viirya commented on code in PR #36675:
URL: https://github.com/apache/spark/pull/36675#discussion_r884468660
##########
sql/core/src/main/java/org/apache/spark/sql/execution/datasources/parquet/VectorizedParquetRecordReader.java:
##########
@@ -270,13 +271,40 @@ private void initBatch(
vectors[i + partitionIdx].setIsConstant();
}
}
+
+ // For Parquet tables whose columns have associated DEFAULT values, this
reader must return
+ // those values instead of NULL when the corresponding columns are not
present in storage (i.e.
+ // belong to the 'missingColumns' field in this class).
+ ColumnVector[] finalColumns = new
ColumnVector[sparkSchema.fields().length];
+ for (int i = 0; i < columnVectors.length; i++) {
+ Object defaultValue = sparkRequestedSchema.existenceDefaultValues()[i];
+ if (defaultValue == null) {
+ finalColumns[i] = vectors[i];
+ } else {
+ WritableColumnVector writable;
+ if (memMode == MemoryMode.OFF_HEAP) {
+ writable = new OffHeapColumnVector(capacity, vectors[i].dataType());
+ } else {
+ writable = new OnHeapColumnVector(capacity, vectors[i].dataType());
+ }
+ Optional<Integer> appended = writable.appendObjects(capacity,
defaultValue);
+ if (!appended.isPresent()) {
+ throw new IOException("Cannot assign default column value to result
column batch in " +
+ "vectorized Parquet reader because the data type is not supported:
" + defaultValue);
+ }
Review Comment:
If `isPresent` is false, throwing exception here, but in Orc reader it
silently ignores that and continues to pull nulls. Why the difference?
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
