clickhouse-sql-parser AGENTS.md

AI Agents Guidline

This document outlines guidance for AI coding agent including project structure, coding style, testing, and contribution practices for the ClickHouse SQL Parser project.

Development Commands

# Build the CLI binary
make

# Run tests
make test

# Update golden fixtures after intentional output changes
make update_test

After editing code, use goimports and gofmt to maintain code style, and run make lint to check for any issues before committing or requesting a review.

Project Structure & Module Organization

• main.go is the CLI entry point (clickhouse-sql-parser) for AST output and SQL formatting.
• parser/ contains core parser code: lexer (lexer.go), AST definitions (ast.go), traversal helpers (walk.go), and grammar-specific parser files (parser_query.go, parser_table.go, parser_alter.go, etc.).
• Tests live next to source as *_test.go files, with fixtures under parser/testdata/.
• Fixture groups are organized by SQL type (basic/, query/, dml/, ddl/), with generated expectations in output/ (AST JSON) and format/ (formatted SQL).

Coding Style & Naming Conventions

• Use Go 1.21 conventions (go.mod) and keep code gofmt/goimports clean (enforced by lint).
• Naming is the most important style aspect, try you best to choose a clear and descriptive name for variables, functions, types, and files. For example, use parseSelect for a function that parses a SELECT statement, and SelectStatement for the corresponding AST node type.
• Place parsing logic in the matching module by statement family (for example, query parsing in parser/parser_query.go).
• Follow existing parser naming patterns such as parseXxx helpers and explicit AST type names.
• Keep AST FormatSQL() output deterministic; formatting changes must be reflected in golden files.
• You must go through the repository before adding new code to ensure consistency with existing patterns and styles. If you are unsure about where to place new code or how to format it, please refer to the existing codebase or ask for guidance.
• Reusing existing code and patterns is encouraged to maintain consistency and reduce redundancy. If you find a similar function or pattern in the codebase, consider adapting it for your needs instead of creating something new from scratch.