Lifetime bounds for `std::str::strip_prefix` should be relaxed

1

Currently the bounds for std::str::strip_prefix are:

pub fn strip_prefix<'a, P>(&'a self, prefix: P) -> Option<'a str>
    where P: Pattern<'a>

But it doesn't really make any sense that the output's lifetime should depend on the lifetime of the pattern. Something like this would make more sense:

pub fn strip_prefix<'s: 'p, 'p, P>(&'s self, prefix: P) -> Option<'s str>
where 
    P: Pattern<'p>

since the Pattern really only needs to stay alive inside the function call itself.

I'm currently dealing with an issue in which a function takes a &mut &str and a Pattern<'static>, but the borrow checker is mad that the &str isn't 'static as well. I haven't found anyone else complaining about this, but I'm sure I'm not the first person who's wanted this changed

2

Pattern<'a> doesn't mean the lifetime of the pattern itself is 'a, but rather of the string we match against.

3

I think the bounds are correct:

So anything requiring Pattern<'static> does require &'static str. I think whatever API you're working with needs to be updated to allow shorter lifetimes.

4

"Can only work on 'static haystacks" is exactly what P: Pattern<'static> means.

Perhaps you or whomever imposed the Pattern<'static> constraint meant P: for<'any> Pattern<'any>.

1 Like
5

Ah gotcha. I probably messed up the bounds on my end somehow

6

Uh...yes. Somehow you fixed my problem without even looking at the code.

In case anyone is looking at this in the future and doesn't understand this like I didn't, the issue is that in fn foo<'a>(pat: impl Pattern<'a>), pat has exactly one lifetime, whereas in fn<P> foo(pat: P) where P: for<'a> Pattern<'a>, the Rust compiler treats pat like it is compatible with all lifetimes at once, meaning it will be compatible with any &str, whether that's a &'a str, &'short str, or a &'static str

3 Likes
7

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.