Conversation
Split comma-separated rule codes in multi-rule fixtures (e.g. CP02,LT01) so detection comparison matches individual violation codes correctly.
… and AM08 Refine detection and autofix patches across multiple rules to close SQLFluff fixture gaps. Key changes: exclude trailing newlines from single-line comment protected ranges in fix engine, improve CV06 semicolon placement with block-comment awareness, add LT03/LT04 config-driven operator and comma positioning, make LT09/LT10 select-target autofix more precise, implement ST02 simple-case detection, and extend ST09 comparison-operator handling.
Tested against 85 production PostgreSQL files to align FlowScope detection with SQLFluff defaults. Key fixes: - CP01: exclude INTERVAL from keyword tracking; skip keywords after period in compound identifiers (e.g. wu.type) - CP05: remove WITH/ZONE from tracked type names (CTE false positives) - LT01: add FILTER/CONFLICT/BY to space-before-paren keyword list; recognise INSERT INTO target table before column-list paren; expand type-name list for bracket spacing (uuid, jsonb, etc.); allow ] followed by ) without space
CP01: Exclude function-like keywords (COALESCE, CAST, ANY, SOME, REPLACE) and TYPE from keyword capitalisation tracking — these are handled by CP03/CP05 in SQLFluff and TYPE is commonly used as a column name. LT09: Skip PostgreSQL DISTINCT ON (...) clause when finding the first select target, preventing false positives on multi-target SELECTs with DISTINCT ON. LT14: Restrict legacy (no-config) inconsistency check to top-level SELECT statements (paren_depth == 0) and top-level clause keywords, preventing false positives on compact EXISTS subqueries like SELECT 1 FROM t WHERE ...
Separate span resolution from autofix generation so violations always get proper source locations even when autofix is blocked (implicit column refs, comments in projection). Previously these fell back to line 1 with no span. Also extract positioned token building into a shared helper to reduce duplication.
ST06: Classify any top-level cast expression as band 1 (simple) to match SQLFluff's behavior where EXTRACT(...)::integer etc. are treated as simple regardless of inner complexity. Adds tests for WITH...INSERT INTO parsing (Statement::Query with SetExpr::Insert). RF01: Fix ORDER BY scope resolution — qualified references in ORDER BY were validated against inherited_sources (empty for top-level queries) instead of the body SELECT's FROM/JOIN scope. Adds order_by_scope_from_body() helper and regression tests. Net effect: ST06 over-reporting reduced +9→+4, RF01 eliminated +12→0.
Add ON CONFLICT / ON DUPLICATE KEY UPDATE support to both visit_expressions (for expression-based rules) and collect_identifier_candidates (for identifier-based rules like CP02). This enables detection of EXCLUDED.* references in PostgreSQL ON CONFLICT DO UPDATE SET clauses.
Switch capitalization rules (CP02, CP03, CP04, CP05) and reference consistency (RF03) from per-statement to per-identifier/per-reference issue emission to match SQLFluff's violation granularity. Add UPDATE/DELETE/subquery traversal to AL01 table alias detection. Parity improvements: AL01 -121→0, CP02 -77→+1, CP03 -18→+7, CP04 -3→0, RF03 -22→-13.
Add is_non_function_sql_keyword blocklist to prevent SQL clause keywords like IN, AS, OVER, USING, etc. from being misidentified as function calls in tokenizer-based CP03 function capitalization checks. CP03 parity: +7 → +1.
Fall back to ON expression span when autofix spans fail, and suppress issues where neither can be resolved to avoid misleading L1:1 reports.
SQLFluff flags only the first reversed join condition per SELECT/FROM clause, not all of them. Match this behaviour by gating detection in check_table_factor_joins with an issues_before counter and breaking out of the from_loop after the first hit. ST09 parity: +9 → 0
…QL UNNEST parity AL03: Remove in_set_rhs early return that suppressed all checks in UNION/INTERSECT/EXCEPT right-hand branches. SQLFluff still requires aliases on scalar literals there, so disable allow_scalar for RHS instead. (Δ: -4 → 0) RF02: Extend value table function recognition to PostgreSQL UNNEST so aliases are treated as column names rather than table references, matching SQLFluff behaviour. (Δ: +2 → 0)
…ype candidates ARRAY[] and DATE() are constructors/function calls, not type annotations. Including them in the type candidate pool created false mixed-case violations against lowercase types like `text`. Skip tracked type names followed by `[` (array constructor) or `(` (function call for types that don't take parenthesised precision). (Δ: +5 → 0)
…-keywords sqlparser's ALL_KEYWORDS includes function names (ROW_NUMBER, DATE_PART) and dialect-specific terms (METADATA, CHANNEL, STATUS, GENERATED) that SQLFluff does not treat as keywords. Exclude compound function names (containing underscores, except CURRENT_*/LOCAL_*/SESSION_*/SYSTEM_* pseudo-functions) and a small set of widely-used non-keyword identifiers from the keyword check. (Δ: +3 → -1)
Previously, all operator layout violations were aggregated into a single issue using the first violation's span. SQLFluff reports one violation per misplaced operator. Emit one issue per violation, each with its own span and autofix edits. (Δ: -1 → 0)
…overy Add rayon-based parallel file processing for both lint and fix passes, with a configurable --jobs flag and thread-safe progress bar. Directory expansion now respects .gitignore by default via the ignore crate, with --no-respect-gitignore to opt out.
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.
No description provided.