Filter reads whose alignment starts past the end of the contig to which it allegedly aligns.

git-svn-id: file:///humgen/gsa-scr1/gsa-engineering/svn_contents/trunk@1188 348d0f76-0448-11de-a6fe-93d51630548a

java/src/org/broadinstitute/sting/gatk/datasources/simpleDataSources/SAMDataSource.java

@@ -58,11 +58,6 @@ public class SAMDataSource implements SimpleDataSource {
     /** Backing support for reads. */
     private final Reads reads;
 
-    /**
-     * A histogram of exactly what reads were removed from the input stream and why.
-     */
-    private SAMReadViolationHistogram violations = new SAMReadViolationHistogram();
-
     /** our log, which we want to capture anything from this class */
     protected static Logger logger = Logger.getLogger(SAMDataSource.class);
 
@@ -87,7 +82,7 @@ public class SAMDataSource implements SimpleDataSource {
      * @return Histogram of reads.  Will not be null.
      */
     public SAMReadViolationHistogram getViolationHistogram() {
-        return violations;
+        return iteratorPool.getViolationHistogram();
     }
 
     /**
@@ -364,8 +359,6 @@ public class SAMDataSource implements SimpleDataSource {
                                                       Double downsamplingFraction,
                                                       Boolean filterZeroMappingQualityReads,
                                                       Boolean beSafeP) {
-        wrappedIterator = new MalformedSAMFilteringIterator(wrappedIterator,violations);
-
         // NOTE: this (and other filtering) should be done before on-the-fly sorting
         //  as there is no reason to sort something that we will end of throwing away
         if (downsamplingFraction != null)
@@ -398,6 +391,11 @@ class SAMIteratorPool extends ResourcePool<ReadStreamPointer, StingSAMIterator>
     /** Source information about the reads. */
     protected Reads reads;
 
+    /**
+     * A histogram of exactly what reads were removed from the input stream and why.
+     */
+    private SAMReadViolationHistogram violations = new SAMReadViolationHistogram();
+
     /** Is this a by-reads traversal or a by-locus? */
     protected boolean queryOverlapping;
 
@@ -422,6 +420,14 @@ class SAMIteratorPool extends ResourcePool<ReadStreamPointer, StingSAMIterator>
         return header;
     }
 
+    /**
+     * Returns a histogram of reads that were screened out, grouped by the nature of the error.
+     * @return Histogram of reads.  Will not be null.
+     */
+    public SAMReadViolationHistogram getViolationHistogram() {
+        return violations;
+    }
+
     protected ReadStreamPointer selectBestExistingResource( DataStreamSegment segment, List<ReadStreamPointer> pointers ) {
         for (ReadStreamPointer pointer : pointers) {
             if (pointer.canAccessSegmentEfficiently(segment)) {
@@ -446,7 +452,7 @@ class SAMIteratorPool extends ResourcePool<ReadStreamPointer, StingSAMIterator>
             iterator = streamPointer.getReadsOverlapping((MappedStreamSegment) segment);
         }
 
-        return new ReleasingIterator(iterator);
+        return new ReleasingIterator(new MalformedSAMFilteringIterator(header, iterator, violations));
     }
 
     protected void closeResource( ReadStreamPointer resource ) {

java/src/org/broadinstitute/sting/gatk/iterators/MalformedSAMFilteringIterator.java

@@ -25,6 +25,7 @@
 package org.broadinstitute.sting.gatk.iterators;
 
 import net.sf.samtools.SAMRecord;
+import net.sf.samtools.SAMFileHeader;
 import org.broadinstitute.sting.gatk.Reads;
 import org.broadinstitute.sting.utils.sam.SAMReadValidator;
 import org.broadinstitute.sting.utils.sam.SAMReadValidationException;
@@ -41,6 +42,11 @@ import java.util.NoSuchElementException;
  */
 
 public class MalformedSAMFilteringIterator implements StingSAMIterator {
+    /**
+     * The header to validate reads against.
+     */
+    private SAMFileHeader header = null;
+
     /**
      * The wrapped iterator.  Get reads from here.
      */
@@ -61,7 +67,8 @@ public class MalformedSAMFilteringIterator implements StingSAMIterator {
      * @param wrapped The wrapped iterator to use as backing data.
      * @param violations A structure to hold a breakdown of validator violations.
      */
-    public MalformedSAMFilteringIterator( StingSAMIterator wrapped, SAMReadViolationHistogram violations ) {
+    public MalformedSAMFilteringIterator( SAMFileHeader header, StingSAMIterator wrapped, SAMReadViolationHistogram violations ) {
+        this.header = header;
         this.wrapped = wrapped;
         this.violations = violations;
         seedNext();        
@@ -118,7 +125,7 @@ public class MalformedSAMFilteringIterator implements StingSAMIterator {
         while( wrapped.hasNext() && next == null ) {
             SAMRecord toTest = wrapped.next();
             try {
-                SAMReadValidator.validate(toTest);
+                SAMReadValidator.validate(header,toTest);
                 next = toTest;
             }
             catch ( SAMReadValidationException ex ) {

java/src/org/broadinstitute/sting/utils/sam/SAMReadValidator.java

@@ -25,6 +25,8 @@
 package org.broadinstitute.sting.utils.sam;
 
 import net.sf.samtools.SAMRecord;
+import net.sf.samtools.SAMFileHeader;
+import net.sf.samtools.SAMSequenceRecord;
 
 /**
  * Validates reads against a specific set of criteria.  If it finds a
@@ -41,9 +43,10 @@ public class SAMReadValidator {
      * Throw an exception if the read fails.
      * @param read the read to validate.  Must not be null.
      */
-    public static void validate( SAMRecord read ) throws SAMReadValidationException {
+    public static void validate( SAMFileHeader header, SAMRecord read ) throws SAMReadValidationException {
         checkInvalidAlignmentStart(read);
         checkInvalidAlignmentEnd(read);
+        checkAlignmentDisagreesWithHeader(header,read);
         checkCigarDisagreesWithAlignment(read);
     }
 
@@ -67,6 +70,13 @@ public class SAMReadValidator {
             throw new SAMReadValidationException("Alignment ends prior to its beginning");
     }
 
+    private static void checkAlignmentDisagreesWithHeader( SAMFileHeader header, SAMRecord read ) {
+        SAMSequenceRecord contigHeader = header.getSequence( read.getReferenceIndex() );
+        if( !read.getReadUnmappedFlag() && read.getAlignmentStart() > contigHeader.getSequenceLength() ) {
+            throw new SAMReadValidationException("Read is aligned to a point after the end of the contig");
+        }
+    }    
+
     /**
      * Check for inconsistencies between the cigar string and the 
      * @param read The read to validate.