suggest a op= b over a = a op b

[oli-obk]

since not everyone likes the assign-ops, I also added a restriction lint to forbid assign-ops. Can we somehow make lints depend on each other? like lint if both assign_ops and assign_op_pattern are active?

[llogiq]

If they want all the warnings, give them all the warnings 😄

[llogiq]

Another thought: Maybe differentiate ops where the assign op is specialized, because those likely improve performance, but may also change behavior.

[oli-obk]

Maybe differentiate ops where the assign op is specialized because those likely improve performance

That's what we want

but may also change behavior.

shudder Maybe I should make an RFC that states that the assign ops should be logically equivalent to the full expression. Is there any reason why they shouldn't?

[llogiq]

The original RFC that introduced the OpAssign traits already states exactly that. However, it is really hard to guarantee. Someone has probably already written a quickcheck template, though.

[oli-obk]

However, it is really hard to guarantee.

we don't need to guarantee it, we can depend on it. If anything, this lint will expose such oddities

[oli-obk]

I added your concern to the known problems field of the docs

[llogiq]

Couldn't we detect if the type impls Op and avoid reporting otherwise?

[oli-obk]

We can, but the point of the restriction is to forbid all assign ops, no matter if there are no alternatives. What I can do is add a lint to detect assign op impls withou an op impl

[llogiq]

Ok then. LGTM.