Parquet: Backport variant BINARY upper bound fix (#16880) to 1.9.x - #17355
Parquet: Backport variant BINARY upper bound fix (#16880) to 1.9.x#17355vishnuprakaz wants to merge 1 commit into
Conversation
* Parquet: Fix variant BINARY upper bound to truncate up * Parquet: Add test for variant BINARY upper bound overflow (cherry picked from commit 470f464)
|
Thank you so much for doing all three backports, really appreciate the effort here. I've merged the 1.11 and 1.10 ones. For 1.9, I have to apologize: I should have checked before asking you to include it. The 1.9.x branch hasn't received any commits since July 3, 2025, and there's been no 1.9 release since 1.9.2 (Jul 16, 2025), So the line is effectively inactive. Given that, I'm going to close this one. Sorry for the extra work on the 1.9 backport. Thanks again for the backports! |
|
No worries at all, @huaxingao . Thanks for merging the 1.10 and 1.11 backports and for the reviews! Separately, while in this code I found a sibling issue (same class as #16880), variant STRING bounds are compared in UTF-16 order instead of UTF-8, which can invert bounds and skip files. Opened #17397 with a fix if you'd like to take a look. |
Backports #16880 to the 1.9.x release branch.
Fixes a silent wrong-results bug where the variant shredded BINARY upper
bound truncated down instead of up, so file-skipping stats could exclude
files that actually contain matching data.
(cherry picked from commit 470f464)