Refactor Rifm into useRifm hook and Rifm component

[lewisl9029]

To address #38

Happy to add some tests specifically for the hook if desired, but it
didn't feel super vital atm as all the code paths in the hook are exercised
by the existing render props tests.

Also happy to help in updating docs as well.

I have these changes currently published as @lewisl9029/rifm@0.02 in case anyone wants to try it out.

[istarkov]

+ 100+ bytes for production to just make it as hook. Rifm is very low level and usually (always) is used inside controls, NumberInput etc. So its very different from hooks I use everyday like useState etc.
Here in issues proposal about hooks was from the first day hooks were announced and we have no decision to be or not to be as there is no real use case when hook Rifm version is better vs render props.
Without changing the size of the production build I see nothing bad to accept this now, but size has changed significantly so it should be real - real use case to accept.

[TrySound]

There is a way to reduce the size. The source of problem here is rest operator { children, ...args } which forces babel to add helper. To prevent this you may not use rest and instead pass props as is to useRifm. Flow will fail. As a workaround cast props to any like this useRifm((props: any )). We do not iterate over props so it's safe to do it in place.

[istarkov]

@TrySound Can we after somehow measure the impact size of using hook only version having that Rifm export could be treeshaked? As with suggestion above hook only version can be smaller vs render props?

[TrySound]

843 -> 834

[lewisl9029]

Great catch! I tried @TrySound's suggestion and it reduced the gzipped size to 861, an 18 bytes increase. Is that more acceptable @istarkov?

I think it might also be worthwhile to consider a v2 that provides only a hook, since it's fairly trivial for users to build a Rifm render props component from a useRifm hook (i.e. the 2 lines implemented here), but as a user, I personally couldn't figure out any good ways to convert the Rifm component into a hook, hence this PR. But that's a much bigger decision that's outside of the scope of this PR. :)

[TrySound]

Hook only will be probably v1. But we will keep render prop for now to make soft upgrade without rewriting everything.

[istarkov]

Great!
So hook version tests left. Everything here is covered with tests.
Additional tests are:
Copy paste of
RifmFormat.test.js -> + RifmFormatHook.test.js
RifmMask.test.js -> + RifmMaskHook.test.js

And additional createExec method createExecHook with hook inside instead of render function
(Same as createExec just with hook)
https://github.com/realadvisor/rifm/blob/master/tests/utils/exec.js#L23

[istarkov]

Also it will be good to add few hook version examples

[TrySound]

@istarkov I would just refactor original tests with hook version. Render prop is a thing wrapper without any logic.

[istarkov]

Possibly you are right, I have no strong opinion on this

[TrySound]

@lewisl9029 Upgrade readme with hook version please and we will merge.

[lewisl9029]

@TrySound @istarkov I gave it a shot here: e1b3846

Let me know what y'all think, and feel free to make any edits you feel is appropriate.

[TrySound]

thanks!