Performance optimization suggestion, lint only failed files + changed files

[sergesemashko]

On my current project introducing stylelint raised watch build time from ~3s to ~25s. I guess, it happened because of big amount of files and amount of rules.

So, maybe we could cache files that failed to lint on initial run and then include them + files that changed for linting. That should decrease overhead from linting all the files that passed initially and hasn't been changed.

It seems this will require slight update how to deal with stylelint formatter, because in order to get failed files it would require both JSON and string output.

If this suggestion sounds good, I'd be happy to poke around and apply the optimization.

[JaKXz]

@sergesemashko this was implemented with lintDirtyModulesOnly right?

[sergesemashko]

@JaKXz, correct.
BTW, cache option was added to stylelint. So, if someone has performance issues, that flag can dramatically decrease lint time.