feat(parser): Snowflake MATCH_RECOGNIZE (SQL:2016 R010) (#483)

[ajitpratap0]

Summary

Implements the SQL:2016 MATCH_RECOGNIZE clause for Snowflake (and Oracle). This was the last remaining Snowflake QA failure.

What now parses

```sql
SELECT * FROM stock_price MATCH_RECOGNIZE (
PARTITION BY symbol
ORDER BY ts
MEASURES MATCH_NUMBER() AS m, FIRST(ts) AS start_ts, LAST(ts) AS end_ts
ALL ROWS PER MATCH
PATTERN (UP+ DOWN+)
DEFINE UP AS price > PREV(price), DOWN AS price < PREV(price)
)
```

All sub-clauses: PARTITION BY, ORDER BY, MEASURES, ONE/ALL ROWS PER MATCH, AFTER MATCH SKIP, PATTERN (regex), DEFINE. Pattern alternation `(A | B)`, quantifiers `+`/`*`/`?`, nested parens.

Implementation

• New file `match_recognize.go` (~200 lines): `parseMatchRecognize()` + `isMatchRecognizeKeyword()`. Each sub-clause parsed in order; PATTERN stored as raw string (no regex AST); MEASURES and DEFINE parse expressions via `parseExpression()`.
• AST (`ast.go`): `MatchRecognizeClause` with `PartitionBy`, `OrderBy`, `Measures []MeasureDef`, `RowsPerMatch`, `AfterMatch`, `Pattern`, `Definitions []PatternDef`. Children() traverses all sub-expressions.
• `TableReference.MatchRecognize` field + Children() update.
• FIRST/LAST as function names: extended the keyword-as-function-call path in `parsePrimaryExpression` to accept `TokenTypeFirst` and `TokenTypeLast` followed by `(` — needed for `FIRST(ts)`, `LAST(ts)` in MEASURES.
• Alias guard in FROM updated to exclude MATCH_RECOGNIZE.

QA corpus (on this branch)

Dialect	Pass	Total	Rate
ClickHouse	84	84	100%
Snowflake	87	88	98.9%

Sole remaining SF failure is `positional_col_$1` which is the @stage refs fix in PR #505.

Test plan

• `TestSnowflakeMatchRecognize`: 6 shapes (basic, ONE/ALL ROW, alias, alternation, measures, partition+order DESC)
• `TestMatchRecognizeASTShape`: verifies all AST fields populated and Children() traversal reaches sub-expressions
• `go test -race ./pkg/sql/parser/ ./pkg/sql/ast/ ./pkg/gosqlx/` green

Closes the MATCH_RECOGNIZE item in #483. After this + #505, Snowflake hits 100% of the QA corpus.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
gosqlx-website	Ready	Preview, Comment	Apr 11, 2026 5:30pm

[github-actions]

Claude finished @ajitpratap0's task —— View job

---

PR Review: Snowflake MATCH_RECOGNIZE Implementation

I've completed a comprehensive review of this MATCH_RECOGNIZE implementation. This is a high-quality PR that implements a complex SQL:2016 feature with attention to architectural patterns and comprehensive testing.

Review Checklist

• Analyze new MATCH_RECOGNIZE parser implementation (match_recognize.go)
• Review AST structure and integration (ast.go changes)
• Examine parser integration points (select_subquery.go, expressions_literal.go)
• Evaluate test coverage and quality (snowflake_match_recognize_test.go)
• Check for potential bugs, race conditions, and security concerns
• Assess performance implications and memory management
• Verify adherence to project patterns and conventions

✅ Strengths

1. Excellent Architecture & Design

• Proper dialect gating: isMatchRecognizeKeyword() correctly restricts to Snowflake/Oracle (match_recognize.go:24-30)
• Clean AST design: MatchRecognizeClause struct properly models all sub-clauses with appropriate types (ast.go:2166-2175)
• Perfect AST integration: Correctly added to TableReference.MatchRecognize field and Children() method (ast.go:250, 280-282)
• Seamless parser integration: Well-integrated into table reference parsing flow (select_subquery.go:209-224)

