Backend pipeline

OOXMLZipBackend

Structural extraction via zipped OOXML parts.

• Sheets, defined names, charts (when defined names are referenced)
• Connections + inferred sources
• No Excel installation required

OpenpyxlBackend

Formula text extraction (not evaluation).

• Formula cell nodes (include_formulas=True)
• Runs with include_formulas=True or include_cells=True
• Can use XML formula-only scans for large sheets (formula_scan_mode)
• Supports sheet scoping and pivot-range skips for formulas
• include_cells does not emit cell blocks in this backend

CalamineBackend

Optional used-range/value block scanning.

• Requires excelminer[calamine]
• Emits cell_block nodes

ComBackend (Windows)

Opt-in Excel automation for enrichment.

• Requires Excel + excelminer[com]
• Opt-in for modern OOXML formats
• Can extract sheets, connections, formulas

PowerQueryZipBackend

Parses xl/queries/*.xml and infers sources from M.

• Emits powerquery and m_script nodes
• Mashup containers attempt best-effort script extraction
• Links to source nodes when possible

PivotZipBackend

Best-effort pivot table + cache extraction.

• Emits pivot_table + pivot_cache
• Links pivots to caches and connections

VbaZipBackend

VBA module extraction for macro-enabled files.

• Emits vba_project nodes
• Captures VBA module text when available

Backend ordering

Default ordering is deterministic for reproducible diffs.

1. OOXML → VBA → Power Query → Pivot
2. Calamine → Openpyxl → COM