Inside `impl`s, complete unimplemented functions

[matklad]

See #1022 (comment)

The idea is that for

trait T {
   fn foo();
}

impl T for () {
   |
}

We should get completions for foo. Specifically, the completion should trigger on fn keyword and on bare identifier (so that users who know about this feature can just type the identifier)

The code, responsible for completions, lives here:

https://github.com/rust-analyzer/rust-analyzer/blob/958df1276e3d97b189dcd9de974564011c70ef62/crates/ra_ide_api/src/completion.rs

To implement this feature, one has to add more info to CompletionContext and to create a separate completion module. Take a look at complete_fn_param which does somewhat similar magical completions.

This should also work for types and consts

[flodiebold]

Also, the completion should complete to the whole fn (/type/const) declaration, e.g. fn foo() {|} :) I assume that's implied, but wanted to make sure ;)

[detrumi]

I'll be working on this. Found the IMPL_BLOCK on ctx.leaf.ancestors(), so I know where to start :)

[hargoniX]

So the PR from detrumi has apparently been closed due to inactivity, does this mean that this issue is open to work on again?

[matklad]

Yes, go for it!

…

On Thursday, 22 August 2019, Henrik Böving ***@***.***> wrote: So the PR from detrumi has apparently been closed due to inactivity, does this mean that this issue is open to work on again? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1046>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AANB3M3X3GPU5WVXPO7BNJTQF26JVANCNFSM4HAZ7ONQ> .

[hargoniX]

So having looked a bit into the code and the docs for this project, should I just take over the exact code from the original PR and continue working from there or should I write my own PR from the base up?

[kdelorey]

Since there hasn't been activity on this in a while, I'm going to attempt it if that's alright? I do have basic functionality working as a proof-of-concept to myself.

[matklad]

@kdelorey 👍

[kdelorey]

Update, I'm still working on this slowly but surely on my free time. I've just been trying to get up to speed so I can accomplish this task in the effective way I can think of (which I think is relying on the hir as much as I can).

[kdelorey]

@matklad (or whomever), is there any recommendations in regards to the completion of an associated constant?

I basically want the magic completion to write the following:

trait SomeTrait {
    const SOMETHING: u8;
}

impl SomeTrait for () {
    // text in between `[` and `]` is a magic completion:
    [const SOMETHING: u8 = ]
}

I'm mostly having trouble getting the type information in order to formulate the completion. It feels maybe a little dirty, but I could walk the syntax tree and copy everything up until the ; (or = in cases where a default value is supplied)?