Add bytes to type reference helper methods to FromBytes

[kupiakos]

One common complaint I receive is how onerous it is to convert from bytes to type references in zerocopy for simple cases. I fully agree; we don't use the Ref type in most of our code. AsBytes::as_bytes makes it easy to go from &T to &[u8], so why is there such a dance to go the other way?

These helper methods intend to improve the developer experience. They follow the same naming and error handling scheme as read_from. I would rather these stay named as such until a refactor of read_from and write_to occurs to match - I find that consistency valuable.

These avoid a tuple return in from_[suffix|prefix] - again for developer UX since we usually threw these out.

    fn ref_from(bytes: &[u8]) -> Option<&Self> where Self: Sized;
    fn ref_from_prefix(bytes: &[u8]) -> Option<&Self> where Self: Sized;
    fn ref_from_suffix(bytes: &[u8]) -> Option<&Self> where Self: Sized;

    fn mut_from(bytes: &mut [u8]) -> Option<&mut Self> where Self: Sized + AsBytes;
    fn mut_from_prefix(bytes: &mut [u8]) -> Option<&mut Self> where Self: Sized + AsBytes;
    fn mut_from_suffix(bytes: &mut [u8]) -> Option<&mut Self> where Self: Sized + AsBytes;

[joshlf]

Awesome, thanks for this! Taking a look now...

Do y'all actually make use of from_[suffix|prefix]? I'm curious what your use cases are since I generally have found that those APIs have very few users.

[kupiakos]

Do y'all actually make use of from_[suffix|prefix]?

We don't use suffix like ever, but we do use prefix. We rarely check the exact size, just that it's large enough.
I'm supportive of removing _suffix, but I do like it there for consistency.

[joshlf]

Do y'all actually make use of from_[suffix|prefix]?

We don't use suffix like ever, but we do use prefix. We rarely check the exact size, just that it's large enough. I'm supportive of removing _suffix, but I do like it there for consistency.

No objection to supporting both; I'm just curious what your use cases are.

[joshlf]

Looks good! A few small nits.

I'm going to hold off on approving until we resolve the discussion in #524, and I also want to wait until some other folks have a chance to chime in who aren't working this week. I expect we can have this all resolved and merged early next week.

[kupiakos]

No objection to supporting both; I'm just curious what your use cases are.

There are probably valid use cases for _from_suffix, maybe for dynamic data, but it's just never come up for us from what I've seen.

It's common for a subsystem like TPM to receive a buffer that is larger than the message it contains when the request buffer is shared for the response. _from_prefix is particularly useful there.

[joshlf]

We had another request for this, so I'm going to land it and follow up on the outstanding comments in a follow-up PR.

[joshlf]

This has been published in 0.7.18.