clintropolis commented on code in PR #12408:
URL: https://github.com/apache/druid/pull/12408#discussion_r865585085
##########
processing/src/main/java/org/apache/druid/segment/data/CompressionStrategy.java:
##########
@@ -344,6 +360,81 @@ public ByteBuffer compress(ByteBuffer in, ByteBuffer out)
}
}
+ public static class ZstdCompressor extends Compressor
+ {
+ private static final ZstdCompressor DEFAULT_COMPRESSOR = new
ZstdCompressor();
+
+ @Override
+ ByteBuffer allocateInBuffer(int inputSize, Closer closer)
+ {
+ ByteBuffer inBuffer = ByteBuffer.allocateDirect(inputSize);
+ closer.register(() -> ByteBufferUtils.free(inBuffer));
+ return inBuffer;
+ }
+
+ @Override
+ ByteBuffer allocateOutBuffer(int inputSize, Closer closer)
+ {
+ ByteBuffer outBuffer = ByteBuffer.allocateDirect((int)
Zstd.compressBound(inputSize));
+ closer.register(() -> ByteBufferUtils.free(outBuffer));
+ return outBuffer;
+ }
+
+ @Override
+ public ByteBuffer compress(ByteBuffer in, ByteBuffer out)
+ {
+ int position = in.position();
+ out.clear();
+ long sizeNeeded = Zstd.compressBound(in.remaining());
+ if (out.remaining() < sizeNeeded) {
+ throw new RuntimeException("Output buffer too small, please allocate
more space. " + sizeNeeded + " required.");
+ }
+ Zstd.compress(out, in, Zstd.maxCompressionLevel());
+ in.position(position);
+ out.flip();
+ return out;
+ }
+ }
+
+ public static class ZstdDecompressor implements Decompressor
+ {
+ private static final ZstdDecompressor DEFAULT_COMPRESSOR = new
ZstdDecompressor();
+
+ @Override
+ public void decompress(ByteBuffer in, int numBytes, ByteBuffer out)
+ {
+ out.clear();
+ // some tests don't use dbb's and zstd jni doesn't allow for non-dbb
byte buffers.
Review Comment:
It seems like it mostly would be replacing `ByteBuffer.wrap` with a direct
buffer, and adding an `@After` to release them, unless I'm missing something?
If you'd rather not modify the tests (though it seems like they would
probably be better tests using direct buffers since it more closely reflects
what would be happening in practice), I would be ok with instead modifying the
code to call `ByteBufferUtils.free` on the cloned direct buffers, so that in
case I missed something and we ever do use heap buffers in production code they
are at least not leaked.
--
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]
