Add JNI methods for detecting and purging non-empty nulls from LIST a…

…nd STRUCT (#12742)

This PR adds methods for detecting and purging non-empty nulls.

Authors:
  - Raza Jafri (https://github.com/razajafri)
  - Nghia Truong (https://github.com/ttnghia)
  - AJ Schmidt (https://github.com/ajschmidt8)

Approvers:
  - Nghia Truong (https://github.com/ttnghia)

URL: #12742

java/src/main/java/ai/rapids/cudf/ColumnView.java

@@ -4639,6 +4639,10 @@ static native long makeCudfColumnView(int type, int scale, long data, long dataS
 
   static native long applyBooleanMask(long arrayColumnView, long booleanMaskHandle) throws CudfException;
 
+  static native boolean hasNonEmptyNulls(long handle) throws CudfException;
+
+  static native long purgeNonEmptyNulls(long handle) throws CudfException;
+
   /**
    * A utility class to create column vector like objects without refcounts and other APIs when
    * creating the device side vector from host side nested vectors. Eventually this can go away or
@@ -4997,4 +5001,37 @@ public HostColumnVector copyToHost() {
       }
     }
   }
+
+  /**
+   * Exact check if a column or its descendants have non-empty null rows
+   *
+   * @return Whether the column or its descendants have non-empty null rows
+   */
+  public boolean hasNonEmptyNulls() {
+    return hasNonEmptyNulls(viewHandle);
+  }
+
+  /**
+   * Copies this column into output while purging any non-empty null rows in the column or its
+   * descendants.
+   *
+   * If this column is not of compound type (LIST/STRING/STRUCT/DICTIONARY), the output will be
+   * the same as input.
+   *
+   * The purge operation only applies directly to LIST and STRING columns, but it applies indirectly
+   * to STRUCT/DICTIONARY columns as well, since these columns may have child columns that
+   * are LIST or STRING.
+   *
+   * Examples:
+   * lists = data: [{{0,1}, {2,3}, {4,5}} validity: {true, false, true}]
+   * lists[1] is null, but the list's child column still stores `{2,3}`.
+   *
+   * After purging the contents of the list's null rows, the column's contents will be:
+   * lists = [data: {{0,1}, {4,5}} validity: {true, false, true}]
+   *
+   * @return A new column with equivalent contents to `input`, but with null rows purged
+   */
+  public ColumnVector purgeNonEmptyNulls() {
+    return new ColumnVector(purgeNonEmptyNulls(viewHandle));
+  }
 }

java/src/main/native/src/ColumnViewJni.cpp

@@ -2457,4 +2457,26 @@ JNIEXPORT jlong JNICALL Java_ai_rapids_cudf_ColumnView_applyBooleanMask(
   CATCH_STD(env, 0);
 }
 
+JNIEXPORT jboolean JNICALL
+Java_ai_rapids_cudf_ColumnView_hasNonEmptyNulls(JNIEnv *env, jclass, jlong column_view_handle) {
+  JNI_NULL_CHECK(env, column_view_handle, "column_view handle is null", 0);
+  try {
+    cudf::jni::auto_set_device(env);
+    auto const *cv = reinterpret_cast<cudf::column_view const *>(column_view_handle);
+    return cudf::has_nonempty_nulls(*cv);
+  }
+  CATCH_STD(env, 0);
+}
+
+JNIEXPORT jlong JNICALL
+Java_ai_rapids_cudf_ColumnView_purgeNonEmptyNulls(JNIEnv *env, jclass, jlong column_view_handle) {
+  JNI_NULL_CHECK(env, column_view_handle, "column_view handle is null", 0);
+  try {
+    cudf::jni::auto_set_device(env);
+    auto const *cv = reinterpret_cast<cudf::column_view const *>(column_view_handle);
+    return release_as_jlong(cudf::purge_nonempty_nulls(*cv));
+  }
+  CATCH_STD(env, 0);
+}
+
 } // extern "C"

