How should headless spreadsheet engines handle stale XLSX cached values? #382
Replies: 2 comments
-
|
Update: The short version is unchanged: embedded XLSX cached formula values are useful triage data, but not an accuracy verdict. For a correctness claim, the expected value should come from a fresh Microsoft Excel recalculation, and workbooks without that oracle should stay labeled Useful links:
If you have a workbook corpus where cached values and recalculated Excel values disagree, sanitized examples would help tune the report format and issue templates. |
Beta Was this translation helpful? Give feedback.
-
|
Update, May 14: The current split is:
Links:
The part I still want feedback on: what fields should be mandatory in a public issue before a formula mismatch is credible enough to call a library bug? Right now I am treating formula text, fresh Excel expected value, Bilig actual value, workbook context, and repro notes as the minimum. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
When people test a headless spreadsheet engine against XLSX files, it is easy to compare against the cached formula values embedded in the file.
That cache is useful, but it can be stale. It should not be treated as the truth for an accuracy claim.
The rule I am using for
@bilig/headlessnow is stricter:missing_excel_oracleinstead of being promoted to a bugThe harness commands are:
Docs: https://proompteng.github.io/bilig/xlsx-corpus-verifier-walkthrough.html
Source: https://github.com/proompteng/bilig/blob/main/docs/xlsx-corpus-verifier-walkthrough.md
I would like feedback from people who evaluate spreadsheet engines or workbook automation libraries:
The goal is not to claim blanket Excel compatibility. The goal is to stop stale caches from producing fake wins or fake bugs.
Beta Was this translation helpful? Give feedback.
All reactions