Ocaml target

[bernardro]

Hi, I would like to start work on an Ocaml target. I have reviewed the page on new Antlr target creation. Besides that does anyone have any ideas, advice, links, etc?

[ericvergnaud]

Hey, great to hear that. We have a requirement that all targets are compatible ie. the same grammar behaves the same. To help with this, we have a strong focus on CI, with the exact same tests running across all targets. So a good sequence could be as follows: - create a test target (similar to the existing ones for other targets) - implement CI scripts for CircleCI Linux and/or GitHub Mac and/or GitHub Windows - start with simple lexer tests, then complex lexer tests - continue with simple parser tests Give up any form of social life, family life or mental health. Feel free to ask questions, don’t feel offended if you don’t get a response quickly. That said, welcome to the community ! Eric

…

Le 23 juin 2022 à 18:51, Bernard Okoth ***@***.***> a écrit : Hi, I would like to start work on an Ocaml target. I have reviewed the page on new Antlr target creation. Besides that does anyone have any ideas, advice, links, etc? — Reply to this email directly, view it on GitHub <#3760>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAZNQJAPB4ZT7YGMCQMT6ZTVQSIYPANCNFSM5ZVAUQBA>. You are receiving this because you are subscribed to this thread.

[KvanTTT]

I'd move CI scripts to the end, they aren't required if other items are not completed.

Creating a new test target is very easy with the new test infrastructure, especially after #3752

[KvanTTT]

Hi, I'm not sure @parrt will integrate new targets, see #3736 (comment) Also I'm not sure OCaml is in high demand.

[parrt]

Hi, yeah, I don't think it makes sense to add this target to the main repository although you are of course free to work on one for your own use.

[bernardro]

Thanks for the ideas and warnings. My approach would be to avoid reinventing the wheel if possible - so I may need to suffer multiple passes through multiple tools to get a stable working end product. Still sifting through the open source offerings for these kinds of tasks though - and there is a ton.

Another one that caught my eye is the sitter one and Facebook's evolution of it. It does trigger mixed feelings because its AI.. and we don't have explicit up-front control over how it makes decisions. Fallacies aside, it might be the way to go to iterate and test a more formally verified tool. Antlr is battle tested however, especially for the early grammars and targets.

As for maintenance - I wonder if there is a way to break up and restructure the Antlr repo so that owners of targets and grammars are responsible for keeping their repos, grammars, etc. up to date. I am new to Antlr so these are just my rumblings at this point.

[parrt]

We originally had to decide between multiple repositories and a single one. We might've even tried separate ones. We currently do this for php and it complicates our lives for testing etc..

[KvanTTT]

Maybe it makes sense to use submodules, but a single repository is more maintainable indeed.