add convenience methods for finding rows by name

gemma-core/src/main/java/ubic/gemma/core/datastructure/matrix/BaseExpressionDataMatrix.java

@@ -34,6 +34,7 @@
 import javax.annotation.Nullable;
 import java.io.Serializable;
 import java.util.*;
+import java.util.stream.Collectors;
 
 /**
  * Base class for ExpressionDataMatrix implementations.
@@ -129,6 +130,14 @@ public BioAssayDimension getBestBioAssayDimension() {
 
     }
 
+    @Override
+    public Collection<Integer> findRowsByName( String name ) {
+        return rowElementMap.entrySet().stream()
+                .filter( entry -> entry.getKey().getName().equals( name ) )
+                .map( entry -> entry.getValue() )
+                .collect( Collectors.toList() );
+    }
+
     @Override
     public BioAssayDimension getBioAssayDimension( CompositeSequence designElement ) {
 
@@ -258,7 +267,7 @@ void addToRowMaps( Integer row, CompositeSequence designElement ) {
      * For example, in the following diagram "-" indicates a biomaterial, while "*" indicates a bioassay. Each row of
      * "*" indicates samples run on a different microarray design (a different bio assay material). In the examples we
      * assume there is just a single biomaterial dimension.
-     * 
+     *
      * <pre>
      * ---------------
      * *****              -- only a few samples run on this platform
@@ -268,7 +277,7 @@ void addToRowMaps( Integer row, CompositeSequence designElement ) {
      * <p>
      * A simpler case:
      * </p>
-     * 
+     *
      * <pre>
      * ---------------
      * ***************
@@ -278,23 +287,23 @@ void addToRowMaps( Integer row, CompositeSequence designElement ) {
      * <p>
      * A more typical and easy case (one microarray design used):
      * </p>
-     * 
+     *
      * <pre>
      * ----------------
      * ****************
      * </pre>
      * <p>
      * If every sample was run on two different array designs:
      * </p>
-     * 
+     *
      * <pre>
      * ----------------
      * ****************
      * ****************
      * </pre>
      * <p>
      * Every sample was run on a different array design:
-     * 
+     *
      * <pre>
      * -----------------------
      * ******

gemma-core/src/main/java/ubic/gemma/core/datastructure/matrix/EmptyExpressionMatrix.java

@@ -83,6 +83,11 @@ public Object[] getColumn( BioAssay bioAssay ) {
         throw new UnsupportedOperationException();
     }
 
+    @Override
+    public Object[][] getRows( Collection<Integer> indices ) {
+        throw new UnsupportedOperationException();
+    }
+
     @Override
     public Object[] getColumn( Integer column ) {
         throw new UnsupportedOperationException();

gemma-core/src/main/java/ubic/gemma/core/datastructure/matrix/ExpressionDataBooleanMatrix.java

@@ -62,6 +62,18 @@ public ExpressionDataBooleanMatrix( Collection<? extends DesignElementDataVector
         this.vectorsToMatrix( selectedVectors );
     }
 
+    @Override
+    public Boolean[][] getRows( Collection<Integer> indices ) {
+        if ( indices == null || indices.isEmpty() ) {
+            return null;
+        }
+
+        return indices.stream()
+                .map( index -> getRow( index ) )
+                .toArray( Boolean[][]::new );
+
+    }
+
     @Override
     public int columns() {
         return matrix.columns();

gemma-core/src/main/java/ubic/gemma/core/datastructure/matrix/ExpressionDataDoubleMatrix.java

@@ -356,6 +356,19 @@ public Double[] getRow( Integer index ) {
         return ArrayUtils.toObject( rawRow );
     }
 
+
+    @Override
+    public Double[][] getRows( Collection<Integer> indices ) {
+        if ( indices == null || indices.isEmpty() ) {
+            return null;
+        }
+
+        return indices.stream()
+                .map( index -> getRow( index ) )
+                .toArray( Double[][]::new );
+
+    }
+
     @Override
     public Double[][] getRows( List<CompositeSequence> designElements ) {
         if ( designElements == null ) {

gemma-core/src/main/java/ubic/gemma/core/datastructure/matrix/ExpressionDataDoubleMatrixUtil.java

