CSS class names are a function of the style object

[charliermarsh]

Resolves #32.

I used sha1 to mirror the way we've done this in the past, but I think that we should replace it with either a fast, non-cryptographically secure hash (like murmur), or even a checksum algorithm (like Adler-32--which React uses to compare client- and server-side markup (granted, collisions there are less important than they are here)--or CRC32). Thoughts?

Another thing to note: before, if we'd already seen styleObj in the global registry, we didn't return anything, which isn't really expected behaviour. Now, we return a style object regardless of whether or not it's in the registry (with a shallow clone).

[Raynos]

👍 I want the hash to be content addressable.

One way to do server side and client side RCSS is to generate a bunch of server side class names from the style object in a deterministic fashion.

Then on the client side you can query the DOM for a list of class names that exist and only append a class name to the DOM for RCSS() calls if it hash does not exist yet.

[chenglou]

Cool let's do this then. Can you update the tests though?

[charliermarsh]

@chenglou Tests updated.

[chenglou]

Thanks! In the future, we can have our own optimized JSON.stringify, since we know all the recognized style keys.

[charliermarsh]

@chenglou No problem. I'm curious as to whether we can actually do better than JSON.stringify given native browser support... Want to create an issue and we can start trying things out?