feat(check-eslint): Check for ESLint peerDependency being met #290
Conversation
| if (rules.length < 1) { | ||
| return; | ||
| } | ||
| if (checkEslint.check() === false) { |
There was a problem hiding this comment.
couldn't we just require('eslint') and make it a peer dep?
There was a problem hiding this comment.
It currently is a peer dep and we just require it in lib/rule-finder, but in my experience peer deps are not installed automatically so I thought it would be good to have a nice warning.
Eventually I want to avoid confusion as in #285
There was a problem hiding this comment.
... and instead of just requiring eslint (that is, in a try-catch, to check for it), this draft of an implementation also pulls the corresponding information from the package.json and tests if the installed version satisfies it.
On a second thought, that might be a good idea to even make it a separate, general purpose module and just use that (and concentrate on core functionality of this tool).
What do you think?
There was a problem hiding this comment.
I think that letting the require throw naturally, and expecting users to run npm ls themselves, is fine - i don’t see much value in an explicit check.
There was a problem hiding this comment.
Ok. I have to admit, I didn‘t realize this check would be as bloated as it got already (and there still is erroneously assumed, the package.json to be tested again would be found at process.cwd, so it would be even more bloated, if done properly).
I‘ll happily close this then.
But to hijack this discussion a little bit: What about a proper usage information (using the functionality provided by yargs)?
There was a problem hiding this comment.
That part sounds awesome; I love yargs and its auto-help feature.
Before anything else, checks that ESLint is installed and present in a sufficient version, exit with code 1 otherwise.
This is currently wip — I just threw it together this evening and wanted to collect some opinions.
There are some minor issues reported by the linter: almost the whole content of
bin/diff.jsmoved into the eslint-check's else-branch and with it all the function declarations. I think that whole thing should be refactored soon anyways, so I'd think it's not that big of a deal ...I planned to make use of
yargsfeature to print out usage instructions too (in case of insufficient arguments), but that would be out of scope for this PR.PS: If not already known, appending
?w=0to the url for the commits of this PR (like here) ignores whitespace changes and makes this PR much more easy to read.