new lint: unnecessary_reserve
#14114
Conversation
rustbot
added
the
S-waiting-on-review
wowinter13
changed the title
unnecessary_reserve
| /// let array: &[usize] = &[1, 2]; | ||
| /// vec.extend(array); | ||
| /// ``` | ||
| #[clippy::version = "1.64.0"] |
There was a problem hiding this comment.
The current version is "1.86.0".
| /// | ||
| /// This lint checks for a call to `reserve` before `extend` on a `Vec` or `VecDeque`. | ||
| /// ### Why is this bad? | ||
| /// Since Rust 1.62, `extend` implicitly calls `reserve` |
There was a problem hiding this comment.
I think you can omit "Since Rust 1.62", since it is more than two years old, while checking the MSRV is a good idea, talking about it in the description may not be necessary.
| impl_lint_pass!(UnnecessaryReserve => [UNNECESSARY_RESERVE]); | ||
|
|
||
| pub struct UnnecessaryReserve { | ||
| msrv: Msrv, |
There was a problem hiding this comment.
The lint also needs to be added to clippy_config/src/conf.rs.
| /// ``` | ||
| #[clippy::version = "1.64.0"] | ||
| pub UNNECESSARY_RESERVE, | ||
| pedantic, |
There was a problem hiding this comment.
Should it be pedantic or complexity?
| let acceptable_types = [sym::Vec, sym::VecDeque]; | ||
| acceptable_types.iter().any(|&acceptable_ty| { | ||
| match cx.typeck_results().expr_ty(struct_calling_on).peel_refs().kind() { | ||
| ty::Adt(def, _) => cx.tcx.is_diagnostic_item(acceptable_ty, def.did()), | ||
| _ => false, | ||
| } | ||
| }) |
There was a problem hiding this comment.
cx.typeck_results() should be called once. Also, using expr_ty_adjusted() would let you handle cases where the Vec or VecDeque is behind a smart pointer such as Box (and you should add a test case for this).
I think you can match in a simpler way by:
- Moving
adt_def_id()fromclippy_utils::ty::type_certaintytoclippy_utils::ty, making it public and adding documentation ("Check iftywith references peeled is anAdtand return itsDefId."). Fix the references by importingsuper::adt_def_id()in thetype_certaintymodule. - Using
adt_def_id()in your function, for example:
| let acceptable_types = [sym::Vec, sym::VecDeque]; | |
| acceptable_types.iter().any(|&acceptable_ty| { | |
| match cx.typeck_results().expr_ty(struct_calling_on).peel_refs().kind() { | |
| ty::Adt(def, _) => cx.tcx.is_diagnostic_item(acceptable_ty, def.did()), | |
| _ => false, | |
| } | |
| }) | |
| if let Some(did) = adt_def_id(cx.typeck_results().expr_ty_adjusted(struct_calling_on)) { | |
| matches!(cx.tcx.get_diagnostic_name(did), Some(sym::Vec | sym::VecDeque)) | |
| } else { | |
| false | |
| } |
| struct_expr: &Expr<'tcx>, | ||
| args_a: &Expr<'tcx>, | ||
| ) -> Option<rustc_span::Span> { | ||
| let args_a_kind = &args_a.kind; |
There was a problem hiding this comment.
It is possible to use args_a.kind directly at line 115, this is clearer and will not be more costly.
| let mut read_found = false; | ||
| let mut spanless_eq = SpanlessEq::new(cx); | ||
|
|
||
| let _: Option<!> = for_each_expr(cx, block, |expr: &Expr<'tcx>| { |
There was a problem hiding this comment.
Right now, this will suggest to remove a call to reserve() which follows a call to extend() as well. It would be better to check only statements following the call to reserve(), as right now there is a risk of getting many false positives.
|
@samueltardieu Samuel, really appreciate your review and points! |
resurrection of #10157
fixes #8982
changelog: [
unnecessary_reserve]: add new lint unnecessary_reserve