Compiler Compiler System with Isomorphic Runtime and Binary Structured Intermediate Representations
Aleksandr F. Urakhchin · Zenodo (CERN European Organization for Nuclear Research) · 2026
The Compiler Compiler System (CCS) is a compiler-construction tool that, given a context-free grammar specification, produces a complete compiler for the language that grammar describes. CCS departs from the YACC, GNU Bison, and ANTLR lineage in six specific ways: (i) it replaces the parse tree as central intermediate representation with a structured first-class model of *Context*, *Name*, *Symbol*, and *Rule* and their relationships, instantiated as a `SyntaxControlledRuntime` object during parsing; (ii) it strictly separates syntax processing from semantic processing, with semantics consuming a grammar-defined application programming interface rather than being embedded in grammar actions; (iii) it defines syntax processing as a two-phase, formally bidirectional procedure between the runtime and a serialized form, the `SyntaxControlledBinary`, so that runtime, binary, and source text are interconvertible; (iv) it makes formal de-compilation back to source text a first-class operation rather than a debugging aid; (v) it derives the runtime and binary application programming interfaces from the grammar itself rather than from hand-written interface definitions; and (vi) the source grammar definition language (SGDL) is self- bootstrapped — its own grammar is defined in itself, and the compiler-compiler executable can regenerate itself from that meta- grammar. We present an empirical evaluation in which a CCS- generated JSON parser, distributed as a flat tagged binary artifact, loads within approximately 2× of simdjson DOM and outperforms simdjson's per-leaf walk on every input in an O(1) grammar-derived walk implementation. The system has been exercised on the self-hosting round-trip oracle, on the JSON and XML target languages, and on integration with multiple AI-agent frameworks. This paper describes the parsing model, the runtime and binary forms, SGDL, the generator subsystem, the round-trip oracle, and the empirical results.