Publish to Facebook from WordPress (theme: "oriental") refinements

[tantek]

When you use Bridgy Publish to post from a WordPress blog (with theme: "oriental') to Facebook, posts the title, date/by, content, "This entry was posted in ...", and originally published, like this:

Person-tag experiment
Posted on December 3, 2015 by BrynLove
So this is me tagging Alex(t,f) and let’s try something else, by tagging Tantek(t,f)
This entry was posted in Uncategorized.
(Originally published at: http://loveiswhoweare.com/?p=552)

This could be improved in two ways, FIRST, only publish title, content, and originally published (drop date/by, "This entry was posted in") :

Person-tag experiment
So this is me tagging Alex(t,f) and let’s try something else, by tagging Tantek(t,f)
(Originally published at: http://loveiswhoweare.com/?p=552)

SECOND, bold the title to distinguish it from content body:

Person-tag experiment
So this is me tagging Alex(t,f) and let’s try something else, by tagging Tantek(t,f)
(Originally published at: http://loveiswhoweare.com/?p=552)

Thanks to @BrynWolf for the suggestions!

[snarfed]

thanks for filing! definitely worth improving.

[snarfed]

you all won't be surprised to hear this, but there are tons of wordpress(.com), blogger, and tumblr themes, and they often have radically different markup and microformats. i try to make the default themes work well enough, but i don't make many guarantees about the rest, like oriental.

i can definitely take a look though! here's a snippet of the parsed mf2 from http://pin13.net/mf2/?url=http%3A%2F%2Floveiswhoweare.com%2F%3Fp%3D552 . the thing that immediately jumps out at me is that This entry was posted in Uncategorized. is only in entry["value"], not in entry["properties"]["content"]. hmm.

{"items": [{
  "type": ["h-feed"],
  "children": [{
    "type": ["h-entry"],
    "properties": {
      "name": ["Person-tag experiment"],
      "content": [{
        "html": "&#xD;\n\t\t<p>So this is me tagging <span class=\"u-category h-card\"><a class=\"u-url p-name\" href=\"http:\/\/alexthejourno.com\/\">Alex<\/a>(<a class=\"u-url\" href=\"https:\/\/twitter.com\/alexthejourno\">t<\/a>,<a class=\"u-url\" href=\"https:\/\/www.facebook.com\/alex.campbell.usmc\">f<\/a>)<\/span> and let\u2019s try something else, by tagging <span class=\"u-category h-card\"><a class=\"u-url p-name\" href=\"http:\/\/tantek.com\/\">Tantek<\/a>(<a class=\"u-url\" href=\"https:\/\/twitter.com\/t\">t<\/a>,<a class=\"u-url\" href=\"https:\/\/www.facebook.com\/tantek.celik\">f<\/a>)<br><\/br><\/span><\/p>\n\t\t\t",
        "value": "So this is me tagging Alex(t,f) and let\u2019s try something else, by tagging Tantek(t,f)"
      }]
    },
    "value": "Person-tag experiment\r\n\t\t\t\t\r\n\t\t\tPosted on December 3, 2015 by BrynLove\t\t\r\n\t\r\n\t\tSo this is me tagging Alex(t,f) and let\u2019s try something else, by tagging Tantek(t,f)\n\t\t\t\r\n\r\n\t\r\n\t\tThis entry was posted in Uncategorized."
  }]
}],
...
}

[snarfed]

ooh, actually, mf2py parses this way differently from php-mf2 (above)! cc @kartikprabhu @kylewm. here are the two logs. the parsed mf2 doesn't actually have entry["properties"]["content"] at all, just entry["name"], which looks like it matches php-mf2's.

{"items": [{
  "type": ["h-entry"],
  "properties": {
    "name": ["New person-tag test\n\r\n\t\t\tPosted on December 4, 2015 by BrynLove \n\n\nWhen at first you don\u2019t like the way it looks, bugTantek \u00c7elik(t,f) and Ryan Barret(t).\n\n\r\n\t\tThis entry was posted in Uncategorized."]
  },
  ...
}

here's the HTML:

