Draft
Conversation
…AND boundaries Single-element Parenthesized expressions like ((a = 1 OR b = 2)) are purely syntactic grouping, but break_predicate_at_and_boundaries didn't see through them. This prevented the optimizer from recognizing OR disjuncts inside parenthesized ON clauses, blocking multi-index OR optimization for queries like `JOIN t ON (a = x OR b = x)`. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Plumbing for CTE/subquery cardinality estimation. The field will be populated by the optimizer after join ordering and used by outer queries to estimate how many rows a subquery/CTE produces. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Previously multi-index OR (and AND-intersection) was only considered for single-table access (lhs.is_none()). This meant queries like `JOIN t ON (a = x OR b = x)` couldn't use the OR-by-union optimization during joins, falling back to full scans instead. Remove the lhs.is_none() guard and instead validate that each disjunct's constraints only reference tables already available on the LHS, ensuring cross-table disjuncts are properly handled. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When MultiIndexScan was selected as the access method, the output cardinality computation fell through to the generic full-scan formula, producing wildly inflated row count estimates. This caused the DP optimizer to avoid plans that use multi-index OR during joins. Propagate estimated_rows from cost functions through the MultiIndexScan params and use them directly in output cardinality computation. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The join optimizer now stores its output cardinality in SelectPlan's estimated_output_rows field (clamped to LIMIT when present). When an outer query references a CTE or subquery as a joined table, base_row_estimate reads this field instead of using the hardcoded fallback, enabling correct join ordering for queries like: WITH small AS (SELECT id FROM t LIMIT 3) SELECT ... FROM big_table JOIN small ON ... For CompoundSelect (UNION ALL, INTERSECT, etc.), branch estimates are combined according to set operation semantics. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Anonymized version of a real customer workload: a knowledge graph with TEXT UUID primary keys and CTE-driven neighbor lookups using OR-based multi-index scans. Tests verify both correct join ordering (CTE outer, multi-index OR on link and article tables) and result correctness with 2-seed and 5-seed variants. Verified against sqlite3 for result equivalence. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
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
1 participant
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.
yes TODO better description