emerentius/ord_subset

Tools for working with types where a subset of values has a total order, like e.g. floats without NaN

ord_subset's Issues

Republish last version due to spurious permissions in tarball

I see that last tarball on crates.io has all files with executable bit which creates some problems when packaging for Fedora.

I would appreciate if you would release new version where tarball would not have those files with exec perms. Thanks in advance!

Clarify unsafe vs "can panic"

In the documentation for ord_subset::OrdVar::new_unchecked it says:

Constructs an OrdVar without validity check. Incorrectly constructed OrdVars may panic during comparisons. This is the only way to construct an OrdVar out of a type that is not OrdSubset.

And it is marked unsafe. It is currently unclear what the unsafe behavior is here. panicing is not unsafe, so if comparisons either succeed or panic then probably nothing should be marked unsafe in this crate. If comparisons (or other functionality) can do something other than succeed or panic, then that behavior should be noted in the documentation along with info on how to ensure correct behavior when using the unsafe features.

Impl Hash for OrdVar if T provides Hash

This will enable PartialEq and PartialOrd types to be used as hash keys if they provide Hash.

warn(where_clauses_object_safety) on recent nightly

On recent nightlies, OrdSubsetSliceExt triggers the where_clauses_object_safety warning:

rust-lang/rust#51443

warning: the trait `slice_ext::OrdSubsetSliceExt` cannot be made into an object
  --> src/slice_ext.rs:42:5
   |
42 | /     fn ord_subset_sort(&mut self)
43 | |     where
44 | |         Self: AsMut<[T]>,
45 | |         T: OrdSubset;
   | |_____________________^
   |
   = note: #[warn(where_clauses_object_safety)] on by default
   = warning: this was previously accepted by the compiler but is being phased out; it will become a hard error in a future release!
   = note: for more information, see issue #51443 <https://github.com/rust-lang/rust/issues/51443>
   = note: method `ord_subset_sort` references the `Self` type in where clauses

(+3 other similar warnings)

$ rustc --version
rustc 1.29.0-nightly (6a1c0637c 2018-07-23)

I don't have time right now, but I'll try to come back to this in a week or so.

Implement ord_subset_sort_by_cached_key

I'm missing this, it would also be nice to have it to implement it to have a complete ord_subset mirror of the slice API