[GitHub] [spark] viirya commented on a change in pull request #34659: [SPARK-34863][SQL] Support complex types for Parquet vectorized reader

[GitBox <gi...@apache.org>]

viirya commented on a change in pull request #34659:
URL: https://github.com/apache/spark/pull/34659#discussion_r834685472



##########
File path: sql/core/src/main/java/org/apache/spark/sql/execution/datasources/parquet/ParquetColumnVector.java
##########
@@ -0,0 +1,379 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements.  See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License.  You may obtain a copy of the License at
+ *
+ *    http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.spark.sql.execution.datasources.parquet;
+
+import java.util.ArrayList;
+import java.util.List;
+import java.util.Set;
+
+import com.google.common.base.Preconditions;
+import org.apache.spark.memory.MemoryMode;

Review comment:
       ```suggestion
   import com.google.common.base.Preconditions;
   
   import org.apache.spark.memory.MemoryMode;
   ```

##########
File path: sql/core/src/main/java/org/apache/spark/sql/execution/vectorized/WritableColumnVector.java
##########
@@ -738,6 +774,12 @@ public WritableColumnVector arrayData() {
    */
   protected boolean isConstant;
 
+  /**
+   * True if this column only contains nulls. Comparing to 'isConstant' above, this doesn't require
+   * any allocation of space.

Review comment:
       ```suggestion
      * True if this column only contains nulls. This means the column values never change, even
      * across resets. Comparing to 'isConstant' above, this doesn't require
      * any allocation of space.
   ```
   
   Based on `reset` change.

##########
File path: sql/core/src/main/java/org/apache/spark/sql/execution/datasources/parquet/VectorizedRleValuesReader.java
##########
@@ -210,48 +213,358 @@ private void readBatchInternal(
       } else if (rowId > rangeEnd) {
         state.nextRange();
       } else {
-        // the range [rowId, rowId + n) overlaps with the current row range in state
+        // The range [rowId, rowId + n) overlaps with the current row range in state
         long start = Math.max(rangeStart, rowId);
         long end = Math.min(rangeEnd, rowId + n - 1);
 
-        // skip the part [rowId, start)
+        // Skip the part [rowId, start)
         int toSkip = (int) (start - rowId);
         if (toSkip > 0) {
           skipValues(toSkip, state, valueReader, updater);
           rowId += toSkip;
           leftInPage -= toSkip;
         }
 
-        // read the part [start, end]
+        // Read the part [start, end]
         n = (int) (end - start + 1);
 
         switch (mode) {
           case RLE:
             if (currentValue == state.maxDefinitionLevel) {
-              updater.readValues(n, offset, values, valueReader);
-            } else {
-              nulls.putNulls(offset, n);
+              updater.readValues(n, state.valueOffset, values, valueReader);
+              state.valueOffset += n;
+            } else if (!state.isRequired && currentValue == state.maxDefinitionLevel - 1) {
+              // Only add null if this represents a null element, but not for the case where a
+              // struct itself is null
+              nulls.putNulls(state.valueOffset, n);
+              state.valueOffset += n;
             }
+            defLevels.putInts(state.levelOffset, n, currentValue);
             break;
           case PACKED:
             for (int i = 0; i < n; ++i) {
-              if (currentBuffer[currentBufferIdx++] == state.maxDefinitionLevel) {
-                updater.readValue(offset + i, values, valueReader);
+              int value = currentBuffer[currentBufferIdx++];
+              if (value == state.maxDefinitionLevel) {
+                updater.readValue(state.valueOffset++, values, valueReader);
               } else {
-                nulls.putNull(offset + i);
+                // Only add null if this represents a null element, but not for the case where a
+                // struct itself is null
+                nulls.putNull(state.valueOffset++);
               }
+              defLevels.putInt(state.levelOffset + i, value);
             }
             break;
         }
-        offset += n;
+        state.levelOffset += n;

Review comment:
       why don't update `state.levelOffset` in place like `valueOffset`? Could be more consistent.




-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org