kazuyukitanimura commented on a change in pull request #34611:
URL: https://github.com/apache/spark/pull/34611#discussion_r758204579
##########
File path:
sql/core/src/main/java/org/apache/spark/sql/execution/datasources/parquet/VectorizedPlainValuesReader.java
##########
@@ -53,19 +53,50 @@ public void skip() {
throw new UnsupportedOperationException();
}
+ private void updateCurrentByte() {
+ try {
+ currentByte = (byte) in.read();
+ } catch (IOException e) {
+ throw new ParquetDecodingException("Failed to read a byte", e);
+ }
+ }
+
@Override
public final void readBooleans(int total, WritableColumnVector c, int rowId)
{
- // TODO: properly vectorize this
- for (int i = 0; i < total; i++) {
- c.putBoolean(rowId + i, readBoolean());
+ int i = 0;
+ if (bitOffset > 0) {
+ i = Math.min(8 - bitOffset, total);
+ c.putBooleans(rowId, i, currentByte, bitOffset);
+ bitOffset = (bitOffset + i) & 7;
+ }
+ for (; i + 7 < total; i += 8) {
+ updateCurrentByte();
+ c.putBooleans(rowId + i, currentByte);
+ }
+ if (i < total) {
+ updateCurrentByte();
+ bitOffset = total - i;
+ c.putBooleans(rowId + i, bitOffset, currentByte, 0);
}
}
@Override
public final void skipBooleans(int total) {
- // TODO: properly vectorize this
- for (int i = 0; i < total; i++) {
- readBoolean();
+ // Using >>3 instead of /8 below. The difference is important when
(total-(8-bitOffset))<0.
+ // E.g. (-1)>>3=(-1) vs. (-1)/8=0. The latter incorrectly enters the
if(numBytesToSkip>=0){.
+ // (total-(8-bitOffset))<0 means there will be still unread bits left in
the currentByte.
+ // In that case, updateCurrentByte() should not be called.
+ int numBytesToSkip = (total - (8 - bitOffset)) >> 3;
+ bitOffset = (bitOffset + total) & 7;
+ if (numBytesToSkip >= 0) {
Review comment:
Thank you @sunchao for finding this! I missed it. Since this logic is
error prone and I didn't like the long comments, I changed the method to follow
the same structure as `readBooleans()`. I have also updated the test case to
cover this case.
--
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]