java/src/test/java/ai/rapids/cudf/ColumnVectorTest.java

@@ -32,6 +32,7 @@
 import java.util.Arrays;
 import java.util.Collections;
 import java.util.List;
+import java.util.Optional;
 import java.util.function.Supplier;
 import java.util.stream.Collectors;
 import java.util.stream.IntStream;
@@ -6691,4 +6692,65 @@ void testApplyBooleanMaskFromListOfStructure() {
       assertColumnsAreEqual(expectedCv, actualCv);
     }
   }
+
+  /**
+   * The caller needs to make sure to close the returned ColumnView
+   */
+  private ColumnView[] getColumnViewWithNonEmptyNulls() {
+    List<Integer> list0 = Arrays.asList(1, 2, 3);
+    List<Integer> list1 = Arrays.asList(4, 5, null);
+    List<Integer> list2 = Arrays.asList(7, 8, 9);
+    List<Integer> list3 = null;
+    ColumnVector input = makeListsColumn(DType.INT32, list0, list1, list2, list3);
+    // Modify the validity buffer
+    BaseDeviceMemoryBuffer dmb = input.getDeviceBufferFor(BufferType.VALIDITY);
+    try (HostMemoryBuffer newValidity = HostMemoryBuffer.allocate(64)) {
+      newValidity.copyFromDeviceBuffer(dmb);
+      BitVectorHelper.setNullAt(newValidity, 1);
+      dmb.copyFromHostBuffer(newValidity);
+    }
+    try (HostColumnVector hostColumnVector = input.copyToHost()) {
+      assert (hostColumnVector.isNull(1));
+      assert (hostColumnVector.isNull(3));
+    }
+    try (ColumnVector expectedOffsetsBeforePurge = ColumnVector.fromInts(0, 3, 6, 9, 9)) {
+      ColumnView offsetsCvBeforePurge = input.getListOffsetsView();
+      assertColumnsAreEqual(expectedOffsetsBeforePurge, offsetsCvBeforePurge);
+    }
+    ColumnView colWithNonEmptyNulls = new ColumnView(input.type, input.rows, Optional.of(2L), dmb,
+        input.getDeviceBufferFor(BufferType.OFFSET), input.getChildColumnViews());
+    assertEquals(2, colWithNonEmptyNulls.nullCount);
+    return new ColumnView[]{input, colWithNonEmptyNulls};
+  }
+
+  @Test
+  void testPurgeNonEmptyNullsList() {
+    ColumnView[] values = getColumnViewWithNonEmptyNulls();
+    try (ColumnView colWithNonEmptyNulls = values[1];
+         ColumnView input = values[0];
+         // purge non-empty nulls
+         ColumnView colWithEmptyNulls = colWithNonEmptyNulls.purgeNonEmptyNulls();
+         ColumnVector expectedOffsetsAfterPurge = ColumnVector.fromInts(0, 3, 3, 6, 6);
+         ColumnView offsetsCvAfterPurge = colWithEmptyNulls.getListOffsetsView()) {
+      assertTrue(colWithNonEmptyNulls.hasNonEmptyNulls());
+      assertColumnsAreEqual(expectedOffsetsAfterPurge, offsetsCvAfterPurge);
+      assertFalse(colWithEmptyNulls.hasNonEmptyNulls());
+    }
+  }
+
+  @Test
+  void testPurgeNonEmptyNullsStruct() {
+    ColumnView[] values = getColumnViewWithNonEmptyNulls();
+    try (ColumnView listCol = values[1];
+         ColumnView input = values[0];
+         ColumnView stringsCol = ColumnVector.fromStrings("A", "col", "of", "Strings");
+         ColumnView structView = ColumnView.makeStructView(stringsCol, listCol);
+         ColumnView structWithEmptyNulls = structView.purgeNonEmptyNulls();
+         ColumnView newListChild = structWithEmptyNulls.getChildColumnView(1);
+         ColumnVector expectedOffsetsAfterPurge = ColumnVector.fromInts(0, 3, 3, 6, 6);
+         ColumnView offsetsCvAfterPurge = newListChild.getListOffsetsView()) {
+      assertColumnsAreEqual(expectedOffsetsAfterPurge, offsetsCvAfterPurge);
+      assertFalse(newListChild.hasNonEmptyNulls());
+    }
+  }
 }