JAL-2112 optimisation of EMBL CDS parsing

diff --git a/src/jalview/datamodel/xdb/embl/EmblEntry.java b/src/jalview/datamodel/xdb/embl/EmblEntry.java
index 9a07c36..f8c0bbe 100644 (file)
--- a/src/jalview/datamodel/xdb/embl/EmblEntry.java
+++ b/src/jalview/datamodel/xdb/embl/EmblEntry.java
@@ -197,7 +197,7 @@ public class EmblEntry
     // dbref
     retrievedref.setMap(new Mapping(null, new int[] { 1, dna.getLength() },
             new int[] { 1, dna.getLength() }, 1, 1));
-    // TODO: transform EMBL Database refs to canonical form
+
     if (dbRefs != null)
     {
       for (DBRefEntry dbref : dbRefs)
@@ -206,6 +206,7 @@ public class EmblEntry
       }
     }
 
+    SequenceIdMatcher matcher = new SequenceIdMatcher(peptides);
     try
     {
       for (EmblFeature feature : features)
@@ -219,7 +220,7 @@ public class EmblEntry
         }
         if (FeatureProperties.isCodingFeature(sourceDb, feature.getName()))
         {
-          parseCodingFeature(feature, sourceDb, dna, peptides);
+          parseCodingFeature(feature, sourceDb, dna, peptides, matcher);
         }
       }
     } catch (Exception e)
@@ -249,7 +250,7 @@ public class EmblEntry
    *          list of protein product sequences for Embl entry
    */
   void parseCodingFeature(EmblFeature feature, String sourceDb,
-          SequenceI dna, List<SequenceI> peptides)
+          SequenceI dna, List<SequenceI> peptides, SequenceIdMatcher matcher)
   {
     boolean isEmblCdna = sourceDb.equals(DBRefSource.EMBLCDS);
 
@@ -259,7 +260,6 @@ public class EmblEntry
     String prname = "";
     String prid = null;
     Map<String, String> vals = new Hashtable<String, String>();
-    SequenceIdMatcher matcher = new SequenceIdMatcher(peptides);
 
     /*
      * codon_start 1/2/3 in EMBL corresponds to phase 0/1/2 in CDS
@@ -283,13 +283,13 @@ public class EmblEntry
         }
         else if (qname.equals("protein_id"))
         {
-          prid = q.getValues()[0];
+          prid = q.getValues()[0].trim();
         }
         else if (qname.equals("codon_start"))
         {
           try
           {
-            codonStart = Integer.parseInt(q.getValues()[0]);
+            codonStart = Integer.parseInt(q.getValues()[0].trim());
           } catch (NumberFormatException e)
           {
             System.err.println("Invalid codon_start in XML for "
@@ -299,7 +299,7 @@ public class EmblEntry
         else if (qname.equals("product"))
         {
           // sometimes name is returned e.g. for V00488
-          prname = q.getValues()[0];
+          prname = q.getValues()[0].trim();
         }
         else
         {

diff --git a/test/jalview/datamodel/xdb/embl/EmblEntryTest.java b/test/jalview/datamodel/xdb/embl/EmblEntryTest.java
index e8760bd..3de5e3f 100644 (file)
--- a/test/jalview/datamodel/xdb/embl/EmblEntryTest.java
+++ b/test/jalview/datamodel/xdb/embl/EmblEntryTest.java
@@ -3,6 +3,7 @@ package jalview.datamodel.xdb.embl;
 import static org.testng.AssertJUnit.assertEquals;
 import static org.testng.AssertJUnit.assertSame;
 
+import jalview.analysis.SequenceIdMatcher;
 import jalview.datamodel.DBRefEntry;
 import jalview.datamodel.Sequence;
 import jalview.datamodel.SequenceI;
@@ -37,6 +38,7 @@ public class EmblEntryTest
     // not the whole sequence but enough for this test...
     SequenceI dna = new Sequence("J03321", "GGATCCGTAAGTTAGACGAAATT");
     List<SequenceI> peptides = new ArrayList<SequenceI>();
+    SequenceIdMatcher matcher = new SequenceIdMatcher(peptides);
     EmblFile ef = EmblTestHelper.getEmblFile();
 
     /*
@@ -48,7 +50,7 @@ public class EmblEntryTest
     {
       if ("CDS".equals(feature.getName()))
       {
-        testee.parseCodingFeature(feature, "EMBL", dna, peptides);
+        testee.parseCodingFeature(feature, "EMBL", dna, peptides, matcher);
       }
     }