Missing Eq implementations for types implementing PartialEq

[whitequark]

Specifically (I looked at everything in std ):

• std::num::ParseIntError
• std::str::ParseBoolError
• std::num::FpCategory

There seems to be no good reason for omitting Eq there, just an oversight. All other similar types already implement Eq.

[hanna-kruppe]

A quick grep suggests there are around 300 types that derive PartialEq but not Eq in all of src/, though I don't have a quick way to tell how many of those can implement Eq. But I bet at least a few dozen can and should. I wish there was an easy way to run clippy on all the code.

[whitequark]

That is really not an accurate way of measuring it because it takes into account all of the rustc-internal types, and worse, tests, for which it probably doesn't matter much whether they implement Eq or not. Here's a nicer way:

git grep '#\[derive' | \
  egrep -v '^(test|doc|libcollectionstest|libcoretest|tools|librust|libunwind|libsyntax|libtest|libterm|librbml|libserialize|libfmt_macros)' | \
  grep PartialEq | \
  grep -v '\bEq'

This produces only 29 examples, some of which implement Eq without deriving it:

libcore/char.rs:733:#[derive(PartialEq, Debug)]
libcore/cmp.rs:187:#[derive(Clone, Copy, PartialEq, Debug, Hash)]
libcore/fmt/num.rs:103:#[derive(Clone, PartialEq)]
libcore/fmt/num.rs:107:#[derive(Clone, PartialEq)]
libcore/fmt/num.rs:111:#[derive(Clone, PartialEq)]
libcore/fmt/num.rs:115:#[derive(Clone, PartialEq)]
libcore/fmt/num.rs:119:#[derive(Clone, PartialEq)]
libcore/fmt/rt/v1.rs:34:#[derive(Copy, Clone, PartialEq)]
libcore/num/dec2flt/mod.rs:159:#[derive(Debug, Clone, PartialEq)]
libcore/num/dec2flt/mod.rs:165:#[derive(Debug, Clone, PartialEq)]
libcore/num/flt2dec/decoder.rs:26:#[derive(Copy, Clone, Debug, PartialEq)]
libcore/num/flt2dec/decoder.rs:43:#[derive(Copy, Clone, Debug, PartialEq)]
libcore/num/mod.rs:2414:#[derive(Copy, Clone, PartialEq, Debug)]
libcore/num/mod.rs:2757:#[derive(Debug, Clone, PartialEq)]
libcore/num/mod.rs:2761:#[derive(Debug, Clone, PartialEq)]
libcore/str/mod.rs:112:#[derive(Debug, Clone, PartialEq)]
liblog/lib.rs:233:#[derive(Copy, Clone, PartialEq, PartialOrd, Debug)]
librand/distributions/mod.rs:272:    #[derive(PartialEq, Debug)]
libstd/collections/hash/table.rs:138:#[derive(PartialEq, Copy, Clone)]
libstd/error.rs:461:    #[derive(Debug, PartialEq)]
libstd/error.rs:463:    #[derive(Debug, PartialEq)]
libstd/ffi/c_str.rs:158:#[derive(Clone, PartialEq, Debug)]
libstd/ffi/c_str.rs:164:#[derive(Clone, PartialEq, Debug)]
libstd/ffi/c_str.rs:170:#[derive(Clone, PartialEq, Debug)]
libstd/net/mod.rs:43:#[derive(Copy, Clone, PartialEq, Debug)]
libstd/net/parser.rs:375:#[derive(Debug, Clone, PartialEq)]
libstd/path.rs:461:#[derive(Copy, Clone, PartialEq, PartialOrd, Debug)]
libstd/sync/mpsc/select.rs:106:#[derive(PartialEq)]
libstd/sys/windows/pipe.rs:168:#[derive(PartialEq, Debug)]

[Mark-Simulacrum]

rg 'derive.*PartialEq' | rg -v Eq returns nothing, so at least those types with derived equivalence seem to all implement both. I'm not sure how many types (or what types) that implement these traits manually don't follow this though. I think it might be possible to add such a thing as a lint on the rustdoc-like level once it's reimplemented atop the RLS api.

[dtolnay]

I checked the search from #36301 (comment) and it looks like all the public types that should implement PartialEq now do. Fixed!