Skip to content

Commit

Permalink
Auto merge of rust-lang#13145 - ChayimFriedman2:unmerge-match-arm, r=…
Browse files Browse the repository at this point in the history
…jonas-schievink

feat: Add a "Unmerge match arm" assist to split or-patterns inside match expressions

Fixes rust-lang#13072.

The way I implemented it it leaves the `OrPat` in place even if there is only one pattern now but I don't think something will break because of that, and when more code will be typed we'll parse it again anyway. Removing it (but keeping the child pattern) is hard, I don't know how to do that.
  • Loading branch information
bors committed Aug 30, 2022
2 parents f02cd0a + 5f132e6 commit 989b09d
Show file tree
Hide file tree
Showing 3 changed files with 321 additions and 0 deletions.
293 changes: 293 additions & 0 deletions crates/ide-assists/src/handlers/unmerge_match_arm.rs
Original file line number Diff line number Diff line change
@@ -0,0 +1,293 @@
use syntax::{
algo::neighbor,
ast::{self, edit::IndentLevel, make, AstNode},
ted::{self, Position},
Direction, SyntaxKind, T,
};

use crate::{AssistContext, AssistId, AssistKind, Assists};

// Assist: unmerge_match_arm
//
// Splits the current match with a `|` pattern into two arms with identical bodies.
//
// ```
// enum Action { Move { distance: u32 }, Stop }
//
// fn handle(action: Action) {
// match action {
// Action::Move(..) $0| Action::Stop => foo(),
// }
// }
// ```
// ->
// ```
// enum Action { Move { distance: u32 }, Stop }
//
// fn handle(action: Action) {
// match action {
// Action::Move(..) => foo(),
// Action::Stop => foo(),
// }
// }
// ```
pub(crate) fn unmerge_match_arm(acc: &mut Assists, ctx: &AssistContext<'_>) -> Option<()> {
let pipe_token = ctx.find_token_syntax_at_offset(T![|])?;
let or_pat = ast::OrPat::cast(pipe_token.parent()?)?.clone_for_update();
let match_arm = ast::MatchArm::cast(or_pat.syntax().parent()?)?;
let match_arm_body = match_arm.expr()?;

// We don't need to check for leading pipe because it is directly under `MatchArm`
// without `OrPat`.

let new_parent = match_arm.syntax().parent()?;
let old_parent_range = new_parent.text_range();

acc.add(
AssistId("unmerge_match_arm", AssistKind::RefactorRewrite),
"Unmerge match arm",
pipe_token.text_range(),
|edit| {
let pats_after = pipe_token
.siblings_with_tokens(Direction::Next)
.filter_map(|it| ast::Pat::cast(it.into_node()?));
// FIXME: We should add a leading pipe if the original arm has one.
let new_match_arm = make::match_arm(
pats_after,
match_arm.guard().and_then(|guard| guard.condition()),
match_arm_body,
)
.clone_for_update();

let mut pipe_index = pipe_token.index();
if pipe_token
.prev_sibling_or_token()
.map_or(false, |it| it.kind() == SyntaxKind::WHITESPACE)
{
pipe_index -= 1;
}
or_pat.syntax().splice_children(
pipe_index..or_pat.syntax().children_with_tokens().count(),
Vec::new(),
);

let mut insert_after_old_arm = Vec::new();

// A comma can be:
// - After the arm. In this case we always want to insert a comma after the newly
// inserted arm.
// - Missing after the arm, with no arms after. In this case we want to insert a
// comma before the newly inserted arm. It can not be necessary if there arm
// body is a block, but we don't bother to check that.
// - Missing after the arm with arms after, if the arm body is a block. In this case
// we don't want to insert a comma at all.
let has_comma_after =
std::iter::successors(match_arm.syntax().last_child_or_token(), |it| {
it.prev_sibling_or_token()
})
.map(|it| it.kind())
.skip_while(|it| it.is_trivia())
.next()
== Some(T![,]);
let has_arms_after = neighbor(&match_arm, Direction::Next).is_some();
if !has_comma_after && !has_arms_after {
insert_after_old_arm.push(make::token(T![,]).into());
}

let indent = IndentLevel::from_node(match_arm.syntax());
insert_after_old_arm.push(make::tokens::whitespace(&format!("\n{indent}")).into());

insert_after_old_arm.push(new_match_arm.syntax().clone().into());

ted::insert_all_raw(Position::after(match_arm.syntax()), insert_after_old_arm);

if has_comma_after {
ted::insert_raw(
Position::last_child_of(new_match_arm.syntax()),
make::token(T![,]),
);
}

edit.replace(old_parent_range, new_parent.to_string());
},
)
}

#[cfg(test)]
mod tests {
use crate::tests::{check_assist, check_assist_not_applicable};

use super::*;

