Add "no-multi-spaces" rule (fixes #133)

[armano2]

This PR implement rule proposed in #133

[armano2]

@mysticatea is there a way that we can have access to VAttributeText or something similar with whitespaces between attributes? or we can just add to VAttribute all spaces before.

for now i had to parse entire VStartTag to remove unnecessary spaces.

This is going to be needed for html-indent rule.

[mysticatea]

I think that this rule should be token-based.

return {
  Program(node) {
    // TODO: Check `context.parserServices.getTemplateBodyTokenStore` exists or not.
    const tokenStore = context.parserServices.getTemplateBodyTokenStore()
    const tokens = tokenStore.getTokens(node.templateBody, {includeComments: true})

    let prevToken = tokens.shift()
    for (const token of tokens) {
      const onSameLine = (prevToken.loc.end.line === token.loc.start.line)
      const spaces = token.loc.start.column - prevToken.loc.end.column
      if (onSameLine && spaces >= 2) {
        // Report.
      }
      prevToken = token
    }
  }
}

[mysticatea]

Spaces which don't change semantic are on outside of tokens, so we can check only extra spaces in this way.

[armano2]

@mysticatea thank you for suggestion, i'm still learning how to work with eslint rules 😄

[armano2]

@mysticatea i took your sugestion and i applied it to my code also i added basic doc for it.

i think we should change error message: Extra whitespace detected. seems bad 🍶

[armano2]

i consider <tag div=""/> and <tag div="" /> as valid but only when there is non or one white-space before /

we can add parameter to for this but i'm not sure if this should be in this rule

[mysticatea]

@armano2 I think better if this rule focuses on only the sequences of multiple spaces. In core rule case, no-multi-spaces is doing that, and for example, space-before-function-parentheses focuses on one or zero spaces which are followed by the parenthesis of function parameters.

[mysticatea]

This is inside of registerTemplateBodyVisitor, so context.parserServices.getTemplateBodyTokenStore should always exist.

[mysticatea]

This method cannot be used here because this method has special behavior that it ignores /* */ style comments which are not comments in HTML context.

[armano2]

ok

[mysticatea]

I'd like to see tests for the expression of Vue.js directives and mustaches. Those are parsed by script parser (espree by default), so those have valid tokens. This rule can report on multiple spaces which are in the expressions.

[mysticatea]

i think we should change error message: Extra whitespace detected. seems bad 🍶

For reference, this is the message of core no-multi-spaces rule.

[mysticatea]

This can be simpler a bit. It's "put a space between the previous token and the current token". Original spaces may be \t.

-                fix: (fixer) => fixer.removeRange([prevToken.range[1] + requiredSpaces, prevToken.range[1] + spaces])
+                fix: (fixer) => fixer.replaceTextRange([prevToken.range[1], token.range[0]], ' ')

[armano2]

in this case:

fix: (fixer) => fixer.replaceTextRange([prevToken.range[1], prevToken.range[1] + spaces], ' '),

is going to be valid cus there can be \n\r symbold before space

[mysticatea]

@armano2 Please use [email protected] since I found a bug about token's range (vuejs/vue-eslint-parser@d670ac9). That version fixed a bug that = tokens include following spaces.

[armano2]

@mysticatea thank you for suggestions, i applied most of them except that i changed a little fixer:

fix: (fixer) => fixer.replaceTextRange(
  [prevToken.range[1], prevToken.range[1] + spaces],
  requiredSpaces ? ' ' : ''
)

[mysticatea]

In core, no-multi-spaces doesn't warn trailing spaces. It's covered by no-trailing-spaces. I think it makes sense.

[armano2]

@mysticatea than its ok now?

[mysticatea]

I think better if this rule focuses on only the sequences of multiple spaces between tokens. Could you remove the following things?:

• The special handling of spaces which are before >. This should be covered by an independent rule with always/never options. Maybe space-before-tag-close or something like.
• Trailing spaces. This is covered by no-trailing-spaces.

[armano2]

@mysticatea ok changes applied

---

first time i didn't understand you

[armano2]

also idk if this is bug in parser but HTMLSelfClosingTagClose and HTMLTagClose has no value.

i like space-before-tag-close, i can work on it tomorrow.

[mysticatea]

LGTM, thank you for the great work, @armano2 !

also idk if this is bug in parser but HTMLSelfClosingTagClose and HTMLTagClose has no value.

It's intentional. Those kinds of tokens don't have variation, so their value property is empty.

[mysticatea]

Ah, also the value property of HTMLAssociation tokens is empty.
I think context.getSourceCode().getText(token) is good for this purpose.

[armano2]

thank you i learned few new useful thinks while building this rule

[armano2]

@mysticatea requested changes applied

[mysticatea]

Thank you!