typst column layout: interpret unitless image widths as pixels and convert to inches

[gordonwoodhull]

Follow-up to #11658.

Currently, the following example will fail because it generates Typst output with unitless lengths:

---
title: Unitless image widths in column layout
format: typst
---

::: {.callout-note}

## Plots

::: {layout-ncol=5}

![]({{< placeholder format=svg >}})

![]({{< placeholder format=svg >}})

![]({{< placeholder format=svg >}}){width=100}

![]({{< placeholder format=svg >}})

![]({{< placeholder format=svg >}}){width=25}

:::
:::

For the images, we already interpret unitless image widths as pixels and convert to inches:

quarto-cli/src/resources/filters/quarto-post/typst.lua

Lines 101 to 110 in 46a7609

	-- REMINDME 2024-09-01
	-- work around until https://github.com/jgm/pandoc/issues/9945 is fixed
	local height_as_number = tonumber(image.attributes["height"])
	local width_as_number = tonumber(image.attributes["width"])
	if image.attributes["height"] ~= nil and type(height_as_number) == "number" then
	image.attributes["height"] = tostring(image.attributes["height"] / PANDOC_WRITER_OPTIONS.dpi) .. "in"
	end
	if image.attributes["width"] ~= nil and type(width_as_number) == "number" then
	image.attributes["width"] = tostring(image.attributes["width"] / PANDOC_WRITER_OPTIONS.dpi) .. "in"
	end

But we also need to use the converted widths for the layout.

Alternatively / in the the long run, we could contribute the fix to Pandoc instead.

jgm/pandoc#9945

[cscheid]

Can we make sure we've checked for how unitless dimensions are interpreted in other formats/settings? I'm thinking specifically:

• format: latex
• fig-width: N in knitr code cells

[gordonwoodhull]

Sure. LaTeX works mostly the same - looks like it needs some margin/padding on its images, and it will not accept SVG placeholders.

In the example from above the 3rd placeholder image has unitless width=100 and the 5th has width=25.

One can recognize that the LaTeX output

has the same intent as Typst output (if not the same execution)

fig-width is new to me, will check.

[cscheid]

Great, thanks. I don't think we need to (or can) do a perfect job here because units are also not a format-agnostic concept, but it will help us to know a bit more of the current behavior.