[
https://issues.apache.org/jira/browse/IGNITE-25327?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Vladislav Pyatkov reassigned IGNITE-25327:
------------------------------------------
Assignee: Vladislav Pyatkov
> Avoid copy bytes in tuple comparator
> ------------------------------------
>
> Key: IGNITE-25327
> URL: https://issues.apache.org/jira/browse/IGNITE-25327
> Project: Ignite
> Issue Type: Improvement
> Reporter: Vladislav Pyatkov
> Assignee: Vladislav Pyatkov
> Priority: Major
> Labels: ignite-3
> Time Spent: 10m
> Remaining Estimate: 0h
>
> h3. Motivation
> We copy bytes from ByteBuffer each time when comparing any field in a tuple.
> The base is
> {code:title=BinaryTupleParser}
> private byte[] getBytes(int begin, int end) {
> byte[] bytes = new byte[end - begin];
> buffer.duplicate().position(begin).limit(end).get(bytes);
> return bytes;
> }
> {code}
> where we do a copy from the byte buffer. Then we compare fields by full
> deserialization value
> {{BinaryTupleComparatorUtils#compareFieldValue}}.
> We can avoid copying if we can compare the field data without deserialization
> in ByteBuffer directly using offset.
> h3. Implementation notes
> For extracting bytes from off-heap memory, use GridUnsafe. It ought to be
> more productive than using DirectByteBuffer.
> Fix it for {{BinaryTupleComparator}} and {{PartialBinaryTupleMatcher}}.
> h3. Definition of done
> Do a comparison without deserialization for all types for which it is
> advisable.
> Binary comparison is certainly useful for {{STRING}} and {{BYTES}}.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
