fix no-unit-unknown rule for variable usage

[laysent]

Which issue, if any, is this issue related to?

Closes #4162

Is there anything in the PR that needs further explanation?

I noticed that lint-staged changed the way how it's configured. From their document, they encourage to use ignore configuration of each linter. And since this file is already in .eslintignore and .prettierignore, I believe it's safe to remote corresponding line in lint-staged config.

[hudochenkov]

I wonder why it wasn't catched by !isStandardSyntaxValue(value) function? I think we need to fix this function instead of proposed fix.

[alexander-akait]

hm, https://github.com/stylelint/stylelint/blob/master/lib/utils/isStandardSyntaxValue.js#L23 should be

[laysent]

I wonder why it wasn't catched by !isStandardSyntaxValue(value) function? I think we need to fix this function instead of proposed fix.

blurInterpolation removes @ and thus it won't be recognized correctly by isStandardSyntaxValue here.

I have changed the way how this issue is fixed, i.e. use node.value instead of transformed value in isStandardSyntaxValue. Please kindly have a look.

[hudochenkov]

Nice solution!

Could remove lint-stages changes, please? We squash commits on merge and this change would be confusing in unrelated commit.

[laysent]

Nice solution!

Could remove lint-stages changes, please? We squash commits on merge and this change would be confusing in unrelated commit.

Sure, I have removed lint-stages changes.

Would you like me to send another PR for this lint-staged changes? Seems at least on my side, lint-staged won't work without changing the config in package.json.

[hudochenkov]

Would you like me to send another PR for this lint-staged changes?

Yes, please :)

[hudochenkov]

Thank you!

[hudochenkov]

• Fixed: no-unit-unknown false positives for at-variables (Less) starting with numbers (#4163).

[alexander-akait]

/cc @hudochenkov it is break other case, for example .selector { property: 20px\9; } - regression

[hudochenkov]

But it's a hack for IE8.

All tests were passed.

[alexander-akait]

@hudochenkov yes, based on code test should be failed 😕 https://github.com/laysent/stylelint/blob/0c68261c43123df7e0ffaabbdc350a8a480fb43a/lib/utils/getUnitFromValueNode.js#L29, maybe postcss-value-parser recognize this as other type for node, i will look on this today

[laysent]

@hudochenkov Hi, just curious when would be the next release of stylelint? With this fixed, I could introduce the tool to our team and use it in our projects. ;-)