sqool

ANTLR-based SQL parser for Java 25 targeting MySQL, PostgreSQL, Oracle, and SQLite.

Build baseline

• Build tool: Gradle
• Target runtime: Java 25
• Current status: Milestone 5 (cross-dialect stabilization) complete; all four dialects (MySQL, SQLite, PostgreSQL, Oracle) have stable parser implementations with cross-dialect conformance tests.

Current parser coverage

• MySQL
  • Real upstream grammar vendored from antlr/grammars-v4 (see sqool-grammar-mysql/UPSTREAM.md).
  • Parser facade with SLL-first parsing and LL fallback; parser metrics via ParseMetrics.
  • Normalized AST: CREATE TABLE, CREATE DATABASE, DELETE, DROP TABLE/DROP DATABASE, derived tables, GROUP BY, HAVING, INSERT/REPLACE, SELECT with joins/aliases/expressions, SHOW statements, TRUNCATE, UNION/UNION ALL, UPDATE, script mode, and more. Unsupported statement kinds fall back to MySqlRawStatement.
  • Conformance and regression suite in sqool-conformance; JMH benchmark in sqool-bench (vs JSqlParser).
• SQLite
  • Grammar vendored in sqool-grammar-sqlite (see UPSTREAM.md). Parser in sqool-dialect-sqlite with SLL/LL and ParseMetrics.
  • Normalized AST for core SELECT (simple queries without complex expressions), and raw-statement fallback for all other constructs. Cross-dialect conformance tests alongside MySQL, PostgreSQL, and Oracle. JMH benchmark in sqool-bench.
• PostgreSQL
  • Grammar vendored in sqool-grammar-postgresql (see UPSTREAM.md). Parser in sqool-dialect-postgresql with SLL/LL and ParseMetrics.
  • Normalized AST: SELECT (including single-table and JOIN queries), with raw-statement fallback for INSERT, UPDATE, DELETE, CREATE TABLE, DROP TABLE, and transaction control. Conformance and JMH benchmark present.
• Oracle
  • SQL-only v1 subset: SELECT (including JOINs), INSERT, UPDATE, DELETE, CREATE TABLE, DROP TABLE, TRUNCATE, transaction control (COMMIT, ROLLBACK, SAVEPOINT). PL/SQL procedural constructs (anonymous blocks, stored procedures, triggers) are explicitly out of scope for v1.
  • Grammar written for Oracle SQL v1 in sqool-grammar-oracle. Uses Oracle-specific types (NUMBER, VARCHAR2, NVARCHAR2, CLOB, BLOB, etc.) and constructs (ROWNUM, ROWID, DUAL, SYSDATE, FETCH FIRST n ROWS ONLY, MINUS set operator).
  • Parser in sqool-dialect-oracle with SLL/LL fallback pattern and ParseMetrics. Normalized AST: SELECT (single-table and JOIN queries); all other statements fall back to OracleRawStatement. Conformance suite in sqool-conformance and JMH benchmark in sqool-bench.
  • See Oracle Risk and Scope for scope limitations and risk mitigation details.

See Dialect Coverage Matrix for a full breakdown of supported constructs and normalization level per dialect. Usage examples (parsing each dialect, script mode, inspecting ParseSuccess/ParseFailure) are in the Dialect Coverage Matrix – Usage examples section.

Common commands

./gradlew build
./gradlew verifyBootstrap
./gradlew :sqool-bench:jmh

See Benchmarks for how to run and capture baseline results.

Design documents

• SQL Parser Technical Design
• SQL Parser High-Level Implementation Checklist
• Dialect Coverage Matrix
• Diagnostics Behavior
• Benchmarks
• Milestone 0 Bootstrap Backlog
• Milestone 0 Review Checklist
• Milestone 4 Oracle Backlog
• Milestone 4 Review Checklist
• Oracle Risk and Scope
• Milestone 5 Cross-Dialect Backlog
• Milestone 5 Review Checklist
• Post-M5 Roadmap and Follow-ups
• Next Milestone Checklist
• Parser Abstractions (design note)
• Release Readiness
• Milestone 6: Remaining Implementation Plan
• Upstream Grammar Refresh Process