Is it possible to perform side effects while parsing?

[softdevteam]

Is it possible to perform side effects while parsing?

[ltratt]

Honest answer: I don't remember. I think "no" but I know I looked at this at some point, and I don't remember what I settled upon.

That said, in general I'd say that in LR parsing it's better to bubble results up then force mutation downwards. Flattening things (as in your example) is a common idiom https://softdevteam.github.io/grmtools/latest_release/book/parsing_idioms.html#define-a-flatten-function -- would that work?

[loewenheim]

Yes, I think that would probably work. It has the minor flaw of allocating lots of individual vectors which are then concatenated, but I have no actual reason to think that it would pose a performance problem in practice.

[ltratt]

Informally, my experience is that unless you have very large inputs (or a very very large grammar), parsing performance tends to be a minor concern on modern machines. If however you do encounter problems, please shout!