analysis of support MAthType conversion for QTI files.

 

Context

File: 2026 Sec 1 G3 Math post WA1 practice.docx Request: Can equations from this file appear in editable form in QTI ZIP and SLS?

---

What's in the DOCX

Format	Count	Storage	Complexity
OLE MathType (Equation.DSMT4)	19	Binary .bin in word/embeddings/	High — proprietary binary
OMML (<m:oMath>)	4	XML in document.xml	Trivial — all just minus signs

Each OLE equation has a WMF (Windows Metafile) preview image stored alongside it in word/media/imageN.wmf. These are small (416–1794 bytes), low-resolution previews. Browsers cannot display .wmf files natively.

---

What the Current Code Already Does

In docx_to_qti.html (lines 1171–1175 and 1263–1265), a previous AI added:

javascript

if (run.getElementsByTagNameNS(W,'object').length) {
  parts.push('$$\\text{[equation]}$$');
  continue;
}

OLE objects are detected and replaced with the literal text $$\text{[equation]}$$. The 4 OMML elements are silently ignored (no handling at all).

Result: All equations are lost from the output — only the placeholder text appears.

---

Approach Comparison

Approach A — Extract WMF Previews as Images

Idea: Each OLE equation already has a WMF preview linked via <v:imagedata r:id="..."/>. Extract the WMF → convert to PNG → include as an image token in the QTI ZIP.

Coverage	19/19 OLE equations (all complex ones)
Editable in SLS?	❌ No — displayed as image, students cannot interact
Effort	High — WMF is a 1990s Windows binary vector format; browsers cannot render it
Blocker	Need server-side conversion (LibreOffice, Inkscape, or ImageMagick) or a JS WMF renderer (none are reliable/maintained)
Feasibility	⚠️ Medium — viable only with a Python preprocessing step on the server

What would be needed:

• Python script using subprocess + LibreOffice headless: soffice --headless --convert-to png olePreview.wmf
• Or use wand (ImageMagick Python binding) to convert WMF → PNG
• Replace $$\text{[equation]}$$ with actual image tokens in iterBlockItems
• This requires the Python backend (app.py) to handle equation pre-processing

---

Approach B — OMML → MathML (browser-feasible, but trivial equations only)

Idea: Parse <m:oMath> elements and convert to MathML using Microsoft's published OMML-to-MathML XSLT stylesheet. MathML can be embedded directly in QTI 2.1.

Coverage	4/4 OMML equations — but all 4 are just minus signs (-)
Editable in SLS?	⚠️ Unknown — SLS QTI MathML support is unverified
Effort	Medium — XSLT is available, browser DOMParser can run it via XSLTProcessor
Blocker	Solves none of the 19 complex equations; SLS MathML support unconfirmed
Feasibility	✅ High technically — but near-zero practical impact for this document

---

Approach C — Parse OLE Binary (MTEF) → MathML/LaTeX

Idea: The .bin files contain MathType's proprietary MTEF (MathType Equation Format) binary. Parse MTEF directly → convert to LaTeX/MathML.

Coverage	19/19 OLE equations
Editable in SLS?	✅ Yes — LaTeX renders in SLS if supported
Effort	🔴 Extremely High — weeks to months
Blocker	MTEF is partially documented but complex; no maintained open-source JS/Python parser
Feasibility	❌ Not recommended

---

Approach D — Server-side Preprocessing Pipeline (Recommended if pursuing this)

Idea: Add a Python preprocessing step that converts the DOCX before the browser tool processes it:

1. Open DOCX, find each <w:object> with Equation.DSMT4
2. Extract the linked WMF preview via its relationship ID
3. Convert WMF → PNG using LibreOffice headless or ImageMagick
4. Replace the OLE object in the XML with a <w:drawing> containing the PNG
5. Save a modified DOCX → pass to existing browser converter

Coverage	19/19 OLE equations
Editable in SLS?	❌ No — images only
Effort	Medium (3–5 days) — Python script + server integration
Feasibility	✅ Viable — but adds server dependency to what is currently a pure browser tool

---

Summary Recommendation

	Editable?	Browser-only?	Effort	Practical?
Current (placeholder text)	❌	✅	Done	Minimal fidelity
A: WMF → PNG	❌	❌ (needs server)	High	Good fidelity, not editable
B: OMML → MathML	⚠️	✅	Medium	Only 4 trivial equations
C: MTEF → LaTeX	✅	❌	Extreme	Not feasible now
D: Server preprocessing	❌	❌	Medium	Best trade-off if server available

Bottom line:

• Editable math equations in QTI/SLS from legacy MathType OLE objects is not currently feasible in a browser-only tool
• If the Python backend (app.py) is extended, server-side WMF→PNG conversion (Approach D) gives the best practical outcome — equations appear as images, which SLS can display
• The cleanest long-term solution is for teachers to re-author equations in Word's native OMML math editor (Insert > Equation) instead of the legacy MathType plugin — OMML can be parsed and converted to MathML in-browser

Files Involved (if implementing Approach D)

• docx_to_qti.html — change OLE handler to support image tokens from pre-processed DOCX
• converter.py — add WMF→PNG preprocessing using LibreOffice/ImageMagick
• app.py — expose preprocessing endpoint or integrate into existing DOCX upload flow
• requirements.txt — add wand or Pillow + LibreOffice headless dependency