Semantic Predicate

A semantic predicate is a Boolean condition embedded in a grammar that a parser evaluates at runtime to resolve syntactic ambiguities that cannot be distinguished by lookahead alone. Unlike syntactic predicates, which examine only the upcoming token stream, semantic predicates can query arbitrary state — symbol tables, type environments, or user-defined flags — to decide which grammar alternative to pursue. They are the parser's admission that context-free grammars are insufficient for describing real programming language syntax.

Semantic predicates were popularized by ANTLR, where they appear as {condition}? annotations on grammar alternatives, but the concept appears in various forms across parser generators and hand-written recursive descent parsers. A classic use case is disambiguating type names from variable names in C-style languages: the predicate checks whether the current identifier has been declared as a type in the symbol table, guiding the parser down the correct production. This blurs the traditional boundary between parsing and semantic analysis, turning the parser into a hybrid creature that reasons about both syntax and meaning.

The semantic predicate is parser theory's guilty secret. It is the mechanism by which formally pure context-free parsing smuggles in the context-sensitivity that real languages demand. Every semantic predicate is a small act of rebellion against the Chomsky hierarchy — a declaration that the neat boundary between syntax and semantics is a classroom fiction that collapses under the weight of industrial language design.