Title:
* SequenceIdMatcher *Description:
- * Routine which does approximate Sequence Id resolution by name using string containment rather than equivalence + * Routine which does 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
@@ -41,35 +44,63 @@ public class SequenceIdMatcher public SequenceIdMatcher(SequenceI[] seqs) { names = new Hashtable(); - for (int i = 0; i < seqs.length; i++) { names.put(new SeqIdName(seqs[i].getName()), seqs[i]); } } - SequenceI findIdMatch(SequenceI seq) + /** + * 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) { - SeqIdName nam = new SeqIdName(seq.getName()); - - if (names.containsKey(nam)) + SequenceI match = null; + if (candName == null || matches == null || matches.size() == 0) { - return (SequenceI) names.get(nam); + return null; } + match = (SequenceI) matches.elementAt(0); + matches.removeElementAt(0); + names.put(new SeqIdName(match.getName()), match); + int matchlen = match.getName().length(); + int namlen = candName.id.length(); + while (matches.size() > 0) + { + // look through for a better one. + SequenceI cand = (SequenceI) matches.elementAt(0); + names.put(new SeqIdName(cand.getName()), cand); + int 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) && + candlen > matchlen) + { + match = cand; + matchlen = candlen; + } + } + return match; + } - return null; + /** + * get SequenceI with closest SequenceI.getName() to seq.getName() + * @param seq SequenceI + * @return SequenceI + */ + SequenceI findIdMatch(SequenceI seq) + { + SeqIdName nam = new SeqIdName(seq.getName()); + return findIdMatch(nam); } SequenceI findIdMatch(String seqnam) { SeqIdName nam = new SeqIdName(seqnam); - - if (names.containsKey(nam)) - { - return (SequenceI) names.get(nam); - } - - return null; + return findIdMatch(nam); } /** @@ -83,44 +114,67 @@ public class SequenceIdMatcher */ 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); + } + private class SeqIdName { String id; SeqIdName(String s) { - id = new String(s); + if (s != null) + { + id = new String(s); + } + else + { + id = ""; + } } public int hashCode() { - return (id.substring(0, 4).hashCode()); + return ( (id.length() >= 4) ? id.substring(0, 4).hashCode() : id.hashCode()); } public boolean equals(Object s) @@ -140,24 +194,53 @@ public class SequenceIdMatcher 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 = "~. |#\\/<>!\"£$%^*)}[@',?_"; + + /** + * 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) { - if (id.startsWith(s.id) || s.id.startsWith(id)) + if (id.length() > s.id.length()) { - return true; + 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; } - - return false; } public boolean equals(String s) { - if (id.startsWith(s) || s.startsWith(id)) + if (id.length() > s.length()) { - return true; + 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; } - - return false; } } }