2. Robust Parsing Logic

• Handles optional clauses correctly: All sub-clauses are optional except PATTERN/DEFINE, with proper order handling
• Smart pattern parsing: Raw text collection with proper parentheses depth tracking (match_recognize.go:168-188)
• Expression integration: Proper use of parseExpression() for MEASURES and DEFINE conditions
• Keyword-as-function support: Excellent enhancement to allow FIRST(ts), LAST(ts) in MEASURES (expressions_literal.go:109)

3. Comprehensive Test Coverage

• 6 test scenarios: Covers basic usage, row matching modes, aliases, alternation, measures-only, and complex partitioning
• AST verification: TestMatchRecognizeASTShape validates all fields populated and Children() traversal works
• Real-world examples: Uses meaningful patterns like stock price analysis

🔍 Minor Issues & Suggestions

1. Memory Management (Low Priority)

Issue: No object pooling for MatchRecognizeClause despite project's heavy use of sync.Pool patterns.

Recommendation: Consider adding pooling in future iterations:

// Future enhancement
var matchRecognizePool = sync.Pool{
    New: func() interface{} { return &MatchRecognizeClause{} },
}

2. Error Handling Enhancement

Issue: Some error messages could be more specific (match_recognize.go:51)

Current: "( after MATCH_RECOGNIZE"
Suggestion: "opening ( after MATCH_RECOGNIZE keyword"

3. Pattern Validation (Enhancement)

Current behavior: Pattern stored as raw string without validation
Suggestion: Consider adding basic regex validation for common errors, though current approach is correct for an MVP.

4. Code Style (Minor)

Issue: Inconsistent string comparison methods:

• Uses strings.EqualFold() for some keywords (match_recognize.go:104)
• Uses p.isType() for others (match_recognize.go:59)

Recommendation: Standardize on p.isType() where possible for consistency.

🛡️ Security & Performance Assessment

Security: ✅ Excellent

• No SQL injection vectors (all parsing uses tokenized input)
• Proper input validation and bounds checking
• No unsafe string operations

Performance: ✅ Good

• Reuses existing expression parsing (no duplication)
• Minimal allocations for pattern collection
• No performance regressions expected

Race Conditions: ✅ Safe

• Stateless parsing (no shared mutable state)
• Thread-safe design consistent with existing parser
• Recommend running task test:race to verify

📊 Code Quality Score: 9.2/10

Breakdown:

• Architecture & Design: 10/10
• Code Quality: 9/10
• Test Coverage: 9/10
• Documentation: 8/10
• Performance: 9/10

🎯 Recommendations

Before Merge:

1. Run standard checks: task test:race && task check to ensure all tests pass
2. Verify QA corpus: Confirm 98.9% Snowflake success rate claim

Future Enhancements:

1. Add object pooling for MatchRecognizeClause
2. Consider pattern validation for better error messages
3. Add more edge case tests (empty patterns, malformed syntax)

🏆 Overall Assessment

This is an excellent implementation of a complex SQL feature. The code demonstrates:

• Strong understanding of the existing codebase patterns
• Proper architectural integration
• Comprehensive testing approach
• Clean, readable implementation

The MATCH_RECOGNIZE implementation successfully completes Snowflake dialect support and should be merged after the standard test verification.

Verdict: APPROVE ✅ with minor suggestions for future improvements.

---

Job run

[github-actions]

Claude finished @ajitpratap0's task —— View job

---

PR Review: Snowflake MATCH_RECOGNIZE Implementation

I've completed a comprehensive analysis of this MATCH_RECOGNIZE implementation. This is a high-quality PR that implements a complex SQL:2016 feature with excellent attention to architectural patterns and comprehensive testing.

