Add track_caller to iterator methods

[xxchan]

I'm trying to add track_caller to Itertools::zip_eq, but:

zip_eq(&a, &b).count();

reports Iterator::fold

thread 'zip_eq_panic1' panicked at 'itertools: .zip_eq() reached end of one iterator before the other', /rustc/9d1b2106e23b1abd32fce1f17267604a5102f57a/library/core/src/iter/traits/iterator.rs:2170:34
stack backtrace:
   0: std::panicking::begin_panic
             at /rustc/9d1b2106e23b1abd32fce1f17267604a5102f57a/library/std/src/panicking.rs:525:12
   1: <itertools::zip_eq_impl::ZipEq<I,J> as core::iter::traits::iterator::Iterator>::next
             at ./src/zip_eq_impl.rs:51:13
   2: core::iter::traits::iterator::Iterator::fold
             at /rustc/9d1b2106e23b1abd32fce1f17267604a5102f57a/library/core/src/iter/traits/iterator.rs:2170:29
   3: core::iter::traits::iterator::Iterator::count
             at /rustc/9d1b2106e23b1abd32fce1f17267604a5102f57a/library/core/src/iter/traits/iterator.rs:201:9
   4: zip::zip_eq_panic1
             at ./tests/zip.rs:66:5
   5: zip::zip_eq_panic1::{{closure}}
             at ./tests/zip.rs:61:1
   6: core::ops::function::FnOnce::call_once
             at /rustc/9d1b2106e23b1abd32fce1f17267604a5102f57a/library/core/src/ops/function.rs:227:5
   7: core::ops::function::FnOnce::call_once
             at /rustc/9d1b2106e23b1abd32fce1f17267604a5102f57a/library/core/src/ops/function.rs:227:5

zip_eq(&a, &b).collect_vec();

reports alloc::vec::...

thread 'zip_eq_panic1' panicked at 'itertools: .zip_eq() reached end of one iterator before the other', /rustc/9d1b2106e23b1abd32fce1f17267604a5102f57a/library/alloc/src/vec/mod.rs:2638:44
stack backtrace:
   0: std::panicking::begin_panic
             at /rustc/9d1b2106e23b1abd32fce1f17267604a5102f57a/library/std/src/panicking.rs:525:12
   1: <itertools::zip_eq_impl::ZipEq<I,J> as core::iter::traits::iterator::Iterator>::next
             at ./src/zip_eq_impl.rs:51:13
   2: alloc::vec::Vec<T,A>::extend_desugared
             at /rustc/9d1b2106e23b1abd32fce1f17267604a5102f57a/library/alloc/src/vec/mod.rs:2638:35
   3: <alloc::vec::Vec<T,A> as alloc::vec::spec_extend::SpecExtend<T,I>>::spec_extend
             at /rustc/9d1b2106e23b1abd32fce1f17267604a5102f57a/library/alloc/src/vec/spec_extend.rs:18:9
   4: <alloc::vec::Vec<T> as alloc::vec::spec_from_iter_nested::SpecFromIterNested<T,I>>::from_iter
             at /rustc/9d1b2106e23b1abd32fce1f17267604a5102f57a/library/alloc/src/vec/spec_from_iter_nested.rs:37:9
   5: <alloc::vec::Vec<T> as alloc::vec::spec_from_iter::SpecFromIter<T,I>>::from_iter
             at /rustc/9d1b2106e23b1abd32fce1f17267604a5102f57a/library/alloc/src/vec/spec_from_iter.rs:33:9
   6: <alloc::vec::Vec<T> as core::iter::traits::collect::FromIterator<T>>::from_iter
             at /rustc/9d1b2106e23b1abd32fce1f17267604a5102f57a/library/alloc/src/vec/mod.rs:2541:9
   7: core::iter::traits::iterator::Iterator::collect
             at /rustc/9d1b2106e23b1abd32fce1f17267604a5102f57a/library/core/src/iter/traits/iterator.rs:1745:9
   8: itertools::Itertools::collect_vec
             at ./src/lib.rs:1985:9
   9: zip::zip_eq_panic1
             at ./tests/zip.rs:66:5

Is it possible to improve it?

[the8472]

Closures can be executed pretty deep inside iterators, plumbing #[track_caller] through all possible call paths seems a bit much. It is very cheap, but not entirely free of cost since it sort of introduces an inlined wrapper function that passes the location information under the hood which makes life a tiny bit harder for the optimizers. It can lead to regressions when you put it on very hot code, e.g. #83359

[cuviper]

Stated another way: the zip_eq function itself is not panicking yet, only returning the lazy ZipEq iterator. The actual panic won't be raised until the iterator is advanced that far, which may happen under any arbitrary combination of iterator methods.

The similar change proposed in rayon-rs/rayon#977 does work because IndexedParallelIterator already knows its length, so its zip_eq can immediately assert that the two iterators have the same length.

[cuviper]

It would be interesting if the #[track_caller] fn zip_eq could directly access and store something like panic::Location within ZipEq, so that could be reported in the later panic message.

[cuviper]

Oh, I think you can do that with Location::caller() -- have you tried that?

[the8472]

The similar change proposed in rayon-rs/rayon#977 does work because IndexedParallelIterator already knows its length, so its zip_eq can immediately assert that the two iterators have the same length.

It might be possible to do this on a best-effort basis by checking whether there's any overlap between the size_hints.