#8986 Image before Title still respects Title metadata

[StephanMeijer]

The way I implemented this:

1. First comes regular stuff, like images or anything else
2. Then a block of meta comes including empty paragraphs
3. Then regular stuff like images and paragraphs again

Then it's processed in this order:

1. Meta without any empty paragraphs
2. Non-meta
   a. First block of regular stuff
   b. All empty paragraphs from within Meta
   c. Second block of regular stuff

---

Our internal ID: NLDOC-837

[StephanMeijer]

[StephanMeijer]

Not sure what to do with this. It is entirely logical this is being converted to a Title (meta) so I would prefer letting that meaning just sit in there and maybe change the example.

[jgm]

Can you explain more fully? Are you saying that the Title style is used for both the document title and the title of the table of contents section?

[StephanMeijer]

@jgm Feel free to review. Also asked my collegeau Paul to review.

[pkrijgsman]

Looks great, this should be a fix for the issues that we have.

[StephanMeijer]

@jgm responded at #9170 (comment)

This approach (skipping images) feels a bit ad hoc to me.

I'm thinking it might be preferable to allow items with class Title, Author, or Date to be counted as metadata no matter whether they occur in the document. In the case of Title or Date, only the first would be counted as part of metadata if there are multiple. In the case of multiple Author, they could all be included in a list.

[StephanMeijer]

@jgm I changed the PR based on your feedback in my other PR.

[jgm]

Actually, my suggestion was to replace all the "splitting" code with the simple behavior I mentioned. Was trying to make it simpler, not more complicated.

[StephanMeijer]

@jgm I think I misunderstood what you mean then, as I implemented (I think) exactly what you described. Could you elaborate?

[jgm]

I think there's a function like this in Data.Maybe

[jgm]

I think you do implement what I suggested. But I guess I'd had in mind what might be a simpler approach. Instead of going through all the bp's and splitting them into metadata and non-metadata, I thought it would be simpler to simply keep a state variable for metadata and populate it as we go through the bp's, sequentially. Wouldn't that allow us to avoid all the splitting code at the beginning, while producing the same result? Let me know if that makes sense.

[StephanMeijer]

@jgm it absolutely makes sense!