So a weekend toy project.

Hand-coded parser-lexers are far more powerful because their diagnostics are useful and are much easier to maintain.

Lex/yacc / flex/bison products by contrast have awful diagnostics and become steaming piles of confusion that have to be rewritten.

PHP is Flex/Bison.

The main point of the discussed project is to alter data structure to be more CPU cache friendly and thus try to be faster than C++. The paradigm.