Sequence uniquify/deuniquify now uses the SequenceIdMatcher to cope with mangled...

[jalview.git] / src / jalview / analysis / SeqsetUtils.java

diff --git a/src/jalview/analysis/SeqsetUtils.java b/src/jalview/analysis/SeqsetUtils.java
index 564e3f1..4432414 100755 (executable)
--- a/src/jalview/analysis/SeqsetUtils.java
+++ b/src/jalview/analysis/SeqsetUtils.java
@@ -114,6 +114,14 @@ public class SeqsetUtils
     return new String("Sequence" + i);\r
   }\r
 \r
+  /**\r
+   * Generates a hash of SeqCharacterHash properties for each sequence\r
+   * in a sequence set, and optionally renames the sequences to an\r
+   * unambiguous 'safe' name.\r
+   * @param sequences SequenceI[]\r
+   * @param write_names boolean set this to rename each of the sequences to its unique_name(index) name\r
+   * @return Hashtable to be passed to @see deuniquify to recover original names (and properties) for renamed sequences\r
+   */\r
   public static Hashtable uniquify(SequenceI[] sequences, boolean write_names)\r
   {\r
     // Generate a safely named sequence set and a hash to recover the sequence names\r
@@ -134,25 +142,72 @@ public class SeqsetUtils
 \r
     return map;\r
   }\r
-\r
+  /**\r
+   * recover unsafe sequence names and original properties for a sequence\r
+   * set using a map generated by @see uniquify(sequences,true)\r
+   * @param map Hashtable\r
+   * @param sequences SequenceI[]\r
+   * @return boolean\r
+   */\r
   public static boolean deuniquify(Hashtable map, SequenceI[] sequences)\r
   {\r
-    // recover unsafe sequence names for a sequence set\r
-    boolean allfound = true;\r
-\r
-    for (int i = 0; i < sequences.length; i++)\r
-    {\r
-      if (map.containsKey(sequences[i].getName()))\r
-      {\r
-        Hashtable sqinfo = (Hashtable) map.get(sequences[i].getName());\r
-        SeqCharacterUnhash(sequences[i], sqinfo);\r
-      }\r
-      else\r
-      {\r
-        allfound = false;\r
+    jalview.analysis.SequenceIdMatcher matcher = new SequenceIdMatcher(sequences);\r
+    SequenceI msq = null;\r
+    Enumeration keys = map.keys();\r
+    Vector unmatched = new Vector();\r
+    for (int i=0, j=sequences.length; i<j; i++)\r
+      unmatched.add(sequences[i]);\r
+    while (keys.hasMoreElements()) {\r
+      Object key = keys.nextElement();\r
+      if (key instanceof String) {\r
+        if ((msq = matcher.findIdMatch((String) key))!=null) {\r
+          Hashtable sqinfo = (Hashtable) map.get(key);\r
+          unmatched.remove(msq);\r
+          SeqCharacterUnhash(msq, sqinfo);\r
+        }\r
+        else\r
+        {\r
+          System.err.println("Can't find '"+((String) key)+"' in uniquified alignment");\r
+        }\r
       }\r
     }\r
+    if (unmatched.size()>0) {\r
+      System.err.println("Did not find matches for :");\r
+      for (Enumeration i = unmatched.elements(); i.hasMoreElements(); System.out.println(((SequenceI) i.nextElement()).getName()))\r
+           ;\r
+      return false;\r
+    }\r
 \r
-    return allfound;\r
+    return true;\r
+  }\r
+  /**\r
+   * returns a subset of the sequenceI seuqences,\r
+   * including only those that contain at least one residue.\r
+   * @param sequences SequenceI[]\r
+   * @return SequenceI[]\r
+   */\r
+  public static SequenceI[] getNonEmptySequenceSet(SequenceI[] sequences) {\r
+      // Identify first row of alignment with residues for prediction\r
+      boolean ungapped[] = new boolean[sequences.length];\r
+      int msflen=0;\r
+      for (int i=0,j=sequences.length; i<j;i++) {\r
+        String tempseq = jalview.analysis.AlignSeq.extractGaps(jalview.util.Comparison.GapChars, sequences[i].getSequence());\r
+        if (tempseq.length()==0)\r
+          ungapped[i]=false;\r
+        else {\r
+          ungapped[i]=true;\r
+          msflen++;\r
+        }\r
+      }\r
+      if (msflen==0)\r
+        return null; // no minimal set\r
+      // compose minimal set\r
+      SequenceI[] mset = new SequenceI[msflen];\r
+      for (int i=0,j=sequences.length,k=0; i<j;i++) {\r
+        if (ungapped[i])\r
+          mset[k++] = sequences[i];\r
+      }\r
+      ungapped = null;\r
+      return mset;\r
   }\r
 }\r