Yes, I think this is fine, assuming the implementation of quasi-quote knows how to handle the implicit quoting in ., which I suppose it must.

👍 I'm assuming JuliaLowering must understand this and for flisp lowering, QuoteNode, gets turned into (inert), so what expr.jl is doing here will look the same to flisp lowering.

Looks fine to me, though I hope for a day these can be in Base

though I hope for a day these can be in Base

Isn't QuoteNode in the AST exactly this?

Or were you saying you want syntax for it?

Just syntax, and not wanting to juggle code between JuliaLowering and JuliaSyntax

We might want to revert this. I didn't know at the time that JuliaLowering uses expr.jl for (pre-lowering) SyntaxTree->Expr conversion.

K"inert" won't be in the tree if we got it from normal parsing, but if it's been through macro expansion (or is from expr->syntaxtree, depending on whether we end up back-converting QuoteNodes) it can show up.

It feels like we should possibly introduce some proper layering here then so that JuliaLowering can add additional forms - it seems weird that the conversion logic for such forms should go into the parser. Unless we think that inert is the only case and there will never be any others.

Maybe. JuliaSyntax.node_to_expr is only defined on pre-lowering trees, so JuliaLowering shouldn't be adding new forms. I was thinking that all pre-lowering (post-parsing up to post-macroexpand) trees should allow the same set of forms, so node_to_expr being defined in one piece is best.

On the other hand, I'll probably be setting up a JuliaLowering layer to node_to_expr anyway for macro hygiene (i.e. do a subset of the gynsyming we do normally) since that doesn't make sense to put in the parser. We could also deal with K"inert" there, but other than requiring gensyms, I don't think the post-parse and post-macroexpand tree definitions will (or should) drift too far apart.

cc @c42f in case you have opinions

Thanks for the mention @mlechu.

I didn't see this go past when it was proposed and I was out of action.

Personally I feel like K"inert" is a weird middle ground - something which should eventually have actual "offical syntax" - even if it's just a Base.@inert macro which emits K"inert". Plain quote (not quasiquote) is a pretty fundamental thing to want, really. Actually, if you look in JuliaLowering I've already had a crack at prototyping @inert (unclear if it's correct yet).

Anyway, so it's probably not very harmful to have removed K"inert" for now, but I kinda feel like it does belong in JuliaSyntax even though we technically don't emit it right now.

The issue of dealing with hygiene is separate - as you say that's definitely a lowering problem and shouldn't intrude on JuliaSyntax stuff so we do need a way to deal with that and I'm not 100% sure what is cleanest but I'm thinking about it.