Last item in lists is not wrapped in <p>

[ericmj]

You can see the formatting issue on https://hex.pm/docs/tasks#hex_publish.

[RobertDober]

Frankly I cannot. Can you provide markdown input and expected html output?

[ericmj]

You can see that some of the text in lists is bold and some is not. If you cannot see it on the website you can inspect the source.

Here is the markdown: https://github.com/hexpm/hex/blob/master/lib/mix/tasks/hex/publish.ex#L39-L52

[RobertDober]

I succeeded to extract an offending case

iex(30)> md = """
...(30)> # HL
...(30)>
...(30)>   * one
...(30)>
...(30)>   * two
...(30)> """
 "# HL\n\n  * one\n \n  * two\n"
iex(31)> Earmark.to_html md
"<h1>HL</h1>\n<ul>\n<li><p>one</p>\n</li>\n<li>two\n</li>\n</ul>\n"

[ericmj]

I was meaning to find a reproducing case but got bogged down on an unrelated issue. I am sorry for making you find it.

Thank you @RobertDober ❤️

[RobertDober]

np @ericmj I just needed to read the title...

[RobertDober]

I kinda have a fix for this, honestly the semantics of list items in paras goes a little over my head right now.
Neither from the code, nor from experiments on Babelmark I managed to really understand when to put them and when not.

Therefore, for starters I decided all <li> are wrapped in <p>s if one is, somehow defying the parser's semantics in the renderer. Ugly and maybe incorrect.

@pragdave your guidance would be appreciated.

And anybody I would be grateful for some pointers or explanations on that account.

[pragdave]

My understanding is that the

s are added if there is a blank line after
the item.

On Sun, Oct 2, 2016 at 5:43 AM, Robert Dober notifications@github.com
wrote:

I kinda have a fix for this, honestly the semantics of list items in paras
goes a little over my head right now.
Neither from the code, nor from experiments on Babelmark I managed to
really understand when to put them and when not.

Therefore, for starters I decided all

• are wrapped in

s if one is, somehow defying the parser's semantics in the renderer. Ugly
and maybe incorrect.

@pragdave https://github.com/pragdave your guidance would be
appreciated.

And anybody I would be grateful for some pointers or explanations on that
account.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#93 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAApmKWWucOJ1zV4v8Jqi7k0BmTa6AGZks5qv4q1gaJpZM4KKbKJ
.

[RobertDober]

Thank you @pragdave I will review the code a little more to be sure to not have violated that principle. After merging this issue branch it might be a good idea to release 1.0.2.

[ericmj]

My understanding is that the <p>s are added if there is a blank line after the item.

Don't you always have to terminate the list with a blank line after the last item?

[RobertDober]

@ericmj that is true, and BTW, if you are interested in technical details, it was the source of the bug, as the parser did not put the "spaced" flag on the last list item, never. And it is this never which still bugs me.

I guess one can trust the tests and the code is releasable in case the bugfix is urgently needed. Otherwise I'd like to investigate some further, but that will take some time.

[ericmj]

A bugfix is not needed urgently so on my part please take your time.

[RobertDober]

thx, well noted

[RobertDober]

I did some research here are three possible behaviors.

1. wrap a li's content into a p if there is a space before or after the item.
   (many implementations do that)
2. Wrap withp in case any line has a space before or after the item.
   (some, and amongst them commonmark, which is a surprise to me, do that)
3. For minimal backward incompatibility wrap with p if an item is followed by an empty line or the last line is preceded by an empty line. N.B. That implies a p for the last but one item too.

I like 2. best for it's simplicity, and because that is what we have right now on master. However it might be the most violent concerning backward incompatibility.

[RobertDober]

@pragdave I guess we have enough in the 1.0.2 milestone to release. Do you agree? Thank you in advance.