flyrain commented on a change in pull request #3287:
URL: https://github.com/apache/iceberg/pull/3287#discussion_r733986727
##########
File path:
spark/v3.0/spark3/src/main/java/org/apache/iceberg/spark/data/vectorized/ColumnarBatchReader.java
##########
@@ -47,18 +68,57 @@ public final ColumnarBatch read(ColumnarBatch reuse, int
numRowsToRead) {
closeVectors();
}
+ Pair<int[], Integer> rowIdMapping = rowIdMapping(numRowsToRead);
+
for (int i = 0; i < readers.length; i += 1) {
vectorHolders[i] = readers[i].read(vectorHolders[i], numRowsToRead);
int numRowsInVector = vectorHolders[i].numValues();
Preconditions.checkState(
numRowsInVector == numRowsToRead,
"Number of rows in the vector %s didn't match expected %s ",
numRowsInVector,
numRowsToRead);
- arrowColumnVectors[i] =
- IcebergArrowColumnVector.forHolder(vectorHolders[i],
numRowsInVector);
+
+ if (rowIdMapping.second() == 0) {
+ arrowColumnVectors[i] =
IcebergArrowColumnVector.forHolder(vectorHolders[i], numRowsInVector);
+ } else {
+ arrowColumnVectors[i] =
ColumnVectorWithFilter.forHolder(vectorHolders[i], rowIdMapping.first(),
+ rowIdMapping.second());
+ }
}
+
+ rowStartPosInBatch += numRowsToRead;
ColumnarBatch batch = new ColumnarBatch(arrowColumnVectors);
- batch.setNumRows(numRowsToRead);
+
+ if (rowIdMapping.second() == 0) {
+ batch.setNumRows(numRowsToRead);
+ } else {
+ batch.setNumRows(rowIdMapping.second());
+ }
return batch;
}
+
+ private Pair<int[], Integer> rowIdMapping(int numRows) {
+ if (deletes != null && deletes.hasPosDeletes()) {
+ Roaring64Bitmap deletedRows = deletes.posDeletedRowIds();
+ if (deletedRows != null) {
+ return buildRowIdMapping(deletedRows, numRows);
+ }
+ }
+
+ return Pair.of(null, 0);
+ }
+
+ private Pair<int[], Integer> buildRowIdMapping(Roaring64Bitmap
deletedRowIds, int numRows) {
+ final int[] rowIdMapping = new int[numRows];
+ int originalRowId = 0;
+ int currentRowId = 0;
+ while (originalRowId < numRows) {
+ if (!deletedRowIds.contains(originalRowId + rowStartPosInBatch)) {
Review comment:
We can do that, but it doesn't help, and needs more cycles.
--
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]
