chore(ci): disable node cache to speed up ci

[boredland]

☕️ Reasoning

in my experience restoring and storing cache costs so much time, that the (tiny if even) gain during install is gone. this tries to prove that for next-auth.

before:

install: 1m 16s
post-setup: 1m 44s

after:

install: 1m 11s
post-setup: 0s

... also yamllint complaint about redundant string indicators.

🧢 Checklist

• Documentation
• Tests
• Ready to be merged

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Comments	Updated
auth-docs	❌ Failed (Inspect)			Mar 13, 2023 at 9:56PM (UTC)

1 Ignored Deployment

Name	Status	Preview	Comments	Updated
next-auth-docs	⬜️ Ignored (Inspect)			Mar 13, 2023 at 9:56PM (UTC)

[balazsorban44]

Interesting observation. We should have a deeper look into our CI at some point. We often have cache misses for the docs build as well.

We can merge this for now and see. Thanks!

[AlexErrant]

In your "before", the workflow file says cache: "pnpm", which is great, but the run does not use the cache: "pnpm cache is not found". This is probably because the package.json or lockfile changed and no cache for that package.json/lock file exists. After this run however, it does exist, so it should be available for subsequent runs and may improve their times. You may still be correct in that restoring from a cache isn't as fast as downloading from scratch, but your before does not demonstrate this.

In my own project (not a fork of NextAuth), caching is faster. Here's a screenshot of my own project showing how the cache improves overall pnpm i times (n=1):

Left. Right.

[SauravMaheshkar]

In your "before", the workflow file says cache: "pnpm", which is great, but the run does not use the cache: "pnpm cache is not found". This is probably because the package.json or lockfile changed and no cache for that package.json/lock file exists. After this run however, it does exist, so it should be available for subsequent runs and may improve their times. You may still be correct in that restoring from a cache isn't as fast as downloading from scratch, but your before does not demonstrate this.

In my own project (not a fork of NextAuth), caching is faster. Here's a screenshot of my own project showing how the cache improves overall pnpm i times (n=1):

Left. Right.

@balazsorban44 thoughts on this 👀 ?

[AlexErrant]

Gonna guess he doesn't care that much about a CI speedup. If you care, you could fork, run it in your own repo, and come back with empirical runtimes.