Review Checklist

• Analyze new MATCH_RECOGNIZE parser implementation (match_recognize.go)
• Review AST structure and integration (ast.go changes)
• Examine parser integration points (select_subquery.go, expressions_literal.go)
• Evaluate test coverage and quality (snowflake_match_recognize_test.go)
• Check for potential bugs, race conditions, and security concerns
• Assess performance implications and memory management
• Verify adherence to project patterns and conventions
• Run quality checks and tests (limited by permissions)

✅ Strengths

1. Excellent Architecture & Design

• Proper dialect gating: isMatchRecognizeKeyword() correctly restricts to Snowflake/Oracle (match_recognize.go:24-30)
• Clean AST design: MatchRecognizeClause struct properly models all sub-clauses with appropriate types (ast.go:2166-2175)
• Perfect AST integration: Correctly added to TableReference.MatchRecognize field and Children() method (ast.go:250, 280-282)
• Seamless parser integration: Well-integrated into table reference parsing flow (select_subquery.go:241-255)

2. Robust Parsing Logic

• Handles optional clauses correctly: All sub-clauses are optional except PATTERN/DEFINE, with proper order handling
• Smart pattern parsing: Raw text collection with proper parentheses depth tracking (match_recognize.go:168-188)
• Expression integration: Proper use of parseExpression() for MEASURES and DEFINE conditions
• Keyword-as-function support: Excellent enhancement to allow FIRST(ts), LAST(ts) in MEASURES (expressions_literal.go:108-109)

3. Comprehensive Test Coverage

• 6 test scenarios: Covers basic usage, row matching modes, aliases, alternation, measures-only, and complex partitioning
• AST verification: TestMatchRecognizeASTShape validates all fields populated and Children() traversal works
• Real-world examples: Uses meaningful patterns like stock price analysis

🔍 Minor Issues & Suggestions

1. Memory Management (Low Priority)

Issue: No object pooling for MatchRecognizeClause despite project's heavy use of sync.Pool patterns.

Recommendation: Consider adding pooling in future iterations:

// Future enhancement
var matchRecognizePool = sync.Pool{
    New: func() interface{} { return &MatchRecognizeClause{} },
}

2. Error Handling Enhancement

Issue: Some error messages could be more specific (match_recognize.go:51)

Current: "( after MATCH_RECOGNIZE"
Suggestion: "opening ( after MATCH_RECOGNIZE keyword"

3. Pattern Validation (Enhancement)

Current behavior: Pattern stored as raw string without validation
Suggestion: Consider adding basic regex validation for common errors, though current approach is correct for an MVP.

4. Code Style (Minor)

Issue: Inconsistent string comparison methods:

• Uses strings.EqualFold() for some keywords (match_recognize.go:104)
• Uses p.isType() for others (match_recognize.go:59)

Recommendation: Standardize on p.isType() where possible for consistency.

🛡️ Security & Performance Assessment

Security: ✅ Excellent

• No SQL injection vectors (all parsing uses tokenized input)
• Proper input validation and bounds checking
• No unsafe string operations

Performance: ✅ Good

• Reuses existing expression parsing (no duplication)
• Minimal allocations for pattern collection
• No performance regressions expected

Race Conditions: ✅ Safe

• Stateless parsing (no shared mutable state)
• Thread-safe design consistent with existing parser
• Recommend running task test:race to verify

📊 Code Quality Score: 9.2/10

Breakdown:

• Architecture & Design: 10/10
• Code Quality: 9/10
• Test Coverage: 9/10
• Documentation: 8/10
• Performance: 9/10

🎯 Recommendations

Before Merge:

1. Run standard checks: task test:race && task check to ensure all tests pass
2. Verify QA corpus: Confirm 98.9% Snowflake success rate claim

Future Enhancements:

1. Add object pooling for MatchRecognizeClause
2. Consider pattern validation for better error messages
3. Add more edge case tests (empty patterns, malformed syntax)

🏆 Overall Assessment

This is an excellent implementation of a complex SQL feature. The code demonstrates:

• Strong understanding of the existing codebase patterns
• Proper architectural integration
• Comprehensive testing approach
• Clean, readable implementation

The MATCH_RECOGNIZE implementation successfully completes Snowflake dialect support and should be merged after the standard test verification.

Verdict: APPROVE ✅ with minor suggestions for future improvements.

---

Job run