Don't require semicolon after macro definition

[aochagavia]

The following code generates a syntax error "expected SEMI" at the position of the cursor

fn test() {
    macro_rules! indent {
        () => {
            for _ in 0..level {
                buf.push_str("  ");
            }
        };
    }<|>
    return;
}

[phansch]

This looks like a nice starter issue. I would like to see if I can fix it today.

[aochagavia]

Feel free to ask for help here if you get stuck 😉

[phansch]

I think I jumped the gun a bit early here. Holiday is coming up next week and I want to finish up a compilers course (this one) so that I start with proper understanding of the basics. I do want to get involved, it may just take a few more weeks =)

If someone else wants to pick this up, feel free!

[aochagavia]

It turns out this is not only a problem for macro definitions, but also macro usage. I have created #225 which is more general.

[DJMcNab]

But surely macro definition and usages are different constructs entirely. I would argue that this is should be simpler to fix than #225. It might only require a change to https://github.com/rust-analyzer/rust-analyzer/blob/477de790b0211196256a772befe4f577d1a8ba14/crates/ra_syntax/src/grammar/expressions/mod.rs#L204-L209

[aochagavia]

macro_rules! is itself a macro, so macro definition is actually a macro usage, as surprising as it may sound

[DJMcNab]

I thought it has different syntax, i.e. macro_rules! $name: ident {$inner: tt}, which you cannot define with a regular macro. I'm not certain though

[aochagavia]

I think it is true that you can't define own macros with that syntax. Still, the parser currently treats macro_rules! as a macro usage (you can test that by displaying the AST in VSCode). We could consider changing that, though I am not sure it is worth it to make a distinction at the parser level.

[CAD97]

This is also the case in the current wg-grammar grammar draft.