Refactor how we handle overflow and remove blanket Pattern impl #23580
Conversation
| @@ -474,22 +474,16 @@ impl<'a, 'b> Pattern<'a> for &'b [char] { | |||
| s, CharEqPattern(s)); | |||
| } | |||
|
|
|||
| /// Searches for chars that are equal to any of the chars in the array | |||
There was a problem hiding this comment.
This should probably be the doc comment of the &str impl instead, or rather explaining that it delegates to the &str impl.
|
Apart from the doc comments, the pattern impls look fine to me. |
|
Is it possible to still have a forwarding impl, but just for references? impl<'a, 'b, P: Pattern<'a> + ?Sized + 'b> Pattern<'a> for &'b P {
...
} |
|
@alexcrichton no, because any type |
nikomatsakis
force-pushed
the
pattern-and-overflow
branch
from
9d68f33 to
8ff0d74
Compare
|
rebased, tweaked docs per @Kimundi's suggestion |
|
@bors: r+ 8ff0d74 |
compilation: this removes all the ungainly code that special cases overflow so that we can ensure it propagates.
possible blanket impls and also triggers internal overflow. Add some special cases for common uses (&&str, &String) for now; bounds-targeting deref coercions are probably the right longer term answer.
nikomatsakis
force-pushed
the
pattern-and-overflow
branch
from
8ff0d74 to
76ead08
Compare
|
@bors r=aturon 7dead08 |
|
🙀 |
|
I think the overflow refactoring here might be what @aturon was referring to in the issue I reported at #23714. The overflow/failure on evaluating the |
See comments in the commits for reasoning
r? @aturon
cc @Kimundi