TYPE: Add inlay parameter hints

[farodin91]

Fixes #1141

[farodin91]

cc @vasily-kirichenko

[vasily-kirichenko]

wow. let me try!

[vasily-kirichenko]

Pure beauty :)

[vasily-kirichenko]

@farodin91 I found quite a lot of bugs :)

[farodin91]

I don't show type if i can't infer a type.

[farodin91]

Adding tests and Options

[vasily-kirichenko]

But isn't 1 + 1 has i32 type?..

[Undin]

Not now:)
At this moment plugin can't infer result of arithmetic operations for primitive types. Their implementations are created with macros but, unfortunately, we can't expand macros yet.

[vasily-kirichenko]

Cool :) But generic parameters are not inferred.

[farodin91]

If you set the parameter it should infer parameters. For example let x:HashMap<String, String> after this is should infer.

[farodin91]

@alygin Could you review?

[alygin]

My main concern here is optionality of this feature. I think it must be possible to switch it off separately for the argument names and for the inferred types.

IDEA has a special setting for it: "Editor -> General -> Appearance -> Show parameter name hints" with the "Configure..." button for extended settings, including per-language options. Unfortunately, other products (I looked in CLion) still don't have such block in the Preferences, so we need to add our own settings, probably to the "Langs & Frameworks -> Rust" panel (RustProjectSettingsPanel).

PS: I'm on vacation now, so sorry for possible delays in my answers :)

[alygin]

My main concern here is optionality of this feature. I think it must be possible to switch it off separately for the argument names and for the inferred types.

IDEA has a special setting for it: "Editor -> General -> Appearance -> Show parameter name hints" with the "Configure..." button for extended settings, including per-language options. Unfortunately, other products (I looked in CLion) still don't have such block in the Preferences, so we need to add our own settings, probably to the "Langs & Frameworks -> Rust" panel (RustProjectSettingsPanel).

PS: I'm on vacation now, so sorry for possible delays in my answers :)

[alygin]

": " + type.toString() could be just ": $type"

[alygin]

Looks like a typo. Did you mean LET_BINDING_HINT?

[alygin]

You also need to return null here if valueParameter.pat? is null, to handle anonymous arguments properly. Here's a snippet to test this case:

trait Foo {
    fn foo(&self, u32);
    fn bar(&self) {
        self.foo(10);    // <-- shouldn't add `null: `
    }
}

[kumbayo]

When calling a method using UFCS syntax, the labels are shifted by 1 for the extra self argument.

struct S;

impl S {
    fn foo_self(self, param1: u32) {}
    fn foo_self_ref(&self, param1: u32) {}
    fn foo_self_mut_ref(&mut self, param1: u32) {}
}

fn test() {
    let mut b = S;

    b.foo_self_ref(1);
    b.foo_self_mut_ref(1);
    b.foo_self(1);

    S::foo_self_ref(&b, 1); //bad
    S::foo_self_mut_ref(&mut b, 1); //bad
    S::foo_self(b, 1); //bad
}

[farodin91]

@alygin For IDEA you can de/select a every single feature. Any idea to integrate it into rust settings panel and in the configure area? At the same time.

[kumbayo]

Is there a reason / use case why you do not want to show the unit type on local variables?
Noise during typing new code?

I think it might be useful to show it for situations like.

let z = accidentally_returns_nothing();

fn accidentally_returns_nothing() -> () {
    ;
}

[kumbayo]

IntelliJ claims this import is unused.

[matklad]

Hm, compilation fails on 2016.3... I need to check if we can make extensions in plugin.xml dependent on the version of IDE...

[farodin91]

But my implementation of settings shouldn't work in any case for 2016.3.

[kumbayo]

It seems the S::foo_self_ref(&b, 1); case has regressed.
In an intermediate version it already worked.
But now it is not working anymore and the test for it is gone again.

[farodin91]

@kumbayo Thank i pushed something wrong.

[farodin91]

@matklad Any progress to fix compilation fails on 2016.3?

[matklad]

@farodin91 looks like the easiest solution is just to wait while we drop support for 2016.3. This'll happen once 2017.2 is out, which is pretty soon.

[vlad20012]

2017.2 just released

[farodin91]

I will rebase after #1487 is merged.

[farodin91]

Resolved

[farodin91]

@matklad Do you like to merge it?

[matklad]

@farodin91 sure!

bors r+

[bors]

Build succeeded

• continuous-integration/appveyor/branch
• continuous-integration/travis-ci/push