import java.util.ArrayList;
import java.util.Arrays;
import java.util.HashMap;
+import java.util.HashSet;
import java.util.List;
-import java.util.Vector;
+import java.util.Set;
/**
* Routines for approximate Sequence Id resolution by name using string
*/
public class SequenceIdMatcher
{
- private HashMap<SeqIdName, SequenceI> names;
+ /**
+ * weak hash for each sequence
+ */
+ private HashMap<SeqIdName, Set<SequenceI>> names;
+
+ // /**
+ // * cache of values removed for each query string.
+ // */
+ // private HashMap<String, List<SequenceI>> resolved;
+
+ /**
+ * do we index sequences on all 'words' in ID string ?
+ */
+ private boolean wordBased = 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)
+ */
+ static String WORD_SEP = "~. |#\\/<>!\"" + ((char) 0x00A4)
+ + "$%^*)}[@',?_";
+
+ /**
+ * @return true if matcher is word-based (ie string key matches one of the
+ * words within the body of one or more sequence IDs)
+ */
+ public boolean isWordBased()
+ {
+ return wordBased;
+ }
+ /**
+ * Construct a standard (non-word based) matcher. To configure word based
+ * matching, use the fully qualified constructor
+ *
+ * @param seqs
+ */
public SequenceIdMatcher(List<SequenceI> seqs)
{
- names = new HashMap<SeqIdName, SequenceI>();
+ this(false, seqs);
+ }
+
+ /**
+ * construct a new matcher for a set of sequences, configured as required.
+ * Note: enabling word based matching
+ *
+ * @param wordBasedMatch
+ * - when true, "myseq" matches "X|myseq" and "myseq"
+ * @param seqs
+ */
+ public SequenceIdMatcher(boolean wordBasedMatch, List<SequenceI> seqs)
+ {
+ wordBased = wordBasedMatch;
+ names = new HashMap<SeqIdName, Set<SequenceI>>();
addAll(seqs);
}
{
for (SequenceI seq : seqs)
{
- // TODO: deal with ID collisions - SequenceI should be appended to list
- // associated with this key.
- names.put(new SeqIdName(seq.getDisplayId(true)), seq);
- SequenceI dbseq = seq;
- while (dbseq.getDatasetSequence() != null)
+ addSeq(seq);
+ }
+ }
+
+ private void addSeqIdName(SeqIdName idname, SequenceI seq)
+ {
+ Set<SequenceI> seqset = names.get(idname);
+ if (seqset == null)
+ {
+ seqset = new HashSet<SequenceI>();
+ names.put(idname, seqset);
+ }
+ seqset.add(seq);
+ }
+
+ public void addSeq(SequenceI seq)
+ {
+ // TODO: deal with ID collisions - SequenceI should be appended to list
+ // associated with this key.
+ addSeqIdName(new SeqIdName(seq.getDisplayId(true)), seq);
+ if (wordBased)
+ {
+ for (SeqIdName key : getWordsFor(seq))
+ {
+ addSeqIdName(key, seq);
+ }
+ }
+ SequenceI dbseq = seq;
+ // TODO add test for database xref resolution
+ while (dbseq.getDatasetSequence() != null)
+ {
+ dbseq = dbseq.getDatasetSequence();
+ }
+ // add in any interesting identifiers
+ if (dbseq.getDBRefs() != null)
+ {
+ DBRefEntry dbr[] = dbseq.getDBRefs();
+ SeqIdName sid = null;
+ for (int r = 0; r < dbr.length; r++)
+ {
+ sid = new SeqIdName(dbr[r].getAccessionId());
+ if (!names.containsKey(sid))
+ {
+ addSeqIdName(sid, seq);
+ }
+ }
+ }
+ }
+
+ /**
+ * generate word based keys for the given sequence
+ *
+ * @param seq
+ * @return list of split keys
+ */
+ public static List<SeqIdName> getWordsFor(SequenceI seq)
+ {
+ ArrayList<SeqIdName> keys = new ArrayList<SeqIdName>();
+ String name = seq.getName(), limits = "/" + seq.getStart() + "-"
+ + seq.getEnd();
+ int namel = name.length();
+ char[] sep = new char[WORD_SEP.length()];
+ // find only the separators present in the ID.
