Generate
Your application writes actual.xlsx.
Excel regression testing
Catch unexpected changes in generated Excel reports before users do.
Run the same workbook comparison after an export job, application release, or data-pipeline change. WorkbookLens CLI returns an actionable exit code and saves evidence for the person who investigates.
A durable test pattern
Keep the baseline, actual file, and evidence together.
Compare
WorkbookLens checks it against approved.xlsx.
Decide
The job reads the exit code and retains the reports.
PowerShell example
.workbooklens.exe ".aselinesapproved.xlsx" ".outputactual.xlsx" --keyed "Orders:A" --html ".artifactsworkbook-diff.html" --json ".artifactsworkbook-diff.json"
if ($LASTEXITCODE -eq 1) { throw "Excel report changed; review workbook-diff.html" }
if ($LASTEXITCODE -eq 2) { throw "Workbook comparison could not run" }Reduce false alarms
Compare business identity, not accidental row position.
Generated reports often sort differently after a database query, dependency update, or harmless implementation change. If a worksheet represents records, use a stable key such as Orders:A so the test follows each order ID.
This turns a noisy row-by-row mismatch into a focused result: records added, records removed, and fields changed within records that still exist.
Example test evidenceA-1002 · Amount changed: 480 → 525A-1003 · Record addedAll other keyed records unchanged
Good regression controls
Make failures easy to reproduce and review.
Approve baselines deliberately
Store a known-good workbook produced from controlled input. Replace it only after a reviewer accepts the intended change.
Archive human and machine output
HTML supports review; JSON supports dashboards, notifications, and downstream analysis.
Separate change from error
Treat exit code 1 as a comparison finding and exit code 2 as an execution problem. They require different investigations.
Implementation questions
Before adding the check to CI.
Does the build agent need Excel?
No. WorkbookLens is self-contained and reads the workbook directly.
Can a reviewer open the report without WorkbookLens?
Yes. The HTML report is self-contained and can be archived as an ordinary build artifact.
Should every workbook difference fail the build?
That depends on your release control. A common pattern is to fail or flag the job, attach the HTML report, and require a reviewer to approve a new baseline when the change is intentional.