comparison with x.as_str() where x suffices

[llogiq]

We recently had some instances where we compared name.as_str() == s albeit there was an impl PartialEq<&str>.

I say this should either be allowed by default or use a blacklist of types where we know that the aforementioned impl has the same semantics as the conversion + str-equality.

[Manishearth]

+1 for a blacklist. If someone can come up with a sound way (no false positives) of doing it that's more precise, I'd love to hear it (matching on an as_str is as precise as the blacklist)

[llogiq]

The only possible sound way would be to prove that .as_str() is homomorphic regarding equality (∀x:X where x.as_str() -> &str.∀s:&str.x.as_str() == s ↔ x == s).

While this may be easily solveable for a small subset (where PartialEq<&str> is implemented by calling as_str() first, then comparing), it is impossible to solve in the general case, and probably quite unwieldy to solve automatically in most useful cases.

[llogiq]

We should probably blacklist (type, method_name) tuples.

[Manishearth]

sgtm

[llogiq]

Thinking a bit more about this, a generic (type, method_name, message) tuple + some code to match could subsume a good number of lints we currently have. We'd just need a way to optionally insert snippets into the message...