Skip to content

Better error for missing tuple pattern in args #45069

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 5 commits into from
Oct 13, 2017
Merged

Better error for missing tuple pattern in args #45069

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
8782d0f
Better error for missing tuple pattern in args (#44150)
sinkuu Oct 6, 2017
b6eef82
Rename variables
sinkuu Oct 6, 2017
c74bda1
Add suggestion
sinkuu Oct 7, 2017
fa14f79
Reword
sinkuu Oct 10, 2017
f577847
Reword
sinkuu Oct 13, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
201 changes: 160 additions & 41 deletions src/librustc/traits/error_reporting.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -711,41 +711,105 @@ impl<'a, 'gcx, 'tcx> InferCtxt<'a, 'gcx, 'tcx> {
}
}

OutputTypeParameterMismatch(ref expected_trait_ref, ref actual_trait_ref, _) => {
OutputTypeParameterMismatch(ref found_trait_ref, ref expected_trait_ref, _) => {
let found_trait_ref = self.resolve_type_vars_if_possible(&*found_trait_ref);
let expected_trait_ref = self.resolve_type_vars_if_possible(&*expected_trait_ref);
let actual_trait_ref = self.resolve_type_vars_if_possible(&*actual_trait_ref);
if actual_trait_ref.self_ty().references_error() {
if expected_trait_ref.self_ty().references_error() {
return;
}
let expected_trait_ty = expected_trait_ref.self_ty();
let found_span = expected_trait_ty.ty_to_def_id().and_then(|did| {
let found_trait_ty = found_trait_ref.self_ty();

let found_did = found_trait_ty.ty_to_def_id();
let found_span = found_did.and_then(|did| {
self.tcx.hir.span_if_local(did)
});

let self_ty_count =
match expected_trait_ref.skip_binder().substs.type_at(1).sty {
let found_ty_count =
match found_trait_ref.skip_binder().substs.type_at(1).sty {
ty::TyTuple(ref tys, _) => tys.len(),
_ => 1,
};
let arg_ty_count =
match actual_trait_ref.skip_binder().substs.type_at(1).sty {
ty::TyTuple(ref tys, _) => tys.len(),
_ => 1,
let (expected_tys, expected_ty_count) =
match expected_trait_ref.skip_binder().substs.type_at(1).sty {
ty::TyTuple(ref tys, _) =>
(tys.iter().map(|t| &t.sty).collect(), tys.len()),
ref sty => (vec![sty], 1),
};
if self_ty_count == arg_ty_count {
if found_ty_count == expected_ty_count {
self.report_closure_arg_mismatch(span,
found_span,
expected_trait_ref,
actual_trait_ref)
found_trait_ref,
expected_trait_ref)
} else {
// Expected `|| { }`, found `|x, y| { }`
// Expected `fn(x) -> ()`, found `|| { }`
let expected_tuple = if expected_ty_count == 1 {
expected_tys.first().and_then(|t| {
if let &&ty::TyTuple(ref tuptys, _) = t {
Some(tuptys.len())
} else {
None
}
})
} else {
None
};

// FIXME(#44150): Expand this to "N args expected but a N-tuple found."
Copy link
Contributor

@nikomatsakis nikomatsakis Oct 6, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why does this say "FIXME"?

Copy link
Contributor Author

@sinkuu sinkuu Oct 7, 2017 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This PR improves message for "expected |(x, y)|, found |x, y|" case but not "expected |x, y|, found |(x, y)|" case. In the latter case, found_trait_ref should be FnMut<((_, _),)>, but it becomes FnMut<(_,)> and I don't know how to fix.

Copy link
Contributor

@estebank estebank Oct 10, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@nikomatsakis, I believe @sinkuu is referring to the following case:

 18 |     f(|(x, y)| x + y);
    |     ^ -------- takes a single tuple as an argument
    |     | |
    |     | help: consider changing to: `|x, y|`
    |     |
    |     expected closure that takes 2 arguments

We should create an issue after this PR lands to track that case.

// Type of the 1st expected argument is somehow provided as type of a
// found one in that case.
//
// ```
// [1i32, 2, 3].sort_by(|(a, b)| ..)
// // ^^^^^^^^
// // expected_trait_ref: std::ops::FnMut<(&i32, &i32)>
// // found_trait_ref: std::ops::FnMut<(&i32,)>
// ```

let (closure_span, closure_args) = found_did
.and_then(|did| self.tcx.hir.get_if_local(did))
.and_then(|node| {
if let hir::map::NodeExpr(
&hir::Expr {
node: hir::ExprClosure(_, ref decl, id, span, _),
..
}) = node
{
let ty_snips = decl.inputs.iter()
.map(|ty| {
self.tcx.sess.codemap().span_to_snippet(ty.span).ok()
.and_then(|snip| {
// filter out dummy spans
if snip == "," || snip == "|" {
None
} else {
Some(snip)
}
})
})
.collect::<Vec<Option<String>>>();

let body = self.tcx.hir.body(id);
let pat_snips = body.arguments.iter()
.map(|arg|
self.tcx.sess.codemap().span_to_snippet(arg.pat.span).ok())
.collect::<Option<Vec<String>>>();

