Acute AccentApr 20, 2017

Doing partial prod builds under controlled conditions for comparison purposes. Cold caches, no shrinkwrap, 3 different package.json in a monorepo-type thing, 1797 dependencies installed, only npm install times shown:

- npm4: 406.01 seconds
- npm5: 60.96 seconds

This is a bit less speedup than what I got excited about yesterday, but still (a bit) more than 6 times faster! 🎆 🍻 ✨

cc @zkat, @[email protected]

Show threadAcute Accent

Also, because I got asked this on twitter. Same setup as original tweet (cold cache, no lockfile, etc):

- yarn: 86.82 seconds

But as discussed (in various places previously): npm & yarn have different tradeoffs and models, and both are valid choices! In the end they both use the same npm registry :)

cc @dx @zkat

Apr 20, 2017 at 4:05amWeb
Show thread✨ Kat ✨Apr 20, 2017
@passcod @dx buh! I'd say the most fair comparison when it comes to npm@5 is to generate yarn.lock and npm-shrinkwrap.json, and _then_ clear your cache. Neither tool really shines without the lockfile.
Show threadAcute AccentApr 20, 2017
@zkat @dx Yeah, but this build is *really* not set up for that. I think I'll do a more rigorous set of comparisons with various real codebases (that should support lockfiles better) when npm5 comes out so I can put it out in a proper blog post.
Show thread✨ Kat ✨Apr 20, 2017
@passcod @dx in the end, your setup happens to be a realistic one. This is pretty important to support, too, even if it's not the optimal repo target for the tools.
Show thread✨ Kat ✨Apr 20, 2017
@dx @passcod and seriously, thanks a ton for taking the time to do this.
Show threadAcute AccentApr 20, 2017
@zkat @dx it was pretty fun!