a few people have been asking me why i forked the unicorn lint plugin, and some not-as-friendly people unfortunately have been telling me i haven't done anything other than "replace a couple of dependencies" so here's just some stuff i got upto on the fork 🧵 - ThreadSky

43081j.com • 12 days ago

a few people have been asking me why i forked the unicorn lint plugin, and some not-as-friendly people unfortunately have been telling me i haven't done anything other than "replace a couple of dependencies"

so here's just some stuff i got upto on the fork 🧵

Comments

matteogabriele.bsky.social•12 days ago

When people complain about someone forking a repository, it clearly shows they have too much time.

43081j.com•12 days ago

i switched from find-up to empathic. even though we don't traverse the FS often in the plugin, this is still a no brainer since empathic is much faster

jackf.io•11 days ago

Omg I didn't know about empathic and I've needed this for years! Most of these functions I've written myself but less efficiently about a million times in various scripts or side projects 😁

43081j.com•11 days ago

There's also a crazy amount of packages to find specific files (package.json, editor config, etc)
We've been able to replace all of them with this one package now. Luke has done great work 🙏

43081j.com•12 days ago

i then did a bunch of CPU profiling and deopt work on the individual rules. some saw a 5x speed up when run inside TSESLint's repo, and half as much memory usage

43081j.com•12 days ago

some patterns like this are awful for performance:
`isMethodCall(astNode, opts)`
opts changes shape all over so v8 can't optimise here. these monolithic helpers were split up into smaller type guards which could be optimised

43081j.com•12 days ago

some rules also match from the wrong end. if you're looking for literals inside something specific, often you're better off matching the "something specific" rather than testing every literal.
this approach cut cpu time massively for a few of the rules

43081j.com•12 days ago

the source repo also includes a custom build of lodash just so it can do some basic object merges, and string casing. this obviously doesn't need a whole build of lodash (seems the source repo will fix this at some point at least)

43081j.com•12 days ago

But I also think we will end up making breaking changes that can't go into upstream (yet). Like deprecating rules or removing them entirely

Comments

Posting Rules

Reply