Some((span, pat_snips, ty_snips))
} else {
None
}
})
.map(|(span, pat, ty)| (Some(span), Some((pat, ty))))
.unwrap_or((None, None));
let closure_args = closure_args.and_then(|(pat, ty)| Some((pat?, ty)));

self.report_arg_count_mismatch(
span,
found_span,
arg_ty_count,
self_ty_count,
expected_trait_ty.is_closure()
closure_span.or(found_span),
expected_ty_count,
expected_tuple,
found_ty_count,
closure_args,
found_trait_ty.is_closure()
)
}
}
Expand All @@ -767,32 +831,87 @@ impl<'a, 'gcx, 'tcx> InferCtxt<'a, 'gcx, 'tcx> {
err.emit();
}

fn report_arg_count_mismatch(&self,
span: Span,
found_span: Option<Span>,
expected: usize,
found: usize,
is_closure: bool)
-> DiagnosticBuilder<'tcx>
{
fn report_arg_count_mismatch(
&self,
span: Span,
found_span: Option<Span>,
expected: usize,
expected_tuple: Option<usize>,
found: usize,
closure_args: Option<(Vec<String>, Vec<Option<String>>)>,
is_closure: bool
) -> DiagnosticBuilder<'tcx> {
use std::borrow::Cow;

let kind = if is_closure { "closure" } else { "function" };

let args_str = |n| format!(
"{} argument{}",
n,
if n == 1 { "" } else { "s" }
);

let mut err = struct_span_err!(self.tcx.sess, span, E0593,
"{} takes {} argument{} but {} argument{} {} required",
if is_closure { "closure" } else { "function" },
found,
if found == 1 { "" } else { "s" },
expected,
if expected == 1 { "" } else { "s" },
"{} takes {}, but {} {} required",
kind,
if expected_tuple.is_some() {
Cow::from("multiple arguments")
} else {
Cow::from(args_str(found))
},
if expected_tuple.is_some() {
Cow::from("a tuple argument")
Copy link
Contributor

@nikomatsakis nikomatsakis Oct 9, 2017 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think I figured out what is bothering me about this wording. The construction is not quite parallel. It says "multiple arguments" are expected, but then says "but a tuple argument", which is not obviously the opposite. Also, our usual construction is "expected X, but found Y", whereas the message here says "found X, but Y is required" (pre-existing, I know, but I still find it a bit surprising).

Perhaps this?

expected a closure that takes a single tuple as argument but found a closure that takes N distinct arguments

Maybe that's too wordy. We could tweak it, though it goes a bit away from the "expected X but found Y" form:

expected closure to take a single tuple as argument but it takes N distinct arguments

or, maybe better:

closure is expected to take a single tuple as argument, but it takes N distinct arguments

(I would probably change the existing message that doesn't mention tuples to match, as well.)

Copy link
Contributor Author

@sinkuu sinkuu Oct 10, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It looks better, thanks! Reworded the message to the last one.

} else {
Cow::from(args_str(expected))
},
if expected == 1 { "is" } else { "are" });

err.span_label(span, format!("expected {} that takes {} argument{}",
if is_closure { "closure" } else { "function" },
expected,
if expected == 1 { "" } else { "s" }));
err.span_label(
span,
format!(
"expected {} that takes {}{}",
kind,
args_str(expected),
if let Some(n) = expected_tuple {
assert!(expected == 1);
Cow::from(format!(", a {}-tuple", n))
} else {
Cow::from("")
}
)
);

if let Some(span) = found_span {
err.span_label(span, format!("takes {} argument{}",
found,
if found == 1 { "" } else { "s" }));
if let (Some(expected_tuple), Some((pats, tys))) = (expected_tuple, closure_args) {
if expected_tuple != found || pats.len() != found {
err.span_label(span, format!("takes {}", args_str(found)));
} else {
let sugg = format!(
"|({}){}|",
pats.join(", "),

