Use shallowEqual from fbjs

[timdorr]

Fixes #590

Also handles things like NaN.

[jimbolla]

We should perf test this, since it's called so much. It's useful to remember that we're not comparing any 2 rando values, but specifically props objects coming from react. I think we can make certain assumptions, like that they're objects, and typically (99% of time?) that the # of props is going to be the same between the 2 args. So the algorithm can optimize for things like that.

[timdorr]

True, but I think the effects are minor. Given this is a hot path, it's going to get optimized pretty heavily. I think the only thing to do is to use the for in loop form instead, as v8 has specific optimizations for that with a hasOwnProperty check. Though it probably won't trigger that with the equality check in there. Maybe if we unroll that into another if?

[jimbolla]

If the perf of the fbjs version is fine, then maybe we should just import it from fbjs. fbjs is a dependency of react, so react-redux has a dependency on it indirectly already.

[timdorr]

AFAIK, shallowEqual hasn't changed in quite some time in fbjs. I'm going to merge this as-is and we can look at switching it to an import at a later date.

[jimbolla]

Works for me. This should be fast enough for normal usage. And if someone really needs a faster implementation than the fbjs, they have the power to override it in the options.

[gaearon]

I wouldn't recommend depending on anything from fbjs directly.

[gaearon]

(Copying is fine 😄 )

[wtgtybhertgeghgtwtg]

This seems like it happens often; recompose just has fbjs as a dependency for shallow equal, and I think react-side-effect used to do that. Do you think fbjs could move their shallow equal implementation into another library? Copying doesn't seem like the best solution, either.