rewrite the parser

[k1ngmang]

There's no point in dragging a dependency like antlr into a project. It's easier to write a regular recursive parser or a Pratt parser. Not only will the project get rid of an obviously unnecessary dependency, but such code will also be very easy to read and write.

[emil14]

@k1ngmang

There're two reasons why I think it's better to stick with current solution:

1. It's declarative to have grammar file as source of true for parsing;
2. Language is not yet stable including it's syntax and parser-generators allow to move faster at MVP stage.

I agree though that manual parser allows to get rid of dependency and perhaps improve performance (it's good to have some parser benchmarks first, perhaps)

[k1ngmang]

@emil14

It's not just about performance, but also about maintainability. While a formalized grammar can be useful in many places (for example, for plugins in popular editors like IntelliJ IDEA), editing a regular recursive parser is still easier.

You mentioned that the language's syntax is unstable, which means that describing complex syntactic constructs will be a real pain in the ass in the future, so it's better to take care of this in advance. For IDE, we can describe a regular BNF grammar.

[emil14]

It's not just about performance, but also about maintainability. While a formalized grammar can be useful in many places (...), editing a regular recursive parser is still easier.

Easier how?

• Grammar is not longer source of true -> you have to manually keep it in sync with the parser source code -> it's possible to get them out of sync (not possible at the moment)
• You change grammar -> make antlr -> update listener/visitor (you even get compile errors after codegen)

language's syntax is unstable, which means that describing complex syntactic constructs will be a real pain in the ass in the future, so it's better to take care of this in advance

Yes and to me it's the reason to have declarative grammar!

---

Not directly answers your message but related to the topic.

I don't honestly think Neva syntax will get more complicated than now, I'm actually thinking about removing existing features (see #935 in case of curiousity)