Attempt to fix verticle alignment of some boxes in latex backend

[wilbowma]

This addresses #260 by providing a default box-adjuster. This requires an extra LaTeX package.

I'm not really sure this is a good solution, but I thought I'd offer a patch for experts to review. It might have unintended consequences if existing code was relying on ignoring the vertical alignment style property. The tests seem to pass, but it seems like a hard thing to write a test for.

[mflatt]

Adding a new package is always worrying, but I'd be happy to have a solution here.

Would using minimal instead of export work and potentially avoid unwanted interactions with uses of includegraphics, or is export reasonably unlikely to break existing documents?

@rfindler Do you think we should give this a try and check the document rendering and a few papers?

[rfindler]

I tried the documentation PDF build and it caused no problems there. I also built a recent paper and it didn't cause any problems there either. I'm not savvy enough to understand how this fixes the problem but it sounds like you both are so I'm in favor!

[wilbowma]

minimal is probably better; I was explicitly trying to avoid using it directly with includegraphics directly.

@rfindler, what was happening was the call to do-render-paragraph did specify the box should be top aligned by passing 't as as-box-mode. However, for paragraphs, this gets implemented by reading the style properties and looking for a box-mode implementation. The pict didn't have one, so it was wrapped in hbox, the default, which coverts to a box but doesn't adjust vertical alignment. This patch uses adjustbox as the default, which both converts to a box and adjusts vertical alignment.

[wilbowma]

When I try minimal, I get complaints that valign is undefined by adjustbox, which is strange.

I also notice a small difference, although I haven't observed it yet: \adjustbox{valign=c} should probably behave like \hbox, since that's what it's replacing. But I think it behaves like \mbox. Unfortunately, I don't really understand how these two things differ.

[wilbowma]

The vertical alignment in tables does not behave as expected. It seems to be only coincidence that my example worked. I'll continue debugging and post in #260, but this patch doesn't really work.

[mflatt]

Thanks for investigating more. My vague memory of table-cell alignment was that the CSS and Latex models were deeply different, and I found some approximation that worked for some common cases and gave up on the rest.

[wilbowma]

Yeah thats what im discovering totally understand why you gave up.

…

-- Sent from my phoneamajig

On Aug 20, 2020, at 17:26, Matthew Flatt ***@***.***> wrote:  Thanks for investigating more. My vague memory of table-cell alignment was that the CSS and Latex models were deeply different, and I found some approximation that worked for some common cases and gave up on the rest. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.