Hey Stephan,

Thanks for raising this! In general, `CONTRIBUTING.md` serves as an
encouragement, but the reality is always case by case. It's not uncommon to
have a 500+ LOC PR if the feature or fix absolutely needs to touch a lot of
code.

In your case, I'd suggest creating the +781LOC PR as-is to showcase the
overall solution. And you can later break it down into smaller PRs if the
reviewer finds it necessary.

Looking forward to your fix!

Best,
Shawn

On Mon, Aug 3, 2026 at 9:25 AM Stephan Berger <[email protected]> wrote:

> Hey iceberg-rust folks,
>
>
> I just filed Issue 2950 (
> https://github.com/apache/iceberg-rust/issues/2950) regarding scaling of
> equality delete file application.
>
> There is already a PR aiming to fix the scaling issue by implementing a
> similar approach to Java, but the original authors' organization (@t3hw)
> has since moved away from iceberg-rust.
>
> I took in inspiration from that approach but rebuild the PR on current
> main and using a `RowFilter` with `ArrowPredicateFn`.
>
> I should state that this PR does not include PR #2873 (fix: avoid
> lost-wakeup hang in get_equality_delete_predicate_for_delete_file_path) and
> PR #2630 (fix(reader): avoid deadlock when loading equality deletes). Both
> are approved by @viirya, so hoping those PRs will land soon I'd rather
> rebase my changes on top of those, but I was thinking to file the PR for
> review anyways.
>
> The changes amount to a diff of +781/-172 (a lot of the added lines are
> tests) and the `CONTRIBUTING.md`says:
>
> > We generally discourage large pull requests that are over 300-500 lines
> of diff.
>
> I don't think this PR could be split up into smaller PRs, so I am asking
> for guidance on how to proceed.
>
>
> Best regards
> Stephan

Reply via email to