// add type annotations if available
if tys.iter().any(|ty| ty.is_some()) {
Cow::from(format!(
": ({})",
tys.into_iter().map(|ty| if let Some(ty) = ty {
ty
} else {
"_".to_string()
}).collect::<Vec<String>>().join(", ")
))
} else {
Cow::from("")
},
);

err.span_suggestion(span, "consider changing to", sugg);
}
} else {
err.span_label(span, format!("takes {}", args_str(found)));
}
}

err
}

Expand Down
4 changes: 4 additions & 0 deletions src/test/ui/mismatched_types/closure-arg-count.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,4 +16,8 @@ fn main() {
[1, 2, 3].sort_by(|tuple| panic!());
[1, 2, 3].sort_by(|(tuple, tuple2)| panic!());
f(|| panic!());

let _it = vec![1, 2, 3].into_iter().enumerate().map(|i, x| i);
let _it = vec![1, 2, 3].into_iter().enumerate().map(|i: usize, x| i);
let _it = vec![1, 2, 3].into_iter().enumerate().map(|i, x, y| i);
}
42 changes: 33 additions & 9 deletions src/test/ui/mismatched_types/closure-arg-count.stderr
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,16 +1,16 @@
error[E0593]: closure takes 0 arguments but 2 arguments are required
error[E0593]: closure takes 0 arguments, but 2 arguments are required
--> $DIR/closure-arg-count.rs:15:15
|
15 | [1, 2, 3].sort_by(|| panic!());
| ^^^^^^^ ----------- takes 0 arguments
| ^^^^^^^ -- takes 0 arguments
| |
| expected closure that takes 2 arguments

error[E0593]: closure takes 1 argument but 2 arguments are required
error[E0593]: closure takes 1 argument, but 2 arguments are required
--> $DIR/closure-arg-count.rs:16:15
|
16 | [1, 2, 3].sort_by(|tuple| panic!());
| ^^^^^^^ ---------------- takes 1 argument
| ^^^^^^^ ------- takes 1 argument
| |
| expected closure that takes 2 arguments

Expand All @@ -23,23 +23,47 @@ error[E0308]: mismatched types
= note: expected type `&{integer}`
found type `(_, _)`

error[E0593]: closure takes 1 argument but 2 arguments are required
error[E0593]: closure takes 1 argument, but 2 arguments are required
--> $DIR/closure-arg-count.rs:17:15
|
17 | [1, 2, 3].sort_by(|(tuple, tuple2)| panic!());
| ^^^^^^^ -------------------------- takes 1 argument
| ^^^^^^^ ----------------- takes 1 argument
| |
| expected closure that takes 2 arguments

error[E0593]: closure takes 0 arguments but 1 argument is required
error[E0593]: closure takes 0 arguments, but 1 argument is required
--> $DIR/closure-arg-count.rs:18:5
|
18 | f(|| panic!());
| ^ ----------- takes 0 arguments
| ^ -- takes 0 arguments
| |
| expected closure that takes 1 argument
|
= note: required by `f`

error: aborting due to 5 previous errors
error[E0593]: closure takes multiple arguments, but a tuple argument is required
--> $DIR/closure-arg-count.rs:20:53
|
20 | let _it = vec![1, 2, 3].into_iter().enumerate().map(|i, x| i);
| ^^^ ------ help: consider changing to: `|(i, x)|`
Copy link
Contributor

@estebank estebank Oct 10, 2017 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could you modify the suggestion text to be freestanding without the snippet? Something along the lines of consider changing the closure to accept a tuple. rustc is not the only way these messages can be presented, so they should make sense even when the suggested code is not presented next to it.

| |
| expected closure that takes 1 argument, a 2-tuple
Copy link
Contributor

@nikomatsakis nikomatsakis Oct 10, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is looking really good! I hope you don't mind me nitpicking some more though. =)

Reading this "expected closure that takes 1 argument, a 2-tuple" -- which I think I may even have initially suggested -- I see now that it is actually ambiguous. In particular, the intention was something like "that takes 1 argument, which is a 2-tuple", but it could also be parsed as a list, like "that takes 1 argument and a 2-tuple". In fact, this second interpretation is what I got this time, and I had to re-read it a few times to understand what was meant.

Therefore, I suggest:

"expected closure that takes a single tuple as argument"

which echoes the wording from above


error[E0593]: closure takes multiple arguments, but a tuple argument is required
--> $DIR/closure-arg-count.rs:21:53
|
21 | let _it = vec![1, 2, 3].into_iter().enumerate().map(|i: usize, x| i);
| ^^^ ------------- help: consider changing to: `|(i, x): (usize, _)|`
Copy link
Contributor

@nikomatsakis nikomatsakis Oct 9, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this looks great

| |
| expected closure that takes 1 argument, a 2-tuple

error[E0593]: closure takes multiple arguments, but a tuple argument is required
--> $DIR/closure-arg-count.rs:22:53
|
22 | let _it = vec![1, 2, 3].into_iter().enumerate().map(|i, x, y| i);
| ^^^ --------- takes 3 arguments
| |
| expected closure that takes 1 argument, a 2-tuple

error: aborting due to 8 previous errors