equals(id) ? true : (WORD_SEP + .indexOf(s.id.charAt(id.length())) > -1)) : false; + } + } + + public boolean equals(String s) + { + if (id.length() > s.length()) + { + return id.startsWith(s) ? (WORD_SEP.indexOf(id.charAt(s.length())) > -1) + : false; + } + else + { + return s.startsWith(id) ? (s.equals(id) ? true : (WORD_SEP + .indexOf(s.charAt(id.length())) > -1)) : false; + } + } + } }