sep_by(any(), char(',')) about combine

[Marwes]

Not really from what I can think of, you may want to make a custom "sep_by" parser to handle that.

With the current parsers combine provides though I think you might be able to hack around it with the recently added look_ahead parser though.

sep_by(look_ahead(satisfy(|c| c != ',')).with(parser(other)), char(','))

It looks a bit weird but I think it should do what you want (might be some edge case I haven't considered though).

EDIT: No that will not actually work. I am not sure that what you want is actually possible in combine.

[Marwes]

I don't believe that what you want is easily possible with combine which uses recursive descent for parsing. If your other parser is able to successfully consume a ',' at some point during its parsing it needs to be "aware" in some way that the sep_by parser should take precedence which is at odds with combine's model in which each parser is unaware of where it is used.

[Qard]

I don't think it's necessarily at odds. The parser just needs to store and consume all tokens until it reaches the tokens that get consumed by the separator parser, then feed everything up to that point into the item parser. It's kind of cheating at LALR, but it might work okay.

[Marwes]

It might work but if you have a grammar which requires you to work around it like that it might be better to use a more powerful tool/library instead (lalrpop comes to mind but I don't know how mature it is yet).

[Marwes]

Closing. Feel free to reopen if there is more to be said on this topic

[PSeitz]

When searching for precendence in this repo, this is the only result, but it seems the combine_language crate enables this? (https://docs.rs/combine-language/3.0.1/combine_language/struct.Assoc.html#structfield.precedence)

[Marwes]

That's a slightly different meaning of precedence but yes. If you need to parse one is expressions, combine-language supports that.