feat: new private node property "_withoutClosingTag"

[SukkaW]

Notable Changes

The PR adds support for a private property _withoutClosingTag:

// fixture
{
  tag: 'div',
  _withoutClosingTag: true // By default the property won't exist
}

will result in

<div>

Commit Message Summary (CHANGELOG)

feat: new private node property "_withoutClosingTag"

Type

• Feature

SemVer

• Fix (:label: Patch)
• Feature (:label: Minor)
• Breaking Change (:label: Major)

Issues

The PR is primarily used to implement maltsev/htmlnano#29, to omit those optional closing tags (without producing invalid HTML). It might also help to handle the case I mentioned in posthtml/posthtml-parser#9 (comment).

So far _withoutClosingTag should be considered as a private property of node (with _ prefixed) and shouldn't be documented, but we could make it public in the future.

Checklist

• Lint and unit tests pass with my changes
• I have added tests that prove my fix is effective/works
• I have added necessary documentation (if appropriate)
• Any dependent changes are merged and published in downstream modules

[Scrum]

The implementation should be done here because this block is responsible for the type of render

posthtml-render/lib/index.js

Lines 192 to 206 in 903037c

	if (isSingleTag(tag)) {
	switch (closingSingleTag) {
	case 'tag':
	result += '></' + tag + '>';
	break;
	case 'slash':
	result += ' />';
	break;
	default:
	result += '>';
	}
	result += html(node.content);

I think that it is necessary to introduce another type, for example, byProperty.

1. You are adding an element that must be single options.singleTags = ['html']
2. Refactor isSingleTag

const asTag = '></' + tag + '>';
const asSlash = ' />';
const asDefault = '>';
switch (closingSingleTag) {
  case 'tag':
    result += asTag;

    break;
  case 'slash':
    result += asSlash;

    break;
  case 'byProperty':
    result += node._closingElement ? asTag : asDefault;
    break;
  default:
    result += asDefault;
}

Don't forget to add this to the documentation.

thanks

[Scrum]

_withoutClosingTag The name is not very suitable because tag means that there is an opening and closing element.

It seems more appropriate to me that the _closingElement

[cossssmin]

Regarding semantics, element in HTML refers to the whole thing - an element has start/opening and end/closing tags.

I agree double negation through the use of without in the name is unnecessary, but if a name such as _closingTag creates confusion, maybe we could use something like _needsClosing or _shouldClose or something similar?

[Scrum]

Regarding semantics, element in HTML refers to the whole thing - an element has start/opening and end/closing tags.

It is considered that if the element is single such as <hr> or <img> then it is the element if the element has a closing element then the tag is used

[cossssmin]

@Scrum I understand this could be a solution to posthtml/posthtml-mso#1 ?

[Scrum]

@Scrum I understand this could be a solution to posthtml/posthtml-mso#1 ?

Yes, you can bypass all child elements by the found match and set them a parameter

[SukkaW]

I think that it is necessary to introduce another type, for example, byProperty.

@Scrum

withoutClosingTag should be considered as when it is not a single tag (like hr, img) while the closing tag could be omitted safely (without introducing invalid HTML). See WHATWG spec: https://html.spec.whatwg.org/multipage/syntax.html#optional-tags

It seems more appropriate to me that the _closingElement

Actually, WHATWG HTML spec uses the word start tag and end tag. I use closing because it feels more consistent with the posthtml options' name.

I agree double negation through the use of without in the name is unnecessary

@cossssmin

By default posthtml-parser will not add _withoutClosingTag to node, thus Boolean(node._withoutClosingTag) will be false, unless a posthtml plugin specify it to be true when travase through tree.

[Scrum]

withoutClosingTag should be considered as when it is not a single tag (like hr, img) while the closing tag could be omitted safely (without introducing invalid HTML). See WHATWG spec: https://html.spec.whatwg.org/multipage/syntax.html#optional-tags

The fact is that we go beyond recommendations by manipulating the html tag as a single tag in conjunction with conditional comments. Therefore we have no right to look in the recommendation because we break them using this construction.

After a little reflection, I realized that we can do even better if we pass the type used for closingSingleTag

result += typeof node._closingElement === string ? as[node._closingElement] : node._closingElement ? asTag : asDefault;