Some HTML <div> content wrapped in <p> unnecessarily

[nkalvi]

F. ex.
With Pandoc 1.14.0.4

echo '<div>
 <span>test</span>
</div>' | pandoc'

results in:

<div>
<p><span>test</span></p>
</div>

where as Pandoc 1.13.0.1 produces expected output:

<div>
<span>test</span>
</div>

[jgm]

Not a bug, I think, just the result of the addition of "native divs." With `native_divs`, pandoc parses whatever is inside the div tags just as it would have parsed it by itself, and sticks the result in the Div container.

[nkalvi]

Thanks again @jgm for clarifying.

Yes, I can get the expected results with -f markdown-native_divs.
But I feel that having this (and native_spans) turned off by default may be better.
Some former users may see unexpected changes in some output when upgrading to 1.14.x or higher.

[jgm]

I just checked: native_divs has been in pandoc since 1.13, and changing the default now would not be a good idea.

This raises the question why you're seeing a difference between 1.13.0.1 and 1.14. Clearly something changed in how native_divs behaves.

I believe this is due to 598d3ee, see #1591.
This change was in 1.13.2.

[nkalvi]

I didn't see that it was in 1.13 - yes, it's better not to change the default.

Yes, it does look like it is related to the change prompted by #1591.
While it makes sense to wrap 'plain' content inside div without requiring extra blank line,
wrapping inline elements in p seems unnecessary.

I'm sorry, I cannot see (don't know Haskell yet) from the diff in the reader why it affects only inline HTML elements. F.ex.

With the following input h1 doesn't get wrapped in <p> in the output:

<div>
  <h1>testing</h1>
</div>

while iframe in the following does:

<div>
  <iframe src="http://www.w3schools.com"></iframe>
</div>

BTW, it looks like it happens when the inline element is indented (i.e. does not start at the left margin).
Indentation doesn't seem to affect block elements.

[jgm]

The h1 doesn't get wrapped because this by itself would be an HTML block, not a paragraph:

<h1>testing</h1>

The iframe does get wrapped because this would not be parsed as an HTML block; the iframe tags get parsed as inline HTML:

  <iframe src="http://www.w3schools.com"></iframe>

You can check this using pandoc -t native:

[Para [RawInline (Format "html") "<iframe src=\"http://www.w3schools.com\">",RawInline (Format "html") "</iframe>"]]

Unfortunately iframe is one of those tags that can be used either in block or inline contexts. In CommonMark we're working out a better way of dealing with these, so authors will at least have the option of getting them treated as block tags.