Solving insert-before and insert-after

[dstoc]

[Enter] has currently has magical and broken behavior.

The reason... it seemed desirable that pressing [Enter] in a list item would create a sibling list item. The workaround [Shift-Enter] creates an adjacent paragraph within the list item.

Following on from that, it seemed reasonable that this behavior would apply to other block types. Pressing [Enter] in a block quote would create a paragraph adjacent to the block quote.

However, this feels wrong when the block-quote contains other items like sections -- because pressing [Enter] in a section still creates an adjacent paragraph in the section.

[dstoc]

Proposal: [Enter] creates adjacent paragraphs,

Except when:

• in a list-item, it will create an adjacent list-item
• when the current paragraph is the last child and is empty, it will relocate the paragraph to be the next sibling of its parent

This may solve insert-after in general, even for transclusions.

As for insert-before. Maybe something similar can work for first-child? E.g. to insert-before a transclusion, move the cursor to the start then press twice (or more? depending on the containing blocks). Sections would need special treatment.

[dstoc]

in a list-item, it will create an adjacent list-item

Unless it's not just a simple list item, that is it contains more than a single paragraph.

[dstoc]

There are lots of special cases.

With sections for example, when there's a section inside a list-item, the adjacent paragraph generated after the section gets consumed by the section in normalization. We probably want to insert a subsequent list-item instead, but what if there's a subsequent section? I think we need to introduce a mechanic that can split the contents of a list-item across multiple list items.

There already exist other problems now that this may help to solve, for example:

* a
  * b

  c 

Attempting to unindent "b", results in:

* a

  c
* b

[dstoc]

Some further thoughts on this, undistilled.

1. The model for [Enter] on an empty paragraph is not necessarily insert-before or insert-after, but could be thought of as "move-outwards". If you press [Enter] enough you'll end up at the top level (or top level section).
2. It may then follow that pressing [Enter] for an otherwise empty list-item is the same as the existing unindent behavior (the above example being an exception).

Splitting seems pretty complex, and I'm not sure there's an inverse. But maybe that's OK?

What can be split?

• A flow of paragraph+, section+ inside a list-item can be split into two adjacent list items
• A flow of paragraph+, section+ inside a blockquote can be split across the blockquote and its parent