Add UUID as a well-known string type

[nichegriffy]

This should resolve #61

As the title states, this PR adds uuid as a well-known string type that verifies it looks like a UUID per RFC 4122.

I decided to add one uuid validator rather than version-specific validators because I figured the most common use case would be verifying that the string looks like a UUID rather than a particular version, as well as to be able to support the special nil UUID.

[rmichela]

Look good to me. There are a few conflicts that need resolving.

[nichegriffy]

@rmichela Thanks! I resolved the conflicts, so this should be ready for another review now 🔍

[rodaine]

Awesome! Thanks for the work here :D

[Fleshgrinder]

@rmichela and @rodaine

What about the simple format that basically saves 4 bytes?

Also, what is the general policy regarding performance? Regular expressions — especially in Java — are extremely slow compared to a custom tailored and hard coded state machine.

I didn’t want to create a dedicated issue for either of these questions because how to proceed is basically depending too much on the answers and it seems fitting to ask here.

[rmichela]

PGV uses the Google RE2J regex library instead of java.util.regex.

RE2 is a regular expression engine that runs in time linear in the size of the input.

[akonradi]

That's true for Go (because that's the implementation in the standard library), but the C++ code doesn't implement regexes and needs #22 to get resolved before implementation. I'm personally on board with using RE2 everywhere but want more buy-in.

As for regexes for parsing by hand, my experience has been that switching from RE2 to hand-written parsing makes for a 4-10x performance improvement, at least in C++. For fixed patterns like for UUIDs, it would make sense to implement the same logic by hand.

[rmichela]

I created a new issue so we don't lose track of the work.

[nichegriffy]

@Fleshgrinder

I did see the mention of the simple format without hyphens in the original issue and was going to support it, but the grammar for how to represent a UUID in string form in RFC 4122 does seem to require the hyphens: https://tools.ietf.org/html/rfc4122#section-3

With that said, the decision to strictly adhere to RFC 4122 was my own and this project may prefer to support more representations, but I wanted to give context for why I left it out in this PR.

[Fleshgrinder]

I’m with @akonradi here and happy to provide a benchmark for Java. There are objects being created in the background and they get persisted throughout the lifetime of the whole program.

@nichegriffy yeah it does and I guess that most people will also prefer to use it with hyphens because it’s the best way to distinguish it from some random hex string but I thought I’d ask to gauge whether there’s interest or not. 😉

[rmichela]

See #222