DOCM to MD Conversion - Headings to not match expected numbering from document.

[Dark1886]

When running my DOCM through anydoc for a markdown output, I commonly get heading/subheading numberings that do not match the input DOCX file. Is there any way to change this or diagnose WHY is is doing this?

Additional Information

• ToC is correct
• Multiple section '1' at the start of the document.
• In 'Section 4' it went in this order

  Section 4
  Section 4.1
  Section 1.1
  Section 4.2

EDIT: Applying a new style and reapplying the intended style fixes the issues. If needed I could try to add a pre/post processing script to check for this, but I need to know what anydoc is interpreting as incorrect.

*I can not attach the document directly as it contains proprietary company information.

[Dark1886]

I managed to get a post-processing fix working for my use case, but this likely shows an issue in how new headings are created when converting DOC* files. See the issue summary and resolution below.

Issue Summary: Heading Numbering Corruption from Ghost <w:numPr> in DOCX/DOCM Conversion

1. Why It Broke (Root Cause Analysis)

• OpenXML Behavior in Word:
  When authors apply list formatting to a heading and later revert it back to a Heading style (e.g., Heading 1, Heading 2), Microsoft Word often leaves behind an orphaned or "ghost" numbering property tag (<w:numPr>) inside the paragraph properties (<w:pPr>), even while retaining <w:pStyle w:val="HeadingX"/>.

  <w:p>
      <w:pPr>
          <w:pStyle w:val="Heading2"/>
          <w:numPr> <!-- Ghost list numbering leftover -->
              <w:ilvl w:val="1"/>
              <w:numId w:val="11"/>
          </w:numPr>
      </w:pPr>
      <w:r><w:t>Generic Section Heading</w:t></w:r>
  </w:p>

• AnyDoc Parsing Precedence & Numbering Resolution:
  When AnyDoc extracts paragraph nodes from word/document.xml:
  1. Cross-contamination of counters: If a paragraph has both a heading style and a <w:numPr>, AnyDoc treats the paragraph's section numbering as part of an active list counter (numId) rather than independent document hierarchy.
  2. Out-of-order numbering: AnyDoc renders the element as a Markdown heading (## ...), but attaches the wrong prefix borrowed from an unrelated list counter (e.g., rendering 4 -> 4.1 -> 1.1 -> 4.2).
  3. Table of Contents / Anchor Breakage: Because AnyDoc generates anchors and ToC links based on the rendered heading text, these corrupted prefixes break anchor targets and document navigation.
• Secondary Edge Case (Style-less headings):
  Some documents have paragraphs that lack an explicit heading pStyle altogether, relying solely on list numbering definitions for their visual styling and numbering.

---

2. Possible Upstream Fixes for AnyDoc

1. Prioritize Heading Style (pStyle) Over <w:numPr> (Recommended)

   • Approach: When parsing a <w:p>, inspect <w:pPr>/<w:pStyle>. If the resolved style maps to a heading style (e.g., Heading 1–Heading 6 or custom heading hierarchy), ignore or discard conflicting <w:numPr> list counters unless the style definition explicitly defines multi-level outline heading numbering.
   • Why it helps: Prevents regular list counters (numId) from leaking into and overwriting heading section numbers.

2. Decouple List Numbering Counters from Heading Numbering Trackers

   • Approach: Maintain separate counter states for body list items vs. heading outline numbering.
   • Why it helps: If a heading element does reference a <w:numPr>, ensure it does not borrow from or advance the state of body bullet/numbered lists.

3. Fallback Numbering from Paragraph Run Text (w:t)

   • Approach: If the paragraph text already contains hardcoded hierarchical numbering (e.g., leading 4.1.2), give precedence to the actual text content rather than synthesizing or prepending a calculated list index from <w:numPr>.

4. Sanitization / Normalization Step in DOCX Ingestion

   • Approach: During the initial OpenXML parsing pass, strip <w:numPr> from any paragraph where pStyle is an intrinsic heading and the list style is not configured as a legal/outline heading list style.

[tomsideguide]

Could you attach a docx that reproduces this with the ghost numPr in it? I'll take a look.

[Dark1886]

Here is a generic document that shows the same error.

ghost_numpr_repro_sample.docx

In Microsoft Word: Proper Heading counts

Anydocs: Heading values restart halfway through.

Could you attach a docx that reproduces this with the ghost numPr in it? I'll take a look.