Parsing & ASTs

Why parse first

• tokens are linear
• parser defines shape and meaning
• later phases rely on stable AST

Grammar contract

• precedence and associativity are encoded in recursive descent layers
• each layer consumes one precedence level

Left associativity pattern

• parse left branch
• while operator matches, parse right and fold
• produces (a-b)-c behavior for a-b-c

AST shape

• Number / Ident / Unary / Binary / Call / Grouping
• span metadata on nodes makes errors actionable

Error model

• expected token + actual token
• include location
• sync on ;, }, and EOF

Quality goals

• deterministic AST for the same source
• parser errors are clear and specific
• bad input should recover when possible

Quick checks

• 1 + 2 * 3 is postfix-style precedence
• f(1, a + 2) keeps arg order
• 1 + yields missing-expression diagnostic
• bad statement with recovery still parses subsequent statements