Fix false negatives for template literal declarations in declaration-empty-line-before

[hudochenkov]

Clearly describe the bug

Warning is not shown for declaration-empty-line-before in styled component.

Which rule, if any, is the bug related to?

declaration-empty-line-before

What CSS is needed to reproduce the bug?

const Separator = styled.div`
  margin-bottom: 20px;
  
  border: 0;
`;

What stylelint configuration is needed to reproduce the bug?

e.g.

{
  "rules": {
    "declaration-empty-line-before": ["never"]
  }
}

Which version of stylelint are you using?

9.10.1

How are you running stylelint: CLI, PostCSS plugin, Node.js API?

CLI with stylelint "**/*.styled.js". Also in demo

Does the bug relate to non-standard syntax (e.g. SCSS, Less etc.)?

Yes, it's related to CSS-in-JS (styled components).

What did you expect to happen?

Warnings to be flagged for border.

What actually happened (e.g. what warnings or errors did you get)?

No warnings.

[hudochenkov]

The issue in

stylelint/lib/rules/declaration-empty-line-before/index.js

Lines 57 to 59 in ad46698

	if (!isStandardSyntaxDeclaration(decl)) {
	return;
	}

stylelint/lib/utils/isStandardSyntaxDeclaration.js

Lines 13 to 15 in ad46698

	if (parent.type === "root") {
	return false;
	}

In postcss-jsx parser properties are children of root.

[jeddy3]

In postcss-jsx parser properties are children of root.

I believe that's there because we don't want this rule to pick up on "declarations" outside of rules e.g. dollar variables. We can either work out another heuristic for this or change postcss-jsx behaviour so these declarations are children of a rule node with selector "div".

[alexander-akait]

I think fix postcss-jsx behavior is right way

[hudochenkov]

@evilebottnawi what this change could be? If postcss-jsx will wrap styles in div on parse, then it will introduce false positives for selector-max-type: 0, for example.

[alexander-akait]

@hudochenkov hm, maybe we can add context.isStyled (same we can do for markdown/html/etc if needed) and handle this cases?

[hudochenkov]

I was reviewing PRs and found #3933 which should fix this problem. This issue renders to be a duplicate of #3930.