<article id="post-552" class="post-552 post type-post status-publish format-standard hentry category-uncategorized kind-">
  <header class="entry-header">
    <h1 class="entry-title">Person-tag experiment</h1>
        <div class="entry-meta">
      Posted on <a href="http://loveiswhoweare.com/?p=552" title="7:43 pm" rel="bookmark"><time class="entry-date" datetime="2015-12-03T19:43:59+00:00" pubdate>December 3, 2015</time></a><span class="byline"> by <span class="author vcard"><a class="url fn n" href="http://loveiswhoweare.com/?author=1" title="View all posts by BrynLove" rel="author">BrynLove</a></span></span>    </div><!-- .entry-meta -->
  </header><!-- .entry-header -->

  <div class="entry-content">
    <p>So this is me tagging <span class="u-category h-card"><a class="u-url p-name" href="http://alexthejourno.com/">Alex</a>(<a class="u-url" href="https://twitter.com/alexthejourno">t</a>,<a class="u-url"
href="https://www.facebook.com/alex.campbell.usmc">f</a>)</span> and let&#8217;s try something else, by tagging <span class="u-category h-card"><a class="u-url p-name" href="http://tantek.com/">Tantek</a>(<a class="u-url" href="https://twitter.com/t">t</a>,<a class="u-url"
href="https://www.facebook.com/tantek.celik">f</a>)<br />
</span></p>
      </div><!-- .entry-content -->

  <footer class="entry-meta">
    This entry was posted in <a href="http://loveiswhoweare.com/?cat=1" rel="category">Uncategorized</a>.  </footer><!-- .entry-meta -->
</article><!-- #post-552 -->

[snarfed]

bridgy is on mf2py 0.2.6, from 2015-05-06. the changes since then look pretty minor: https://github.com/tommorris/mf2py/blob/master/CHANGELOG.md

[snarfed]

i think we have enough now to punt to the mf2py people, either a new bug or reopening kartikprabhu/mf2py#45.

thanks again for the report, @tantek and @BrynWolf!

[kylewm]

@snarfed can you do me a favor and run

import requests, mf2py, pprint
r = requests.get("http://loveiswhoweare.com/?p=552", headers={"user-agent": "bridgy-testing"})
p = mf2py.parse(url="http://loveiswhoweare.com/?p=552", doc=r.text)
pprint.pprint(p)

in your virtualenv? I just did it here and it's definitely getting name, url, and content... 😕

[snarfed]

aha, you're right. i get the same (correct) thing in my virtualenv. thanks! narrowing down...

[BrynWolf]

Hey all,
thanks for including me in on this, as it's great for me to learn (even
though I now need to figure out what several words you guys typed mean...
like, "punt", which sounds like a football term...).
Hopefully I can try and implement some of your ideas and solutions and try
this myself.

Have a great Sunday. =)
Bryn

On Sat, Dec 5, 2015 at 12:02 PM, Ryan Barrett notifications@github.com
wrote:

aha, you're right. i get the same (correct) thing in my virtualenv.
thanks! narrowing down...

—
Reply to this email directly or view it on GitHub
#563 (comment).

Bryn "Wolf" Cartwright
Anthropologist, Yoga Teacher, Writer, Musician
Website: WWW.LoveIsWhoWeAre.com

music:
http://www.youtube.com/user/BrynShine

[armingrewe]

the only question being, American football or real football ;-)

[BrynWolf]

definitely American football. =)

On Sun, Dec 6, 2015 at 1:20 PM, Armin Grewe notifications@github.com
wrote:

the only question being, American football or real football ;-)

—
Reply to this email directly or view it on GitHub
#563 (comment).

Bryn "Wolf" Cartwright
Anthropologist, Yoga Teacher, Writer, Musician
Website: WWW.LoveIsWhoWeAre.com

music:
http://www.youtube.com/user/BrynShine

[kylewm]

It looks like the culprit is this bit https://github.com/snarfed/bridgy/blob/master/webmention.py#L83 where it has a special-case for Tumblr's markup to augment it with mf2 classes.

It happens that this Wordpress theme also has .content > .post too, so it replaces hentry with h-entry which defeats backcompat parsing on the rest of the entry.

Some ideas:

• Maybe there is a unique class in Tumblr pages that we could use to differentiate them?
• Instead of .post search for .post:not(.hentry,.h-entry)
• Only run the augment step if we first run mf2py and it doesn't find any entries on the page

[snarfed]

oh wow. nice sleuthing!

i don't feel strongly about how to fix this. the last idea seems like maybe the most robust? not sure.

[kylewm]

the last idea seems like maybe the most robust?

oh nice, that's where I was leaning too. sounds like a winner.