phylotastic hackathon at NESCENT 120606

[jalview.git] / forester / java / src / org / forester / util / SequenceIdParser.java

diff --git a/forester/java/src/org/forester/util/SequenceIdParser.java b/forester/java/src/org/forester/util/SequenceIdParser.java
index 8def260..d75a6dd 100644 (file)
--- a/forester/java/src/org/forester/util/SequenceIdParser.java
+++ b/forester/java/src/org/forester/util/SequenceIdParser.java
@@ -32,7 +32,6 @@ import java.util.regex.Matcher;
 import java.util.regex.Pattern;\r
 \r
 import org.forester.phylogeny.data.Identifier;\r
-import org.forester.ws.uniprot.DatabaseTools;\r
 \r
 public final class SequenceIdParser {\r
 \r
@@ -58,8 +57,14 @@ public final class SequenceIdParser {
                                                                          .compile( "(?:\\A|.*[^a-zA-Z0-9])([A-Z]{2}\\d{6})(?:[^a-zA-Z0-9]|\\Z)" );\r
     private final static Pattern GENBANK_PROTEIN_AC_PATTERN      = Pattern\r
                                                                          .compile( "(?:\\A|.*[^a-zA-Z0-9])([A-Z]{3}\\d{5})(?:[^a-zA-Z0-9]|\\Z)" );\r
-    private final static boolean DEBUG                           = true;\r
+   \r
+    // RefSeq accession numbers can be distinguished from GenBank accessions \r
+    // by their distinct prefix format of 2 characters followed by an\r
+    // underscore character ('_'). For example, a RefSeq protein accession is NP_015325. \r
+    private final static Pattern REFSEQ_PATTERN      = Pattern\r
+    .compile( "(?:\\A|.*[^a-zA-Z0-9])([A-Z]{2}_\\d{6,})(?:[^a-zA-Z0-9]|\\Z)" );\r
 \r
+   \r
     \r
     /**\r
      * Returns null if no match.\r
@@ -68,11 +73,12 @@ public final class SequenceIdParser {
     public final static Identifier parse( final String s ) {\r
         String v = parseGenbankAccessor( s );\r
         if ( !ForesterUtil.isEmpty( v ) ) {\r
-            return new Identifier( v, "ncbi" );\r
+\r
+            return new Identifier( v, Identifier.NCBI );\r
         }\r
         v = parseRefSeqAccessor( s );\r
         if ( !ForesterUtil.isEmpty( v ) ) {\r
-            return new Identifier( v, "ncbi" );\r
+            return new Identifier( v, Identifier.REFSEQ );\r
         }\r
         return null;\r
     }\r
@@ -80,11 +86,8 @@ public final class SequenceIdParser {
     /**\r
      * Returns null if no match.\r
      * \r
-     * @param query\r
-     * @param db \r
-     * @return\r
      */\r
-    static public String parseGenbankAccessor( final String query ) {\r
+    public static String parseGenbankAccessor( final String query ) {\r
         Matcher m = GENBANK_NUCLEOTIDE_AC_PATTERN_1.matcher( query );\r
         if ( m.lookingAt() ) {\r
             return m.group( 1 );\r
@@ -106,26 +109,16 @@ public final class SequenceIdParser {
         }\r
     }\r
     \r
-    public final static String parseRefSeqAccessor( final String query ) {\r
-        Matcher m = GENBANK_NUCLEOTIDE_AC_PATTERN_1.matcher( query );\r
+    /**\r
+     * Returns null if no match.\r
+     * \r
+     */\r
+    private final static String parseRefSeqAccessor( final String query ) {\r
+        Matcher m = REFSEQ_PATTERN.matcher( query );\r
         if ( m.lookingAt() ) {\r
             return m.group( 1 );\r
         }\r
-        else {\r
-            m = GENBANK_NUCLEOTIDE_AC_PATTERN_2.matcher( query );\r
-            if ( m.lookingAt() ) {\r
-                return m.group( 1 );\r
-            }\r
-            else {\r
-                m = GENBANK_PROTEIN_AC_PATTERN.matcher( query );\r
-                if ( m.lookingAt() ) {\r
-                    return m.group( 1 );\r
-                }\r
-                else {\r
-                    return null;\r
-                }\r
-            }\r
-        }\r
+        return null;\r
     }\r
     \r
     \r