#[test]
fn unmerge_match_arm_single_pipe() {
check_assist(
unmerge_match_arm,
r#"
#[derive(Debug)]
enum X { A, B, C }
fn main() {
let x = X::A;
let y = match x {
X::A $0| X::B => { 1i32 }
X::C => { 2i32 }
};
}
"#,
r#"
#[derive(Debug)]
enum X { A, B, C }
fn main() {
let x = X::A;
let y = match x {
X::A => { 1i32 }
X::B => { 1i32 }
X::C => { 2i32 }
};
}
"#,
);
}

#[test]
fn unmerge_match_arm_guard() {
check_assist(
unmerge_match_arm,
r#"
#[derive(Debug)]
enum X { A, B, C }
fn main() {
let x = X::A;
let y = match x {
X::A $0| X::B if true => { 1i32 }
_ => { 2i32 }
};
}
"#,
r#"
#[derive(Debug)]
enum X { A, B, C }
fn main() {
let x = X::A;
let y = match x {
X::A if true => { 1i32 }
X::B if true => { 1i32 }
_ => { 2i32 }
};
}
"#,
);
}

#[test]
fn unmerge_match_arm_leading_pipe() {
check_assist_not_applicable(
unmerge_match_arm,
r#"
fn main() {
let y = match 0 {
|$0 0 => { 1i32 }
1 => { 2i32 }
};
}
"#,
);
}

#[test]
fn unmerge_match_arm_multiple_pipes() {
check_assist(
unmerge_match_arm,
r#"
#[derive(Debug)]
enum X { A, B, C, D, E }
fn main() {
let x = X::A;
let y = match x {
X::A | X::B |$0 X::C | X::D => 1i32,
X::E => 2i32,
};
}
"#,
r#"
#[derive(Debug)]
enum X { A, B, C, D, E }
fn main() {
let x = X::A;
let y = match x {
X::A | X::B => 1i32,
X::C | X::D => 1i32,
X::E => 2i32,
};
}
"#,
);
}

#[test]
fn unmerge_match_arm_inserts_comma_if_required() {
check_assist(
unmerge_match_arm,
r#"
#[derive(Debug)]
enum X { A, B }
fn main() {
let x = X::A;
let y = match x {
X::A $0| X::B => 1i32
};
}
"#,
r#"
#[derive(Debug)]
enum X { A, B }
fn main() {
let x = X::A;
let y = match x {
X::A => 1i32,
X::B => 1i32
};
}
"#,
);
}

#[test]
fn unmerge_match_arm_inserts_comma_if_had_after() {
check_assist(
unmerge_match_arm,
r#"
#[derive(Debug)]
enum X { A, B }
fn main() {
let x = X::A;
match x {
X::A $0| X::B => {},
}
}
"#,
r#"
#[derive(Debug)]
enum X { A, B }
fn main() {
let x = X::A;
match x {
X::A => {},
X::B => {},
}
}
"#,
);
}
}
2 changes: 2 additions & 0 deletions crates/ide-assists/src/lib.rs
Original file line number Diff line number Diff line change
Expand Up @@ -185,6 +185,7 @@ mod handlers {
mod replace_string_with_char;
mod replace_turbofish_with_explicit_type;
mod split_import;
mod unmerge_match_arm;
mod sort_items;
mod toggle_ignore;
mod unmerge_use;
Expand Down Expand Up @@ -278,6 +279,7 @@ mod handlers {
sort_items::sort_items,
split_import::split_import,
toggle_ignore::toggle_ignore,
unmerge_match_arm::unmerge_match_arm,
unmerge_use::unmerge_use,
unnecessary_async::unnecessary_async,
unwrap_block::unwrap_block,
Expand Down
26 changes: 26 additions & 0 deletions crates/ide-assists/src/tests/generated.rs
Original file line number Diff line number Diff line change
Expand Up @@ -2207,6 +2207,32 @@ fn arithmetics {
)
}

#[test]
fn doctest_unmerge_match_arm() {
check_doc_test(
"unmerge_match_arm",
r#####"
enum Action { Move { distance: u32 }, Stop }
fn handle(action: Action) {
match action {
Action::Move(..) $0| Action::Stop => foo(),
}
}
"#####,
r#####"
enum Action { Move { distance: u32 }, Stop }
fn handle(action: Action) {
match action {
Action::Move(..) => foo(),
Action::Stop => foo(),
}
}
"#####,
)
}

#[test]
fn doctest_unmerge_use() {
check_doc_test(
Expand Down

0 comments on commit 989b09d

Please sign in to comment.