Evgeny Gryaznov

Yes, I agree and will fix this. Originally, I wanted to support the following style of writing productions but in practice it turned out to be very confusing. ``` a...

Yes, I agree that reserving a bunch of nice identifiers (such as Type, Parent, Child, and Children) is unfortunate for the grammar author but this is the only way for...

> Because the node types structs only contain this pointer, they actually fit into the second value of the interface data, and therefore the returned value doesn't require an allocation...

> For what it's worth I have some other tricks up my sleeve to reduce allocations, though they may be a bit too much of micro-optimization for some tastes. I'd...

This is working as intended and is caused by a design decision to not have "roles" in ASTs and model them with intermediate nodes instead. All children nodes within one...

I used to derive semantic actions producing ASTs from declarative annotations. This worked okay-ish but was not easily testable and the declarative syntax was quite sophisticated to make it usable....

Automatic annotation is a double edged sword: there will be cases when it will be helpful but in many cases you'll have to disambiguate yourself. I have enough experience with...

Yes, template instantiation happens before EBNF expansion, and this is rather non-trivial to combine in one step. I'll add better error reporting though. Another input that does not seem to...

Yeah, this is working as expected with the given data model. The parser produces a bunch of ranges, and the AST builder looks at how they are nested to produce...

I think we can introduce a generation option and have both, with 2 being the default. With the go rewrite, we will get declarative inlining, which should make a lot...