From: Jim Procter <jprocter@issues.jalview.org>
Date: Mon, 29 Aug 2016 21:20:23 +0000 (+0100)
Subject: JAL-2154 don’t synthesise multiple CDS|<Acc> sequences when one is already available... 
X-Git-Tag: Release_2_10_0~47^2~4^2~32
X-Git-Url: http://source.jalview.org/gitweb/?a=commitdiff_plain;ds=sidebyside;h=4bf04707b9bb755a7cdaa829eeec02ab87b4a228;hp=-c;p=jalview.git

JAL-2154 don’t synthesise multiple CDS|<Acc> sequences when one is already available on dataset (fixes crossRefs2XML failure - CDS|ENST00000288602 found duplicated
"Pass (0,0,0): before start of pass3: ENSEMBL ENSG00000157764 -> UNIPROT{0}:”)
---

4bf04707b9bb755a7cdaa829eeec02ab87b4a228
diff --git a/src/jalview/analysis/AlignmentUtils.java b/src/jalview/analysis/AlignmentUtils.java
index e0ec22b..b0a2269 100644
--- a/src/jalview/analysis/AlignmentUtils.java
+++ b/src/jalview/analysis/AlignmentUtils.java
@@ -1680,15 +1680,20 @@ public class AlignmentUtils
            * didn't find mapped CDS sequence - construct it and add
            * its dataset sequence to the dataset
            */
-          cdsSeq = makeCdsSequence(dnaSeq.getDatasetSequence(), aMapping);
+          cdsSeq = makeCdsSequence(dnaSeq.getDatasetSequence(), aMapping,
+                  dataset).deriveSequence();
           // cdsSeq has a name constructed as CDS|<dbref>
           // <dbref> will be either the accession for the coding sequence,
           // marked in the /via/ dbref to the protein product accession
           // or it will be the original nucleotide accession.
-          SequenceI cdsSeqDss = cdsSeq.createDatasetSequence();
+          SequenceI cdsSeqDss = cdsSeq.getDatasetSequence();
+
           cdsSeqs.add(cdsSeq);
+
           if (!dataset.getSequences().contains(cdsSeqDss))
           {
+            // check if this sequence is a newly created one
+            // so needs adding to the dataset
             dataset.addSequence(cdsSeqDss);
           }
 
@@ -1870,9 +1875,14 @@ public class AlignmentUtils
    * 
    * @param seq
    * @param mapping
+   * @param dataset
+   *          - existing dataset. We check for sequences that look like the CDS
+   *          we are about to construct, if one exists already, then we will
+   *          just return that one.
    * @return CDS sequence (as a dataset sequence)
    */
-  static SequenceI makeCdsSequence(SequenceI seq, Mapping mapping)
+  static SequenceI makeCdsSequence(SequenceI seq, Mapping mapping,
+          AlignmentI dataset)
   {
     char[] seqChars = seq.getSequence();
     List<int[]> fromRanges = mapping.getMap().getFromRanges();
@@ -1907,6 +1917,40 @@ public class AlignmentUtils
     String mapFromId = mapping.getMappedFromId();
     String seqId = "CDS|" + (mapFromId != null ? mapFromId : seq.getName());
     SequenceI newSeq = new Sequence(seqId, newSeqChars, 1, newPos);
+    if (dataset != null)
+    {
+      SequenceI[] matches = dataset.findSequenceMatch(newSeq.getName());
+      if (matches != null)
+      {
+        boolean matched = false;
+        for (SequenceI mtch : matches)
+        {
+          if (mtch.getStart() != newSeq.getStart())
+          {
+            continue;
+          }
+          if (mtch.getEnd() != newSeq.getEnd())
+          {
+            continue;
+          }
+          if (!Arrays.equals(mtch.getSequence(), newSeq.getSequence()))
+          {
+            continue;
+          }
+          if (!matched)
+          {
+            matched = true;
+            newSeq = mtch;
+          }
+          else
+          {
+            System.err
+                    .println("JAL-2154 regression: warning - found (and ignnored a duplicate CDS sequence):"
+                            + mtch.toString());
+          }
+        }
+      }
+    }
     // newSeq.setDescription(mapFromId);
 
     return newSeq;