Support descriptions on additional interactive

[mitchkeller]

Sending this as three commits just to make clearer where the changes were. I think this now covers the various interactive elements (Desmos, GeoGebra, CalcPlot3D, CircuitJS, and arbitrary iframe) for descriptions.

• Two changes in assembly:
  • Ensure that interactive[@geogebra] with children has them copied when restructuring to use interactive/slate.
  • Move an unstructured description to be a shortdescription when the child of interactive
• Changes to pretext-html.xsl to replicate earlier work done for interactive[@platform] to have descriptions.
• A mixture of test cases (no shortdescription, both shortdescription and description, unstructured description) added to the sample article.

[rbeezer]

Thanks, @mitchkeller! I'll be able to look at this closer in a couple of days.

[rbeezer]

As discussed, review from Claude coming next.

• Some typos to fix up.
• Some duplicated code that should be isolated in a new template. let me know if you want me to do that, I can add it.

A force-push is fine. Thanks for this one!

[rbeezer]

Review of #2981 — works end-to-end; one consolidation suggestion

I built the sample article on the branch (its edits are the test cases) and inspected the rendered output for all five interactive types. Everything lands correctly:

Test case	Result
desmos-natural-log (short + long)	title="An interactive graph of a function.", long description block follows ✓
geogebra-right-triangle (shorthand @geogebra, structured desc)	title="described in detail…" + aria-describedby + description ✓
geogebra-train-distance (@platform, unstructured desc)	repaired to shortdescription → title="This is a test…" ✓
interactive-circuitjs-attribute (structured desc)	title + aria-describedby + description ✓
calcplot3d-intersection-planes (short + long)	title from short + description block ✓
calcplot3d-probability-wavefunction (long only)	title + aria-describedby + description ✓

The geogebra shorthand path is the subtle one and it checks out. interactive[@geogebra] is restructured in assembly to <interactive platform="geogebra"> (it explicitly adds @platform), and the new <xsl:apply-templates select="node()" mode="enrichment"/> copies the authored <description> across before the manufactured <slate>. By the time HTML sees it, it matches the existing interactive[@platform] description templates — which is why @geogebra correctly doesn't appear in the new HTML match list, and the copied description renders correctly in the output.

The repair extension is sound. Adding interactive/description[not(*[not(self::var)])] alongside image/description[…] correctly moves an unstructured interactive description to shortdescription (verified by geogebra-train-distance).

Consolidation opportunity (optional — the PR is correct as-is). This work brings the count of one accessibility block to five. The title/aria-describedby choose is now byte-identical across all five mode="iframe-interactive" templates — interactive[@desmos], [@calcplot3d], [@circuitjs], [@iframe], and the pre-existing [@platform]:

<xsl:choose>
    <xsl:when test="not(string(shortdescription) = '')">
        <xsl:attribute name="title"><xsl:apply-templates select="shortdescription"/></xsl:attribute>
    </xsl:when>
    <xsl:when test="description">
        <xsl:attribute name="title"><xsl:text>described in detail following the image</xsl:text></xsl:attribute>
        <xsl:attribute name="aria-describedby"><xsl:apply-templates select="." mode="describedby-id"/></xsl:attribute>
    </xsl:when>
</xsl:choose>

Since all five are identical and all are mode="iframe-interactive", lifting this into one shared template and calling <xsl:apply-templates select="." mode="iframe-accessibility-attributes"/> from each site would be output-neutral and collapse 5 copies to 1. Two smaller repetitions sit alongside it in the four new templates and could fold into the same helper: the <xsl:apply-templates select="." mode="iframe-dark-mode-attribute"/> after the choose is a redundant second call (it's already applied earlier in each template — harmless, last-wins, but dead), and the <xsl:apply-templates select="." mode="description"/> after </iframe> repeats five times. None of this affects output; it's purely DRY-ing up the five near-identical templates while they're all being touched anyway.

A few typos in the new sample-article descriptions:

• sample-article.xml:9633 (geogebra-right-triangle): "decomposed into four pices" → pieces; and "…shapes of the pieces in the left triangle. on the right…" → capital On (new sentence).
• sample-article.xml:9705 (calcplot3d-probability-wavefunction): the description labels the surface <m>\phi_{n_x,n_y}(x,y)</m>, but the interactive's jaxlabel is \psi and the caption is "Probability wavefunction" — should be \psi, not \phi.
• sample-article.xml:9666 (desmos-natural-log): "…at x=0, where the y-value is approximately 1.5." For ln(x^2+5)-3 that value is ln(5)-3 ≈ -1.39; the shape is described correctly, but the number looks off (likely ln(5)≈1.6 without the -3).

Non-blocking observation: for the short+long combination, the long description is in the DOM but the iframe is aria-describedby-linked only when there's no shortdescription. That's the existing @platform behavior faithfully replicated, so it's not this PR's concern — just flagging for whenever the broader pattern is revisited.

Net: functionally complete and correct across all five interactive types. No blocking change; the consolidation is a nice-to-have given how much of these templates is now duplicated, and the description typos are worth a quick fix since this is sample content.

Claude Opus 4.8, acting as a review assistant for Rob Beezer

[mitchkeller]

I've pushed the typo fixes, and I'd like to do the consolidation bit myself, as I think doing that will improve my understanding of how some of these things work. Questions before I dig into that:

• Is there a preferred location in pretext-html.xsl where I should put the iframe-accessibility-attributes template? Not sure if there's an overarching structure to where templates get located.
• I need to review more carefully, but it looks like there are two other attributes (@html-id-attribute and @interactive-sizing-style-attribute) that are addressed in all of these templates, but it's only the @iframe-dark-mode-attribute that accidentally got a redundant addition. Would it be better to leave all three of those attributes behind in each individual template or take them all along into the new template with the code Claude flagged? Maybe then accessibility attributes isn't the best descriptor though. (Perhaps iframe-common-attributes or iframe-shared-attributes?)
• Mention of the call <xsl:apply-templates select="." mode="description"/> after </iframe> that's repeated across the templates is a bit unclear to me. I can't take the </iframe> off into the new template, so is there any action to be done with this, or is that just Claude making an observation?

[rbeezer]

• Put it close by. I usually put it after use, so if a code-reader wants more detail they will get there eventually.
• Let's keep topical things together. You could put those three together in a template with a generic name (I like common).
• "a bit unclear to me" - maybe me too. I think that maybe lives outside the iframe, and needs to? So there is no real benefit in grouping it with other repeats? I saw in the diff and thought about it. Maye just leave it as-is.

[mitchkeller]

Two topical templates as suggested are now after the five places they're used. Left the mode="description" where it was, because I think we need to not be putting that inside the iframe. Should be ready for you/your assistant to review.