Completion based on trait bounds

[kngwyu]

Related: #706, #829

General roadmap

• for FnArg
  already implemented in Completion by trait bounds for fnarg #828, Support trait bound in where clause #937

fn f<T: Clone>(a: T) {
    let b = a.clo~
}

• for struct member
  implemented in Completion based on impl<T: Bound>  #948 , Type inference for struct field via function bound #975

struct S<T> {
    mem: T
}
fn f<T: Clone>(a: S<T>) {
    let b = a.mem.clo~
}
impl<T: Clone> S<T> {
    fn f(self) {
        let b = a.mem.clo~
    } 
}

• for closures

fn fun<F: Fn() -> Option<i64>>(f: F) {
    let a = f().unwra~
}

• associated functions and types
  implemented in Completes associated function & type for type parameter #991

Precise roadmaps

refactor generics support

• Intoroduce ImplHeader Intoroduce 'ImplHeader' #938
• Replace MatchType::Impl and MatchType::TraitImpl with ImplHeader
• Remove Match::generics_args and Match::generic_types
• Modify MatchType::Enum and etc to have generics
• Intoroduce MatchType::Method which has ImplHeader
  - [ ] Remove PathSearch TOO HARD
• Refactor TraitBounds again
  - [ ] Use GenericsArgs to resolve type params, instead of taking Match Now I think there's no use

enhance scope_start's inefficiency

• Have scope cache

Method search

• rewrite method search using ImplHeader

TypeName resolution

• resolve type parameter in ImplHeader when resolve_path is called

Related issues

• Generic type with multiple type parameter(Not handle Box<Trait + Send> #900)
  I think this will be easily fixed after refactoring all codes.

[letmutx]

Hi! Is struct member method completions working? I can't seem to get completions with the latest master. I am using the same example above

struct S<T> {
    mem: T
}
fn f<T: Clone>(a: S<T>) {
    let b = a.mem.clo~
}

$ RUST_LOG=racer=trace ../racer/target/debug/racer complete 5 21 src/main.rs 2> trace.txt
PREFIX 71,74,clo
END

trace.txt

Weirdly enough, the testcases in tests/trait_bounds.rs pass.

[kngwyu]

Thanks.

Weirdly enough, the testcases in tests/trait_bounds.rs pass.

Tests contain only 'bounded by impl' cases, like

impl<T: Clone> S<T> {
    fn f(self) {
        let b = a.mem.clo~
    } 
}

, so it's not weird.
I don't remember it's simply not implemented or broken, but anyway fix it later.

[letmutx]

Right. I was obviously overlooking that. Sorry. I would like to take a go at this.

…

On Wed, Oct 17, 2018 at 7:29 AM Yuji Kanagawa ***@***.***> wrote: Thanks. Weirdly enough, the testcases in tests/trait_bounds.rs pass. Tests contain only 'bounded by impl' cases, like impl<T: Clone> S<T> { fn f(self) { let b = a.mem.clo~ } } , so it's not weird. I don't remember it's simply not implemented or broken, but anyway fix it later. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#939 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AZcHBLvVtSfCCmh02uQ5gALD8sQAYbcvks5ulo8bgaJpZM4WTPiI> .

[kngwyu]

@letmutx
Thanks, but I want to fix this by myself, to recall what I implemented in last month.

[letmutx]

@kngwyu Okay. I have some progress here: https://github.com/letmutx/racer/tree/add-struct-member-trait-completions