+ for (int i = 0; i < sep.length; i++)
+ {
+ sep[i] = WORD_SEP.charAt(i);
+ if (seq.getName().indexOf("" + sep[i]) == -1)
{
- dbseq = dbseq.getDatasetSequence();
+ sep[i] = 0;
}
- // add in any interesting identifiers
- if (dbseq.getDBRefs() != null)
+ }
+ ;
+ // make words
+ for (int i = 0; i < sep.length; i++)
+ {
+ if (sep[i] > 0)
{
- DBRefEntry dbr[] = dbseq.getDBRefs();
- SeqIdName sid = null;
- for (int r = 0; r < dbr.length; r++)
+ int p = 0, m = -1;
+ while ((m = name.indexOf(sep[i], p)) > p)
{
- sid = new SeqIdName(dbr[r].getAccessionId());
- if (!names.containsKey(sid))
+
+ if (m > 0 && m - p > 5)
+ {
+ // split to end of word m with this delimiter
+ keys.add(new SeqIdName(name.substring(p, m - 1) + limits));
+ }
+ if (namel - m > 5)
{
- names.put(sid, seq);
+ // index word after this delimiter m
+ keys.add(new SeqIdName(name.substring(m + 1) + limits));
}
+ p = m + 1;
+ }
+ if (namel - p > 4)
+ {
+ // index word after this delimiter m
+ keys.add(new SeqIdName(name.substring(p) + limits));
}
}
}
+ return keys;
}
/**
- * convenience method to make a matcher from concrete array
+ * convenience method to make a matcher from concrete array Note: in order to
+ * support word based matching, use the fully qualified constructor
*
* @param sequences
*/
}
SequenceI match = matches.remove(0);
best.add(match);
- names.put(new SeqIdName(match.getName()), match);
+ addSeq(match);
int matchlen = match.getName().length();
int namlen = candName.id.length();
while (matches.size() > 0)
{
// look through for a better one.
SequenceI cand = matches.remove(0);
- names.put(new SeqIdName(cand.getName()), cand);
+ addSeq(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)
+ boolean is_closer = ((q = Math.abs(matchlen - namlen)) > (w = Math
+ .abs(candlen - namlen)) && candlen > matchlen);
+ // if not closer, then check if current best is actually identical in case
+ // as
+ // well
+ if (is_closer
+ || (candName.equalsCase(cand.getName()) && !candName
+ .equalsCase(best.get(0).getName())))
{
best.clear();
match = cand;
matchlen = candlen;
best.add(match);
}
- if (q == w && candlen == matchlen)
+ else
{
- // record any ties
- best.add(cand);
+ if (q == w && candlen == matchlen)
+ {
+ // equivalently good, and matches with case as well. so
+ // record any ties
+ best.add(cand);
+ }
}
}
if (best.size() == 0)
* @return SequenceI
*/
private SequenceI findIdMatch(
- jalview.analysis.SequenceIdMatcher.SeqIdName nam)
+ jalview.analysis.SeqIdName nam)
{
- Vector matches = new Vector();
+ ArrayList<SequenceI> matches = new ArrayList<SequenceI>();
while (names.containsKey(nam))
{
- matches.addElement(names.remove(nam));
+ matches.addAll(names.remove(nam));
}
return pickbestMatch(nam, matches);
}
* @return SequenceI[]
*/
private List<SequenceI> findAllIdMatches(
- jalview.analysis.SequenceIdMatcher.SeqIdName nam)
+ jalview.analysis.SeqIdName nam)
{
ArrayList<SequenceI> matches = new ArrayList<SequenceI>();
while (names.containsKey(nam))
{
- matches.add(names.remove(nam));
+ matches.addAll(names.remove(nam));
}
List<SequenceI> r = pickbestMatches(nam, matches);
return r;
}
-
- private class SeqIdName
- {
- String id;
-
- SeqIdName(String s)
- {
- if (s != null)
- {
- id = new String(s).toLowerCase();
- }
- else
- {
- id = "";
- }
- }
-
- @Override
- public int hashCode()
- {
- return ((id.length() >= 4) ? id.substring(0, 4).hashCode() : id
- .hashCode());
- }
-
- @Override
- public boolean equals(Object s)
- {
- if (s == null)
- {
- return false;
- }
- 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 check_wordmatch(id, s.id);
- }
- else
- {
- return check_wordmatch(s.id, id);
- }
- }
-
- private boolean check_wordmatch(String longer, String shorter)
- {
- boolean elen = longer.length() == shorter.length();
- int sp = longer.indexOf(shorter);
- if (sp == -1)
- {
- return false;
- }
-
- if (sp == 0)
- {
- // end of match is word boundary
- return elen ? true : (WORD_SEP.indexOf(longer.charAt(shorter
- .length() + sp)) > -1);
- }
- if (WORD_SEP.indexOf(longer.charAt(sp - 1)) > -1)
- {
- if (sp + shorter.length() == longer.length())
- {
- return true;
- }
- else
- {
- // end of match is word boundary
- return elen ? false
- : sp + shorter.length() == longer.length() ? true
- : (WORD_SEP.indexOf(longer.charAt(shorter
- .length() + sp)) > -1);
- }
- }
- else
- {
- // prefix of match is not a word boundary
- return false;
- }
- }
-
- public boolean equals(String s)
- {
- s = s.toLowerCase(); // TODO: employ faster to lower case operation
- if (id.length() > s.length())
- {
- return check_wordmatch(id, s);
- }
- else
- {
- return check_wordmatch(s, id);
- }
- }
- }
}
git://source.jalview.org / jalview.git / blobdiff