pingzh opened a new issue, #7142: URL: https://github.com/apache/datafusion-comet/issues/7142
### What is the problem the feature request solves? PR #7117 introduces a `make_date` evaluation guard, while the open #5533/#5972 stack introduces overlapping shared evaluation-mask machinery for other expressions. Consolidate these implementations without losing the consumers or safe native paths protected by either approach. The public configuration semantics and landing order need agreement before the overlapping changes merge. The implementation consolidation can be a separate follow-up to the correctness fix, ideally completed before release. Both proposed opt-outs are public configuration keys; once released, their scope and behavior must be preserved under the configuration compatibility policy. Mapping a date-only opt-out directly to a global switch would also disable protection for other expressions and is not a behavior-preserving rename. ### Describe the potential solution - Agree on configuration scope and landing order, including how operator-level opt-outs relate to expression-level safeguards. - Share expression enrollment, planner protection, and eager-child analysis while preserving each expression's behavior. Simply enrolling `MakeDate` in the existing shared pass is insufficient: carry over the additional join, sampling, `mapPartitions`, and `Dataset.rdd` consumers covered by #7117. - Preserve conditional-projection and filter evaluation, aggregates, limits, materialization boundaries, AQE, and reused plans. Reconcile whole-stage-codegen eligibility and filter input ordering using Spark's actual evaluation rules. - Retain a safe fast path for plans without enrolled expressions or existing protection tags, and run the combined regressions with native execution controls across the applicable Spark profiles. - Consolidate public configuration before release where possible. If a key has shipped, preserve its original scope through migration rather than broadening its meaning with an alias. ### Additional context Follow-up to [the review on #7117](https://github.com/apache/datafusion-comet/pull/7117#discussion_r4332013624). Related to #6006, which tracks the broader audit of data-dependent errors and evaluation masks. This issue specifically tracks consolidation of the overlapping implementations and public configuration in #7117, #5533, and #5972. -- 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]
