[as_conversions] No alternative exists to float to/from int conversion using `as`

[casey]

The as_conversions lint warns if the as operator is used for int-to-float and float-to-int conversions, however I don't believe that there is actually an alternative.

It might be good to relax the lint so it doesn't complain if either argument is a float until an alternative exists.

[flip1995]

IMO it is an restriction lint, that should warn exactly about this. A float-to-int or int-to-float conversion is just such a dangerous conversion, that can be silently lossy.
Playgroun

Allowing this, would defeat the purpose of this lint. For the cases, where into/from exists, we have the cast_* lints.

Side note: restriction lints are called like this, because they "restrict" the language, without claiming that the thing is per se bad. In this case the lint removes as from the language and therefore restricts the language.

[casey]

I think that's fair, but looking at the cast_* lints, it's not immediately obvious to me that there's a set of lints which would warn me about casts where a better alternative exists, i.e. try_from and from, but not warn me about casts where no alternative exists. Or am I not seeing it?

I.e. it seems reasonable to want 0u8 as 0u16 to be disallowed (into exists), 0u16 as 0u8 to be disallowed (try_into exists), but u8 as f64 to be allowed, since there's no alternative, but no combination of lints can give you that.

[flip1995]

That is because an int-float-conversion has no alternatives, so we can't warn on int as float code, since it cannot be improved. This is where this lint comes into play. Just disallow every as conversion and if you are sure, that the conversion you're doing is save, add an #[allow(...)]. Kind of like unsafe.

[casey]

I guess that's reasonable. Thanks for the help!