Should we have FromVarULE?

[sffc]

Part of #1082

Split from #1179

If we want to support cases where the target type borrows from the VarULE, FromVarULE can work (playground):

trait FromVarULE<'a, U: ?Sized>: 'a {
    fn from_var_ule(ule: &'a U) -> Self;
}

impl FromVarULE<'_, str> for String {
    fn from_var_ule(ule: &str) -> Self {
        ule.to_string()
    }
}

impl<'a> FromVarULE<'a, str> for Cow<'a, str> {
    fn from_var_ule(ule: &'a str) -> Self {
        Cow::Borrowed(ule)
    }
}

Thoughts? Needs input from:

• @zbraniecki
• @Manishearth

[Manishearth]

Yeah so I do want this, but my general thoughts are that this function will basically be called by users not the VZV/ZV APIs, so it's not important to have a trait for it unless we're doing a custom derive (which we should absolutely do eventually). My suggestion to Zibi for now was to implement this as an inherent method on the ULE type (.as_foo(), we have .as_varzerovec() on VZVULE for example). Happy to have a trait early so that we can document this though; it's personally low-priority for me since it can be done as an inherent method.

[sffc]

Would a method such as get_decoded be useful?

impl<'a, U> VarZeroVec<'a, U>
where
    U: VarULE + ?Sized
{
    fn get_decoded<T>(&'a self) -> Option<T>
    where
        T: FromVarULE<'a, U>
    {
        todo!()
    }
}

[Manishearth]

yes! it would need an index argument, but yeah that sounds useful

DecodeFromVarULE?

We can actually merge EncodeAs and DecodeFrom but there are types that are encodable but not decodable (e.g. String can't be "decoded" as a cheap borrow)

[sffc]

Would ZeroCopyFrom do the trick, instead of FromVarULE?

[Manishearth]

Yes, I think so

[Manishearth]

Though I think ZCF is a bit weird bc it requires Yokeable. I'll have to think about that

[sffc]

FromVarULE is a special case of ZCF: the case when the source (the Yokeable) has no lifetime parameter.

Let's not combine encode/decode like we did with the old AsVarULE. I prefer smaller traits when possible, and I think it results in a better outcome anyway.

[Manishearth]

Nah, we need the lifetime parameter I think, and we also need the Yokeable. That's somewhat unfortunate but it is what it is.

[Manishearth]

When we have GATs a lot of this should simplify, but that might be a ways off

[sffc]

Discussion: add FromVarULE as its own trait for now, and see if we can unify it with ZeroCopyFrom later in #1186.

[Manishearth]

Related: #1255 . If we can make that happen first then we won't need a separate trait

[Manishearth]

Now that ZCF is simpler I think we should just use ZCF here. It might be worth having the proc macro autogen the impl if asked, though!

[Manishearth]

I'm going to close this issue then