Order auto-imports by relevance
Fixes#10337.
Basically we sort the imports according to how "far away" the imported item is from where we want to import it to. This change makes it so that imports from the current crate are sorted before any third-party crates. Additionally, we make an exception for builtin crates (`std`, `core`, etc.) so that they are sorted before any third-party crates.
There are probably other heuristics that should be added to improve the experience (such as preferring imports that are common elsewhere in the same crate, and ranking crates depending on the dependency graph). However, I think this is a first good step.
PS. This is my first time contributing here, so please be gentle if I have missed something obvious :-)
feature: `Merge imports` assist can merge multiple selected imports.
The selected imports have to have a common prefix in paths.
Select imports or use trees to merge:
```rust
$0use std::fmt::Display;
use std::fmt::Debug;
use std::fmt::Write;$0
```
Apply `Merge imports`:
```rust
use std::fmt::{Display, Debug, Write};
```
Closes#12426
internal: Shorten main thread names
Linux effectively has a 15 byte limit, which resulted in `rust-analyzer s` and `rust-analyzer p`. That's still unambiguous, but probably not obvious.
The selected imports have to have a common prefix in paths.
Before
```rust
$0use std::fmt::Display;
use std::fmt::Debug;$0
```
After
```rust
use std::fmt::{Display, Debug};
```
fix(extract_module) resolving import panics and improve import resolution
- Should solve #11766
- While adding a test case for this issue, I observed another issue:
For this test case:
```rust
mod x {
pub struct Foo;
pub struct Bar;
}
use x::{Bar, Foo};
$0type A = (Foo, Bar);$0
```
extract module should yield this:
```rust
mod x {
pub struct Foo;
pub struct Bar;
}
use x::{};
mod modname {
use super:❌:Bar;
use super:❌:Foo;
pub(crate) type A = (Foo, Bar);
}
```
instead it gave:
```rust
mod x {
pub struct Foo;
pub struct Bar;
}
use x::{};
mod modname {
use x::Bar;
use x::Foo;
pub(crate) type A = (Foo, Bar);
}
```
So fixed this problem with second commit
minor: Update commit/PR style guide
What matters for the changelog is having PR title/description from the user's perspective, while commits can be internal.
fix(ide-db): correct single-file module rename
Fixes a bug where rust-analyzer would emit `WorkspaceEdit`s with paths to dirs instead of files for the following project layout.
lib.rs
```rust
mod foo;
```
foo.rs
```rust
mod bar {
struct Bar;
}
```
Also fixes emitted paths for modules with mod.rs.
The bug resulted in panic in helix editor when attempting to rename a module.
fix: Retrigger visibility completion after parentheses
close#12390
This PR add `(` to trigger_characters as discussed in original issue.
Some questions:
1. Is lsp's `ctx.trigger_character` from `params.context` is the same as `ctx.original_token` inside actually completions?
1. If not what's the difference?
2. if they are the same, it's unnecessary to pass it down from handler at all.
3. if they are the same, maybe we could parse it from fixture directly instead of using the `check_with_trigger_character` I added.
2. Some completion fixtures written as `($0)` ( https://github.com/rust-lang/rust-analyzer/blob/master/crates/ide-completion/src/tests/fn_param.rs#L105 as an example), If I understand correctly they are not invoked outside tests at all?
1. using `ctx.original_token` directly would break these tests as well as parsing trigger_character from fixture for now.
2. I think it make sense to allow `(` triggering these cases?
3. I hope this line up with #12144
fix overflow during type inference for tuple struct patterns
The following code causes integer overflow during type inference for (malformed) tuple struct patterns.
```rust
struct S(usize);
let S(.., a, b) = S(1);
```
It has been panicking only in debug builds, and working in a way in release builds but it was inconsistent with type inference for tuple patterns:
```rust
struct S(usize);
let S(.., a, b) = S(1); // a -> unknown, b -> usize
let (.., a, b) = (1,); // a -> usize, b -> unknown
```
With this PR, the overflow no longer happens by utilizing `saturating_sub()` like in other places and type inference for tuple struct patterns is in line with that for tuple patterns.