@@ -94,8 +94,16 @@ public static ExpressionDataDoubleMatrix filterAndLog2Transform( ExpressionDataD
             if ( dmatrix.rows() < r ) {
                 ExpressionDataDoubleMatrixUtil.log.info( ( r - dmatrix.rows() ) + " rows removed due to too many identical values" );
             }
-        }
 
+            /*
+             * As noted in https://github.com/PavlidisLab/GemmaCuration/issues/218, this filter can still do things
+             * that are counter-intuitive because we are using quantile-normalization. In some cases very highly expressed genes
+             * will be assigned multiple identical values, though they weren't identical in the raw data.
+             *
+             * Fixing this "once and for all" might mean making the filter less stringent at very high expression levels (like the top 1% of genes)
+             * or changing the approach.
+             */
+        }
         return dmatrix;
 
     }

gemma-core/src/main/java/ubic/gemma/core/datastructure/matrix/ExpressionDataIntegerMatrix.java

@@ -62,6 +62,19 @@ public Integer[][] get( List<CompositeSequence> designElements, List<BioAssay> b
         throw new UnsupportedOperationException();
     }
 
+
+    @Override
+    public Integer[][] getRows( Collection<Integer> indices ) {
+        if ( indices == null || indices.isEmpty() ) {
+            return null;
+        }
+
+        return indices.stream()
+                .map( index -> getRow( index ) )
+                .toArray( Integer[][]::new );
+
+    }
+
     @Override
     public Integer[] getColumn( BioAssay bioAssay ) {
         int index = this.columnAssayMap.get( bioAssay );

gemma-core/src/main/java/ubic/gemma/core/datastructure/matrix/ExpressionDataMatrix.java

@@ -220,6 +220,23 @@ public interface ExpressionDataMatrix<T> {
      */
     T[][] getRows( List<CompositeSequence> designElements );
 
+    /**
+     * Access a submatrix
+     * @param rowIndices of integers to select
+     * @return T[][] or null if rowIndices is null or empty.
+     */
+    T[][] getRows( Collection<Integer> rowIndices);
+
+
+    /**
+     * Convenience function to locate the indices with an (exact match) CompositeSequence.name
+     *
+     * @param name The CompositeSequence name to look for
+     * @return array of row indices matching (usually will just be one value)
+     */
+    Collection<Integer> findRowsByName(String name);
+
+
     /**
      * @return true if any values are null or NaN (for Doubles); all other values are considered non-missing.
      */

gemma-core/src/main/java/ubic/gemma/core/datastructure/matrix/ExpressionDataStringMatrix.java

@@ -87,6 +87,20 @@ public String[][] get( List<CompositeSequence> designElements, List<BioAssay> bi
         throw new UnsupportedOperationException();
     }
 
+
+    @Override
+    public String[][] getRows( Collection<Integer> indices ) {
+        if ( indices == null || indices.isEmpty() ) {
+            return null;
+        }
+
+        return indices.stream()
+                .map( index -> getRow( index ) )
+                .toArray( String[][]::new );
+
+    }
+
+
     @Override
     public String[] getColumn( BioAssay bioAssay ) {
         int index = this.columnAssayMap.get( bioAssay );

gemma-core/src/test/java/ubic/gemma/core/datastructure/matrix/ExpressionDataDoubleMatrixTest.java

@@ -170,6 +170,8 @@ public void testConstructExpressionDataDoubleMatrix() {
             log.debug( aRow );
         }
 
+        assertEquals(2, expressionDataDoubleMatrix.getRows( Arrays.asList( new Integer[]{1,2} )).length );
+
         Double[][] dMatrix = expressionDataDoubleMatrix.getRawMatrix();
         assertEquals( dMatrix.length, 200 );
         assertEquals( dMatrix[0].length, 59 );