eomiks opened a new pull request, #8868:
URL: https://github.com/apache/hbase/pull/8868
Fixes `PrefixFilter#filterRowKey` comparing bytes past the end of a row that
is shorter than the prefix.
HBASE-28621 removed the row length check from `filterRowKey`, so the filter
compares `prefix.length` bytes starting at the row offset even when the row is
shorter than the prefix.
- When the row is followed by other bytes of the cell (family length,
family, ...), those bytes are compared as if they were part of the row. With
family `f` and rows `ab`, `ab\x00x`, `ab\x01f1`, `PrefixFilter("ab\x01f")`
returns `[ab, ab\x01f1]` instead of `[ab\x01f1]`, and `PrefixFilter("ab\x00")`
returns `[]` instead of `[ab\x00x]`.
- When the backing array holds nothing but the row, as in the fake cells
built from the scan start row, the comparison throws
`ArrayIndexOutOfBoundsException`. Before HBASE-27691 a scan with explicit
columns and a start row shorter than the prefix handed such a cell to the
filter.
`filterRowKey` now compares no more than the row itself. A row shorter than
the prefix that matches the prefix as far as it goes is before the prefix, so a
forward scan seeks to the prefix with the hint added by HBASE-28621 and a
reversed scan is done.
Tests:
- `TestPrefixFilter`: shorter rows in a row-only cell, a `KeyValue` and a
`ByteBufferKeyValue`. Five of the six new tests fail without the fix (two with
`ArrayIndexOutOfBoundsException`).
- `TestPrefixFilterShorterRow` (new, mini cluster): the two scans above fail
without the fix. The scans with explicit columns and a shorter start row pass
on master with or without the fix because of HBASE-27691; on a 2.6.6 based
branch without HBASE-27691 the forward ones fail with
`ArrayIndexOutOfBoundsException` before the fix and pass after it.
Details: https://issues.apache.org/jira/browse/HBASE-30537
--
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]
