word-omml-mathtype

word-omml-mathtype

A Codex skill for converting existing Word OMML equations in a .docx into genuine MathType OLE objects on Windows.

It handles display equations, inline formulas, section-scoped conversion, residual $ cleanup, and verification. It was developed and tested against a document containing 71 OMML equations.

Install as a Codex skill

Clone this repository into a skills folder, for example:

git clone https://github.com/YYK-beep/word-omml-mathtype.git `
  "$env:USERPROFILE\.codex\skills\word-omml-mathtype"

The folder must contain SKILL.md at its root. Codex can then discover the skill by name or description.

Requirements

• Windows
• Microsoft Word desktop
• MathType 7 with MathType Commands 2016.dotm loaded in Word
• Python 3
• pypandoc-binary or pandoc

Install or verify the runtime dependencies:

powershell -ExecutionPolicy Bypass -File scripts/setup_environment.ps1

Quick start

Convert only one section of an application or thesis:

powershell -ExecutionPolicy Bypass -File scripts/convert_omml_to_mathtype.ps1 `
  -SourcePath "D:\docs\input.docx" `
  -TargetPath "D:\docs\input_mathtype.docx" `
  -SectionHeading "四、课题研究内容和工作方案" `
  -EndHeading "五、基础条件和优势"

Run a small probe first:

powershell -ExecutionPolicy Bypass -File scripts/convert_omml_to_mathtype.ps1 `
  -SourcePath "D:\docs\input.docx" `
  -TargetPath "D:\docs\probe.docx" `
  -SectionHeading "四、课题研究内容和工作方案" `
  -EndHeading "五、基础条件和优势" `
  -MaxFormulas 2

Verify:

powershell -ExecutionPolicy Bypass -File scripts/verify_mathtype.ps1 `
  -Path "D:\docs\input_mathtype.docx" `
  -SectionHeading "四、课题研究内容和工作方案" `
  -EndHeading "五、基础条件和优势"

Expected result:

• omathsInRange : 0
• mathTypeShapes : <number of converted formulas>
• remainingDollarCharacters : 0

Scripts

Performance

• 100 ms polling instead of 300 ms intervals.
• Standalone alphabetic formulas are pre-wrapped as \mathrm{...} instead of being attempted once and retried after a timeout.
• The document is saved every 20 formulas, not every 5.
• Delimiter $ cleanup runs once at the end instead of after every formula.

In a 5-formula probe, total runtime including Word startup was about 12.6 s; after the first formula, each formula typically completes in 0.4-0.5 s.

Design notes

• The source is copied first; the original document is never edited in place.
• Conversion processes OMaths from last to first so earlier ranges remain stable.
• Progress is saved every five formulas.
• Single alphabetic formulas such as Ber or KL are retried with \mathrm{} because MathType’s TeX toggle may otherwise refuse to create an object for a standalone word.
• Display equation numbers remain as normal text; formulas use the existing center and right tab stops.

Limitations

• Windows and Word desktop only.
• MathType 7 is required for real MathType objects.
• Pandoc must be able to read the document’s OMML equations.
• Complex nested or unusually malformed equations may still require manual review after conversion.
• The script cannot convert formulas in floating text boxes unless they are part of the document story processed by Word.

License

MIT