Add FromIntoRef and TryFromIntoRef

[oblique]

Conversions from references are rare, however a user who controls the implementation of the type, can implement Into<T> from &'a MyType and use this to avoid cloning.

[codecov]

Codecov Report

Merging #618 (babfaeb) into master (e096f0b) will increase coverage by 0.58%.
The diff coverage is 97.29%.

@@            Coverage Diff             @@
##           master     #618      +/-   ##
==========================================
+ Coverage   63.63%   64.22%   +0.58%     
==========================================
  Files          35       36       +1     
  Lines        2090     2127      +37     
==========================================
+ Hits         1330     1366      +36     
- Misses        760      761       +1     

Impacted Files	Coverage Δ
serde_with/src/de/impls.rs	58.66% <80.00%> (+0.21%)	⬆️
serde_with/src/ser/impls.rs	91.66% <100.00%> (+0.36%)	⬆️
serde_with/tests/serde_as/fromintoref.rs	100.00% <100.00%> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[jonasbb]

Thanks for the contribution. I think I see the utility. Can you please confirm: the new types are almost identical to FromInto and TryFromInto, except that the SerializeAs trait bound does not require Clone by instead including the reference in the Into/TryInto bounds?
Do you have a use case for which you need this?

However, the “borrow” name part seems wrong here. In the serde ecosystem, borrow usually refers to borrowing during deserialization. The new types do not enable borrowing during serialization either, they only remove the Clone requirement. The reason is that for<'a> &'a T: Into<U> does not allow U to borrow from T. What the for<'a> means is any lifetime, no matter how short. Therefore, for any lifetime of U you can find a smaller 'a, thus U cannot borrow from T.
I don't have any name suggestions right now. Do you have any idea?

[oblique]

Can you please confirm: the new types are almost identical to FromInto and TryFromInto, except that the SerializeAs trait bound does not require Clone by instead including the reference in the Into/TryInto bounds?

Yes I confirm the above.

Do you have a use case for which you need this?

I'm working on a WASM project that has many serializations of data that resides in heap and I wanted to minimize cloning whenever I can.

However, the “borrow” name part seems wrong here.
.....
I don't have any name suggestions right now. Do you have any idea?

I agree borrow doesn't feel right. I didn't know what name to choose. These are some ideas:

1. FromRefInto/TryFromRefInto
2. RefFromInto/RefTryFromInto
3. FromIntoRef/TryFromIntoRef
4. FromIntoNonCloning/TryFromIntoNonCloning

[jonasbb]

Thanks for the info. I like FromIntoRef/TryFromIntoRef the most. Could you please update the name? The rest looks good. You have tests and documentation written. The only thing missing is a changelog entry, but I can create that when releasing the crate.

[oblique]

I changed the names. I also changed the first line of the doc comments. Let me know if you prefer other wording there.

[jonasbb]

Thank you for the changes. This looks good now.

bors r+

[bors]

Build succeeded!

The publicly hosted instance of bors-ng is deprecated and will go away soon.

If you want to self-host your own instance, instructions are here.
For more help, visit the forum.

If you want to switch to GitHub's built-in merge queue, visit their help page.

• Rustsec Audit
• Tests result