Syntex solves the problem of parsing source code by relying on Clang, but how does it parse Syntex queries themselves? Ideally, we would parse Syntex queries with Clang, then unify the parsed queries and parsed source code to identify matches. The deserialization process builds an index of tree node types (corresponding to grammar symbols) to deserialized nodes, which enables Syntex to quickly discover candidates that could match against a query’s root grammar symbol.