Per-conjunct pruning statistics for PruningPredicate#22235
Draft
adriangb wants to merge 2 commits into
Draft
Conversation
Introduce `OptionalFilterPhysicalExpr`, a transparent `PhysicalExpr` wrapper that marks a filter as *optional* — droppable without affecting query correctness. It delegates every `PhysicalExpr` method to the inner expression, so it is behavior-neutral until a consumer explicitly checks for the marker. This is the foundation for adaptive filter scheduling: a scan can detect the wrapper and drop a performance-hint filter (e.g. a hash-join dynamic filter) when it is not cost-effective, knowing correctness is enforced elsewhere. Also adds proto serialization (`PhysicalOptionalFilterNode`) so physical plans containing the wrapper round-trip faithfully. No caller wraps anything yet — that arrives with the adaptive parquet scan later in the stack. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Add an opt-in way to learn, per individual conjunct, how effective each predicate was during pruning — without running any extra pruning passes. - `PruningPredicate::try_new_tagged_conjuncts` builds a predicate from AND-conjuncts, each carrying a caller-supplied tag. - `PruningPredicate::prune_per_conjunct` returns the usual prune mask plus per-conjunct `PerConjunctPruneStats` (rows/containers seen vs. skipped) as a side effect of the pruning iteration that already runs. - `RowGroupAccessPlanFilter::prune_by_statistics_with_per_conjunct_stats` and `PagePruningAccessPlanFilter::prune_plan_with_per_conjunct_stats` surface those stats for row-group and page-index pruning respectively. The existing untagged `prune` / `prune_by_statistics` / `prune_plan_with_page_index` paths are preserved and unchanged; the new methods return empty stats on the untagged path. No in-tree caller uses the tagged path yet — the adaptive parquet scan consumes it later in the stack as a selectivity prior. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
github-actions
Bot
added
physical-expr
This was referenced May 15, 2026
|
Thank you for opening this pull request! Reviewer note: cargo-semver-checks reported the current version number is not SemVer-compatible with the changes in this pull request (compared against the base branch). Details |
Member
|
This would also be extremely useful for bootstrapping a cold stats in catalog via execution and runtime stats, I hope this feature lands! |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Which issue does this PR close?
Rationale for this change
Adaptive filter scheduling needs to know how effective each individual predicate conjunct was, so it can decide where to place that conjunct. Today pruning only reports an aggregate result. This PR surfaces per-conjunct effectiveness as a free side effect of the pruning pass that already runs — no extra passes.
What changes are included in this PR?
PruningPredicate::try_new_tagged_conjuncts— build a predicate from AND-conjuncts, each carrying a caller-supplied tag.PruningPredicate::prune_per_conjunct— returns the usual prune mask plus per-conjunctPerConjunctPruneStats.RowGroupAccessPlanFilter::prune_by_statistics_with_per_conjunct_statsandPagePruningAccessPlanFilter::prune_plan_with_per_conjunct_stats— surface those stats for row-group and page-index pruning.Existing untagged
prune/prune_by_statistics/prune_plan_with_page_indexpaths are preserved and unchanged. No in-tree caller uses the tagged path yet.Are these changes tested?
Yes — unit tests for the tagged-conjunct constructor and per-conjunct stat accounting.
Are there any user-facing changes?
New public API on
PruningPredicate. Purely additive; no behavior change.Stacked PR — diff is cumulative against
main. Review the top commit "feat: per-conjunct pruning statistics for PruningPredicate"; the commit below it is PR #22234.Stack (review/merge in order):