divijvaidya commented on code in PR #13135:
URL: https://github.com/apache/kafka/pull/13135#discussion_r1083831592
##########
clients/src/main/java/org/apache/kafka/common/record/DefaultRecordBatch.java:
##########
@@ -273,20 +272,32 @@ public int partitionLeaderEpoch() {
public DataInputStream recordInputStream(BufferSupplier bufferSupplier) {
final ByteBuffer buffer = this.buffer.duplicate();
buffer.position(RECORDS_OFFSET);
- return new DataInputStream(compressionType().wrapForInput(buffer,
magic(), bufferSupplier));
+ final InputStream decompressedStream =
compressionType().wrapForInput(buffer, magic(), bufferSupplier);
+ return decompressedStream instanceof DataInputStream ?
(DataInputStream) decompressedStream : new DataInputStream(decompressedStream);
}
private CloseableIterator<Record> compressedIterator(BufferSupplier
bufferSupplier, boolean skipKeyValue) {
final DataInputStream inputStream = recordInputStream(bufferSupplier);
if (skipKeyValue) {
// this buffer is used to skip length delimited fields like key,
value, headers
- byte[] skipArray = new byte[MAX_SKIP_BUFFER_SIZE];
+ final ByteBuffer skipBuffer =
bufferSupplier.get(compressionType().getRecommendedDOutSize());
Review Comment:
I have mistakenly added that I have changed it to the description which I
will fix.
I would like to make this (skipping) a separate code change because I think
that it is a performance regression since it has an extra buffer allocation.
This is because zstd-jni implementation of skip() (de)allocates a "skip"
buffer (from the buffer pool) for skipping [1]. Alternatively, in current
implementation, we read all data in the same 16KB output buffer (which is
allocated only once). In both cases, the amount of data copy from native to
Java is same. The only difference is whether we read & skip in our code or we
read & skip in zstd-jni code. Pushing skip to zstd-jni would be beneficial when
it further pushed it down to native layer.
[1]
https://github.com/luben/zstd-jni/blob/master/src/main/java/com/github/luben/zstd/ZstdInputStreamNoFinalizer.java#L228
--
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: jira-unsubscr...@kafka.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
