Short sentence standalone paragraphs missing from reader view of https://www.royalroad.com/fiction/63759/super-supportive/chapter/1449598/one-hundred-two-what-kind-of-wordchain

[jaredhirsch]

As described in https://bugzilla.mozilla.org/show_bug.cgi?id=1890310, this page contains a number of very short paragraphs (one or a few words) that are lost in reader view, even though most of the content is correctly displayed.

It's worth noting that this website's design is to wrap each paragraph's content in <div><div><span>, with inline styles applied to all three elements, rather than just putting the content in sibling <p> elements. So the page structure looks like a few hundred blocks like this:

<div style="margin: 1em 0; caret-color: #000000">
  <div style="direction: ltr">
    <span style="font-family: arial, sans-serif">Some content here</span>
  </div>
</div>

Some examples of missed content:

• Alden stared at it.
• “Let me think about it.”
• “Gotcha.”

Slightly longer sentences not dropped from reader view:

• “I was thinking the same thing.”
• “That’s very efficient,” said Alden.

[cmkm]

One potential fix might be to add chapter to the positive regex.

[inhumantsar]

adding chapter to the positive regex didn't help. looks like these short paragraphs are being filtered by _cleanConditionally:

 Reader: (Readability) Removing node: !isList && headingDensity < 0.9 && contentLength < 25 && (img === 0 || img > 2) && !this._hasAncestorTag(node, 'figure') (headingDensity=0, contentLength=9, img=0)

adding && linkDensity > 0 to that check lets them through though. seeing 221 child elements in the expected output vs 222 in the source, with the 1 missing element being an empty div. this does affect a number of test cases, but they seem manageable so far. i'll open PR (at least a draft) for this shortly.