[Feature Request]: C target

[davidfetter]

Antlr3 had one, and it'd be awesome if Antlr4 did, too.

[AmatanHead]

I'm currently working with c++ target and there are some feature requests that I'd like to see in C target:

• parsing trees should be traversable without visitors. It means contexts should store alt number by default;
• for each context we should generate an enum containing all alternative names so that we can later switch-case between alternatives;
• tokens should have line and position of the last matched character;
• C target should support memory arenas (seeing how many unnecessary allocations there are in C++ target makes me sad). It should also utilise constant views (in C++ target, many std::strings can be replaced with std::string_views. In C, it is natural to use non-owning char*s to reference strings, so this should not be an issue);
• there should be customization points for altering error messages.

I'd also like to see CI with sanitizers and fuzzing (C++ target have some UBs and leaks that I can't properly trace).

[AmatanHead]

FYI i've started looking into how ATN works and making some drafts for C runtime.

[davidfetter]

FYI i've started looking into how ATN works and making some drafts for C runtime.

Fantastic. Thank you!

[davidfetter]

Bump :)

[AmatanHead]

Well I'm still on it, but things are moving quite slowly because my job takes all the time.

So far I've implemented basic support structures — list, hashmap, interval set, memory pool etc., I also have ATN classes and deserialization for them. I'm working on lexer simulator and DFA.

You can track progress in my fork. Some code review would be appreciated.

[Nic30]

Maybe it would be better to create C wrapper for C++ runtime implementation. I am not saying the that C target is bad idea. But considering the time required for development of the C++ target, the complexity of the implementation....

It makes me think that we should really work on single C compatible runtime. It seems to me that the C++ runtime is not finished and I am not sure when it will be done.

[JackNWhite]

@AmatanHead, may I ask if your C target is still in progress? Your fork mentioned above is inaccessible.

[jimidle]

Personally, I think it is too much work and maintenance to create a C target for v4. Or I would have probably done it.

Not impossible, but there are is a lot of OOP stuff in v4 that means you either have a ton of simulation or a completely different system.

Even a link compatible C++ runtime would be work.