excel-stress-tester-builder

name	excel-stress-tester-builder
description	Creates Excel workbooks that exercise every content type in the Workbook Graph extraction spec. Use when building stress-test workbooks, measuring parser completeness, or regression-testing the xlsx_parser. Covers cells, formulas, tables, charts, shapes, conditional formatting, data validation, and edge cases.

Excel Stress-Test Workbook Builder

Create one or more Excel workbooks that collectively exercise every content type required by the extraction spec, for parser completeness measurement and regression testing.

Inputs

Input	Type	Description
output_dir	path	Where to write workbooks and manifests
seed	int	For deterministic generation
scale	enum	tiny | small | medium | large — rows/cols, chart count
formats	list	["xlsx","xlsm","xlsb"] — include what tooling supports
features	dict	Optional overrides: pivots, power query, slicers, shapes, external links

Outputs

• Workbooks: KS_ExcelStress_v1.xlsx, KS_ExcelStress_v1_reportlike.xlsx, KS_ExcelStress_v1_edges.xlsx
• Manifest JSON per workbook: what features exist and where (sheet, range, object id)
• Golden expectations expected_extraction_min.json: must-extract entities for the "Practical minimum" set

Coverage Checklist (Must Include)

Workbook layer

• Properties: created/modified, custom properties
• Calculation mode + date system (1900 vs 1904 if possible)
• Workbook-level defined names (named ranges, named formulas like =LET, =LAMBDA)
• External link placeholder (formula referencing another workbook, may be broken)
• Connections/queries metadata (or placeholder sheet + OOXML manifest note if not embeddable)

Sheet layer

• Visible + hidden + very hidden sheets
• Tab colors
• Freeze panes + split panes
• Row/col widths, hidden rows/cols
• Merged cells
• Print area + header/footer
• Sheet-scoped defined names

Object layer

Grid objects

• Cells: numeric, text, bool, date, error, blank
• Number formats: currency, %, accounting, custom date, fraction
• Formulas: relative, absolute, 3D refs, structured refs, dynamic arrays (spill)
• Rich text runs in a cell
• Hyperlinks: external + internal
• Data validation: list from range, custom formula
• Conditional formatting: formula rule, color scale, icon set
• Tables (ListObjects): header, totals, calculated column
• PivotTable (if feasible): cache from table, row/col/value fields
• AutoFilter with criteria

Canvas objects

• Charts: line, column, pie — series referencing ranges and table columns
• Text boxes, shapes, arrows, callouts — overlaps, z-order, grouping
• Images with anchors
• Slicer/timeline (optional)

Manifest (derived artifacts)

• Sheet purpose label: input / calc / dashboard / report / raw / lookup
• KPI cells and label cells
• Reading-order anchors for textboxes and blocks

Construction Strategy

1. Deterministic: Given seed, always place the same objects in the same cells.
2. Three workbook archetypes:
   • Tabular model: tables, pivots, charts
   • Report layout: blocks, merged cells, text boxes, KPI callouts
   • Edge-cases: spills, circular refs, weird number formats, rich text, hidden sheets

Failure Modes

If PivotTables, PowerQuery, or slicers cannot be generated with current libraries:

• Do not fake the file
• Add a NotGenerated section in the manifest
• Provide manual creation script/instructions to add in Excel and re-save

Acceptance Criteria

• Every checklist item present and recorded in manifest.json (sheet, range/object id, expected fields)
• Workbook opens without corruption in Excel/LibreOffice
• expected_extraction_min.json usable as test oracle for the extraction pipeline

Reference

See DESIGN.md for the parser architecture and docs/PARSER_KNOWN_ISSUES.md for known limitations.