Repository files navigation

Compare thenpm install times forlocal-npm versus regularnpm, using some popular JavaScript repos.

• test-local-npm-speed
• Summary-nolanlawson/tiny-queue-lodash/lodash-substack/node-browserify-strongloop/express-facebook/react-gulpjs/gulp-pouchdb/pouchdb
• Full data
• nolanlawson/tiny-queue
  • local-npm
  • Regular npm
• lodash/lodash
  • local-npm
  • Regular npm
• substack/node-browserify
  • local-npm
  • Regular npm
• strongloop/express
  • local-npm
  • Regular npm
• facebook/react
  • local-npm
  • Regular npm
• gulpjs/gulp
  • local-npm
  • Regular npm
• pouchdb/pouchdb
  • local-npm
  • Regular npm
• Notes
• Reproduce these results

I started off with a small repo of mine,nolanlawson/tiny-queue, then moved on to some popular libraries likelodash/lodash andfacebook/react. In each case, I'm just cloning the code from Github and runningnpm install.

TLDR: Regular npm is faster for the firstnpm install, but afterwardslocal-npm is always faster, even afternpm cache clear. Sometimes it's even 2x or 3x faster (e.g. 1m15.752s vs 3m50.467s to installfacebook/react the 2nd time). This is kind of the point oflocal-npm: it gets faster the more you run it, because it aggressively caches everything.

• 45.69% slower for firstnpm install
• 3.22% faster for secondnpm install
• 16.99% faster afternpm cache clear

• 34.26% slower for firstnpm install
• 6.78% faster for secondnpm install
• 2.75% slower afternpm cache clear

• 33.00% slower for firstnpm install
• 5.64% faster for secondnpm install
• 20.27% slower afternpm cache clear

• 29.44% slower for firstnpm install
• 9.86% faster for secondnpm install
• 15.16% faster afternpm cache clear

• 26.71% slower for firstnpm install
• 8.89% slower for secondnpm install
• 2.76% faster afternpm cache clear

• 10.40% slower for firstnpm install
• 6.68% faster for secondnpm install
• 12.54% slower afternpm cache clear

• 8.16% slower for firstnpm install
• 6.13% faster for secondnpm install
• 12.62% faster afternpm cache clear

• 1st npm install: 0m2.392s
• 2nd npm install: 0m1.263s
• After cache clean: 0m2.447s

• 1st npm install: 0m1.299s
• 2nd npm install: 0m1.305s
• After cache clean: 0m2.948s

• 1st npm install: 1m22.140s
• 2nd npm install: 0m48.855s
• After cache clean: 1m19.177s

• 1st npm install: 0m53.995s
• 2nd npm install: 0m52.407s
• After cache clean: 1m17.000s

• 1st npm install: 0m41.307s
• 2nd npm install: 0m26.133s
• After cache clean: 0m49.842s

• 1st npm install: 0m27.677s
• 2nd npm install: 0m27.695s
• After cache clean: 0m39.738s

• 1st npm install: 0m19.571s
• 2nd npm install: 0m13.149s
• After cache clean: 0m16.645s

• 1st npm install: 0m13.810s
• 2nd npm install: 0m14.587s
• After cache clean: 0m19.619s

• 1st npm install: 2m30.568s
• 2nd npm install: 1m54.844s
• After cache clean: 2m39.366s

• 1st npm install: 1m50.349s
• 2nd npm install: 1m44.638s
• After cache clean: 2m43.893s

• 1st npm install: 0m40.063s
• 2nd npm install: 0m36.125s
• After cache clean: 1m4.591s

• 1st npm install: 0m35.895s
• 2nd npm install: 0m38.711s
• After cache clean: 0m56.494s

• 1st npm install: 2m37.697s
• 2nd npm install: 2m2.765s
• After cache clean: 2m40.102s

• 1st npm install: 2m24.823s
• 2nd npm install: 2m10.778s
• After cache clean: 3m3.219s

The test is a little unfair, because mylocal-npm has already downloaded all the metadata, but this is a typical use-case forlocal-npm. (It replicates the npm metadata on first run, which can take a few hours.) I did clear out the tarballs before the test, though.

Also, repos tested later in the test may benefit from a slight boost, if they share any modules with a previous repo. But again, this is a typical use-case forlocal-npm – the more you use it, the more that common tarballs will be pre-cached.

I find it interesting thatlocal-npm is faster even after the secondnpm install, which is when npm's cache is supposed to kick in. This suggests that npm is not caching as aggressively as it could be.

First,npm install -g local-npm && local-npm. Then just clone this repo and runnpm test. The python script can print a summary afterwards.

Warning: the test will override your local~/.npmrc file! So you may want to donpmrc -c testing && npmrc testing beforehand.