Improve social media preview

[svenseeberg]

Short description

Improve the social media previews.

More info: https://chat.tuerantuer.org/digitalfabrik/pl/rnoez75xxjyhugxo97mbjuqmte & https://tasks.tuerantuer.org/projects/infra/work_packages/3734/activity?query_id=59

Previews on Signal and WhatsApp are not yet working.

Proposed changes

• Improve edge case handling in excerpts.

Side effects

• N/A

---

Pull Request Review Guidelines

[JoeyStk]

Why do we need this line? :)

[svenseeberg]

For example

Hallo<br>
Welt

will evaluate to Hallo Welt as we replaced newlines and br tags with empty spaces in line 51.

[PeterNerlich]

Wouldn't re.sub(r"\s+", " ", stripped_content.strip()) work better in this case? Who says we only have two spaces?

[PeterNerlich]

Just yesterday I came across a post about implementing preview cards for forgejo. Maybe it helps to cross-compare this: https://codeberg.org/forgejo/forgejo/pulls/6053/files#diff-f9ae2dae24595a0ce188650ca882e187de8cc998

Edit: No, doesn't look like they implement anything but the og:image yet

[PeterNerlich]

Suggested change

	return stripped_content[:100].strip().rsplit(' ', 1)[0]+" ..."
	return stripped_content[:100].strip().rsplit(' ', 1)[0]+" …"

[PeterNerlich]

Looking at https://ogp.me/, we might benefit from adding/improving

• og:type (seems to already be defined elsewhere…? investigate where)
• og:url should be a fixed (= unchanging) url of the content and functions as an ID of the node on the graph.
  Investigate whether url is a permalink or contains the slug that might change for a page.
• og:locale Language of the content (e.g. en_US)
• og:locale:alternate Alternate locales the content is available in (repeat this meta tag for each alternative)
• og:site_name Name of the overall website the content is part of. I guess this would just be Integreat and should probably not be part of og:title if defined here…?
• og:image:secure_url Described as "An alternate url to use if the webpage requires HTTPS" – Might be out of date, but won't hurt to add it. Will be the same value as og:image
• og:image:alt Alt text for the image (not a caption). I could imagine this to be the problem on WhatsApp and/or signal, as ogp.me says "If the page specifies an og:image it should specify og:image:alt."
• og:image:type Mime type for the image