A context-free grammar has the form of
A = A1 | A2
A1 = B C
...
which looks very much like definitions of algebraic datatypes. This is by no coincidence - both formalisms reflect an underlying model with which we build complex concepts from constituents.
Given this correspondence, we could express a context-free grammar entirely as datatypes in a programming language. For example, in Java 17,
sealed interface A permits A1, A2{}
record A1(B b, C c) implements A{}
Such datatypes can be transliterated into a grammar, which is fed to a particular parser generator to build a parser. During the parsing process, constructors of datatypes are invoked, eventually outputting a parse tree in the very same datatypes.
PegParser<A> parser = PegParser.of(A.class);
A a = parser.matchFull(input);
Rekex implements this novel idea for Java 17, addressing issues of practical concerns, proving that it is an elegant approach to making parsers.
Read More:
Create by Zhong Yu. I am looking for a Java job; helps appreciated.