Forbid comments with more than two dashes

[steventango]

Description of the Change

Any comment ending with more than 2 dashes is now considered as invalid.illegal.bad-comments-or-CDATA.xml

Alternate Designs

Benefits

Grammar no longer accepts ---> comment ending that is forbidden by XML grammar.

Possible Drawbacks

Applicable Issues

#84

[winstliu]

Thanks for the PR, @TangSteven! Just one question :).

Additionally, would you mind adding a new spec for this to ensure that this behavior doesn't regress? Thanks again 🙇

[winstliu]

Is there a reason this is {3,} and not {3}?

[steventango]

Well, if 3 dashes is invalid grammar, anything longer than 2 should also be invalid.

So if the end arrow was four long (---->)
{3} would only highlight - --->
{3,} highlights the entire arrow which makes more sense, ---->

[winstliu]

Hmm, I feel like we should only be highlighting the last three.
@Ingramz any opinions on this?

[steventango]

I'm not quite sure how to add a new spec, would you mind pointing me in the right direction?

[winstliu]

I'm not quite sure how to add a new spec, would you mind pointing me in the right direction?

Yeah, of course! Specs are added to https://github.com/atom/language-xml/blob/master/spec/xml-spec.coffee. It looks like there aren't a lot of specs yet for this language, but you can take a look at language-html's for inspiration.

[Ingramz]

The specification mentions that the grammar for comments is following:

Comment ::= '<!--' ((Char - '-') | ('-' (Char - '-')))* '-->'

Simplified:

Comment ::= '<!--' (CharNoDash | ('-' CharNoDash))* '-->'

Which can be interpreted as -- is illegal if it is not followed by >.

This pull request in its current state only covers the example provided in specification, but not the general case, which leaves cases such as the following valid:

<!-- Hello -- World -->

Do note that the following is still valid:

<!--- Hello World -->

I think it would be easier to just make a rule within the body of comment that just searched for --(?!>) and highlighted that part as the source of error. This way it doesn't really matter if it is 3, 4 or 10 hyphens and whether they are part of the comment body or the ending tag. Should be cleaner future maintenance wise too.

Edit: Also the comments for JSP (<%-- --%>) and XML (<!-- -->) should be broken apart as I think the double hyphen is only disallowed in XML comments and not in JSP.

[steventango]

I added a negative lookbehind (?<!<[!%]), as without it incorrectly matches the <!-- of the next comment. Example:

With the negative lookbehind:

How would I break apart the JSP and XML comments?

[Ingramz]

Just create two separate rules like this:

  'comments':
    'patterns': [
      {
        'begin': '<%--'
        'captures':
          '0':
            'name': 'punctuation.definition.comment.xml'
        'end': '--%>'
        'name': 'comment.block.xml'
      }
      {
        'begin': '<!--'
        'captures':
          '0':
            'name': 'punctuation.definition.comment.xml'
        'end': '-->'
        'name': 'comment.block.xml'
      }
    ]

Then work only on the second one.

The lookbehind should be unnecessary. From the point where the error is first registered, the rest of the document will be invalid. To make it a little clearer where the error comes from, we can highlight the first occurrence only by making use of begin/end rule:

        'patterns': [
          {
            'begin': '--(?!>)'
            'beginCaptures':
              '0':
                'name': 'invalid.illegal.bad-comments-or-CDATA.xml'
          }
        ]

[steventango]

Updated!