AST-Based Formatting

Modern formatters parse → AST → format → output.

Each AST node type has its formatting rule. Trees compose: a binary op containing function calls renders both correctly.

Comments are tricky:

• AST drops them.
• Need: associate each comment with a "next" or "prev" AST node.
• During formatting, emit comments adjacent to their node.

Most formatters use CST (concrete syntax tree) instead, retaining comments + whitespace info.

Tree-sitter produces CST. Many newer formatters use it.

Pros of AST formatting:

• Robust: works regardless of input formatting.
• Can refactor: convert function expression to declaration, etc.

Cons:

• Loses original details (formatting hints, comment alignment).
• Bug in parser → formatter dies.

Hybrid: AST for the "real" code; preserve trivia (whitespace, comments) verbatim.

Example: prettier in JavaScript:

• Uses ESTree-style AST.
• Special-case comments: leading vs trailing.
• Preserves choice of ' vs " in strings (configurable).

Black: AST-based. Strict: removes optional parens, normalizes everything.