spreadsheet-analyser

Formula-logic analysis for spreadsheets — the lens-family member that reads a workbook's reasoning (formulas, dependencies, error cells, named ranges, hygiene smells), not its data.

records-analyser reads .xlsx via pd.read_excel → computed values only, profiled as a data table. This reads the same bytes through openpyxl with data_only=False → formula logic: the SUM/IF/VLOOKUP/INDIRECT calls, the cell-dependency graph, the #REF! cells, the =A1*1.075 magic numbers. Because both members ingest .xlsx, this one is explicit-only (auto_routable: false) — .xlsx continues to auto-route to records-analyser by default; invoke spreadsheet-analyser deliberately when you want formula-logic signals.

Install

pip install spreadsheet-analyser

Use

Python:

from spreadsheet_analyser import SpreadsheetAnalyser, SpreadsheetAnalysis

result: SpreadsheetAnalysis = SpreadsheetAnalyser().analyse("budget.xlsx")
print(result.overall_formulas.unique_functions)        # ['IF', 'SUM', 'VLOOKUP']
print(result.overall_formulas.max_nesting_depth)       # 4
print(result.dependencies.circular_references)         # [['Sheet1!A1', 'Sheet1!B1', 'Sheet1!A1']]
print(result.overall_errors.error_cells)               # {'#REF!': 2, '#DIV/0!': 1}

CLI:

spreadsheet-analyser budget.xlsx          # human summary
spreadsheet-analyser budget.xlsx --json   # raw JSON
spreadsheet-analyser serve                # HTTP API on port 8011
spreadsheet-analyser manifest             # capability manifest

HTTP (spreadsheet-analyser serve on port 8011):

curl -F file=@budget.xlsx http://localhost:8011/analyse
curl http://localhost:8011/health
curl http://localhost:8011/manifest

Signals

Per sheet and rolled up across the workbook:

• Formulas — count, formula-vs-constant ratio, unique functions used, max nesting depth, longest formula, volatile functions (NOW/RAND/OFFSET/INDIRECT), hardcoded magic numbers in formulas (a hygiene smell — =A1*0.15 instead of =A1*TaxRate).
• Dependencies — cross-sheet references, dependency-chain depth, circular references.
• Errors — #REF! / #DIV/0! / #VALUE! / #N/A / #NAME? / #NULL! / #NUM! cells.
• Structure — sheet count, used range per sheet, named ranges, tables, charts present.

The family

Part of the lens analyser family — small single-purpose engines that compose via a shared contract (lens-contract).

Limits

• .xlsx and .xlsm only for v1 (.ods via odfpy is a possible follow-on).
• Cycle detection is direct cell→cell only (not via named-range indirection).
• "Hardcoded magic numbers" is a heuristic (numbers other than 0/1 inside formulas after stripping cell refs) — useful as a hygiene signal, not a definitive smell.

License

MIT