The worst part of designing a language is the parsing stage.

Simple enough to do it by hand, but there’s a lot of boilerplate and bureaucracy involved that is painfully time-wasting unless you know exactly what syntax you are going for.

But if you adopt a parser-generator such as Flex/Bison you’ll find yourself learning and debugging and obtuse language that has to be forcefully bent to your needs, and I hope your knowledge of parsing theory is up-to-scratch when you’re facing with shift-reduce conflicts or have to decide whether LR or LALR(1) or whatever is most appropriate to your syntax.

Not even PEG is gonna come to your rescue.