Cache miss async rendering

[88kami88]

When using an async cache strategy and there's a cache miss, segments were being rendered out of order or not at all. This is caused by the fact that the cache segment is returning a promise, but the sequence continued to allow the emission of segments before the cache had resolved.

This PR makes almost everything a promise and builds a chain of promises where the next item in the sequence must wait for the previous item to render.

@divmain @maxgalbu

Resolves #87

TODO:

• [x] tests
• [x] benchmarks

rapscallion{cache-miss}$ npm run benchmark

> rapscallion@2.1.7 benchmark /Users/jkaminetsky/dev/oss/rapscallion
> babel-node benchmark/benchmark.js

Starting benchmark for 10 concurrent render operations...
renderToString took 11.445220171 seconds
rapscallion, no caching took 43.507448408 seconds; ~0.26x faster
rapscallion, caching DIVs took 8.118688945 seconds; ~1.4x faster
rapscallion, caching DIVs (second time) took 8.136480110 seconds; ~1.4x faster
rapscallion, caching Components took 0.644866469 seconds; ~17.74x faster
rapscallion, caching Components (second time) took 13.052715387 seconds; ~0.87x faster
rapscallion (pre-rendered), no caching took 24.778552483 seconds; ~0.46x faster
rapscallion (pre-rendered), caching DIVs took 8.361157752 seconds; ~1.36x faster
rapscallion (pre-rendered), caching DIVs (second time) took 7.835752739 seconds; ~1.46x faster
rapscallion (pre-rendered), caching Components took 13.324269605 seconds; ~0.85x faster
rapscallion (pre-rendered), caching Components (second time) took 13.924382979 seconds; ~0.82x faster

[88kami88]

@jvivs Check this out.

@divmain

[maxgalbu]

Something's not right in that benchmark, how can the second pass with cached components take longer than the first pass of caching?

[divmain]

I'm going to want to take a look at this more closely. Those benchmarks indicate a significant performance regression. There might be a better-performing solution to the issues raised here.

[88kami88]

My first guess would be overhead from naive promise implementation, specifically waiting for all segments to resolve in series, even when there's not an async cache get.

[clifton]

I'm not sure how much you should index to the benchmark when async caching doesn't seem to be working properly in the first place. ex: https://github.com/Rymar/rapscalion-cache is the exact problem I'm finding when trying to do async caching in a production application

[88kami88]

Honestly the benchmarks 2.1.7 don't look much different:

renderToString took 11.486885729 seconds
rapscallion, no caching took 43.693205031 seconds; ~0.26x faster
rapscallion, caching DIVs took 16.163418342 seconds; ~0.71x faster
rapscallion, caching DIVs (second time) took 14.280983280 seconds; ~0.8x faster
rapscallion, caching Components took 0.701809205 seconds; ~16.36x faster
rapscallion, caching Components (second time) took 15.668481997 seconds; ~0.73x faster
rapscallion (pre-rendered), no caching took 35.028802432 seconds; ~0.32x faster
rapscallion (pre-rendered), caching DIVs took 18.111153076 seconds; ~0.63x faster
rapscallion (pre-rendered), caching DIVs (second time) took 17.383266673 seconds; ~0.66x faster
rapscallion (pre-rendered), caching Components took 16.657128768 seconds; ~0.68x faster
rapscallion (pre-rendered), caching Components (second time) took 16.814332255 seconds; ~0.68x faster```

[88kami88]

Now back on this cache-miss branch, executed immediately afterwards with similar load on same machine:

renderToString took 11.203203946 seconds
rapscallion, no caching took 41.365911048 seconds; ~0.27x faster
rapscallion, caching DIVs took 15.757512965 seconds; ~0.71x faster
rapscallion, caching DIVs (second time) took 13.955430961 seconds; ~0.8x faster
rapscallion, caching Components took 0.529333810 seconds; ~21.16x faster
rapscallion, caching Components (second time) took 12.946540060 seconds; ~0.86x faster
rapscallion (pre-rendered), no caching took 25.377965259 seconds; ~0.44x faster
rapscallion (pre-rendered), caching DIVs took 13.801301708 seconds; ~0.81x faster
rapscallion (pre-rendered), caching DIVs (second time) took 13.337650601 seconds; ~0.83x faster
rapscallion (pre-rendered), caching Components took 13.305775981 seconds; ~0.84x faster
rapscallion (pre-rendered), caching Components (second time) took 12.985989510 seconds; ~0.86x faster

It might actually be faster, granted the smallest of sample sizes in a personal computing environment.

[divmain]

I pulled this down and you're right - the benchmarks don't look much difference between v2.1.7 and your branch, although it seems to perform much better against React on my machine.

@ethersage, can you clean up the merge conflicts? I'll follow up with publishing a new version.

[88kami88]

I will look at merge conflicts

[88kami88]

@fleg Please review where I've tagged you to make sure I merged correctly.

[fleg]

lgtm, what about tests?

[88kami88]

Tests look good to me.

[divmain]

Alrighty, I think this is ready to ship. Thanks for your work on this!