RUST-687 Support deserializing binary of UuidOld

[kenkoooo]

Since I have a lot of existing binary data of subtype=0x03, which is a subtype of old UUID, I added support to deserialize such binary data to UUID.

[pooooodles]

Tracked in https://jira.mongodb.org/browse/RUST-687

[patrickfreed]

Hi @kenkoooo! Thank you or the PR, and thank you @jcdyer for bringing up the UUID spec! I think this implementation looks great so far, but I think we'll want to go for a slightly different API to be consistent with how we currently handle regular subtype 4 UUIDs.

By that I mean, instead of separate types for each of the different UUIDs, can we change this to introduce separate serde_helper modules for each? This matches the current UUID subtype 4 API.

e.g.

#[derive(Serialize, Deserialize)]
struct MyUuids {
    // This is the existing UUID helper
    #[serde(with = "serde_helpers::uuid_as_binary")]
    regular: Uuid,

    // Below would be the new ones for each legacy type
    #[serde(with = "serde_helpers::uuid_as_java_legacy_binary")]
    java_legacy: Uuid,

    #[serde(with = "serde_helpers::uuid_as_python_legacy_binary")]
    python_legacy: Uuid,

    #[serde(with = "serde_helpers::uuid_as_c_sharp_legacy_binary")]
    c_sharp_legacy: Uuid,
}

Check out serde_helpers.rs for an example implementation for the subtype 4 version.

For datetimes, we originally did go the route of introducing a new type (bson::DateTime) which wraps the regular chrono::DateTime and has special Serialize and Deserialize implementations, but users tended to find it confusing / hard to use / hard to discover. As a result, we opted to take the serde_helper approach for Uuid instead as it is a bit more explicit / configurable, so I think we should continue to use that pattern for these legacy variants too.

[pooooodles]

Hey @kenkoooo just wanted to drop a comment to make sure you saw Patrick's feedback, are these changes you would like to make or would you prefer we push up changes to this branch ourselves? Thanks for all the hard work!

[kenkoooo]

@agolin95 Thank you for your confirmation. Let me try to implement as @patrickfreed suggested.

[kenkoooo]

Hi, sorry for the very slow work. I update my PR as @patrickfreed suggested. Could you take a look into it? Thanks!

[patrickfreed]

These changes look great! I only have a few minor comments:

• Can we update the module names to include _0_8_ after uuid? We're versioning these helpers in order to support multiple semver-breaking versions of uuid as they come out. See #255 for more info (you can leave the exising uuid_as_binary one alone, that'll be updated in that PR)
• Can we add top level re-exports of the serialize and deserialize functions similar to what we have for the other ones? Check the top of the file for the pattern there.
• There are a few clippy failures, it looks like there are a few unneeded clones of uuid's in src/tests/serde.rs. Don't worry about the decimal128 ones--those are pre-existing.

Thanks again for your contribution!

[kenkoooo]

@patrickfreed Thanks for your review! I updated it.

[patrickfreed]

Looks like rustfmt needs to be run, but other than that it looks good to me! (note: we use the nightly version of rustfmt).

tagging in @isabelatkinson for review as well.

[isabelatkinson]

LGTM once the formatting errors are fixed, thanks for your contribution!

[kenkoooo]

@isabelatkinson @patrickfreed Thanks! I reformat by cargo +nightly fmt.

[patrickfreed]

LGTM! Thank you for your contribution @kenkoooo !