Check whether the outcome improved
Exercise early and late exceptions
Place a mixed-type, oversized, or otherwise relevant exceptional value inside and beyond the preview sample. Compare inferred settings and expected outcomes for each controlled fixture. Include an ordinary file of similar size so the correction does not unnecessarily reject a valid import merely because the sample is bounded.
Check the readiness statement
Observe whether the interface distinguishes sampled display, completed validation, and queued execution. Confirm any claim about full-file checking against the actual path. When a later problem is found, inspect its location and correction instructions so users can identify the relevant record rather than repeatedly editing the visible sample.
Inspect final accepted and rejected state
Execute the supported fixture cases and compare stored values or rejection outcomes with independent expectations. Check that a corrected file can be retried through the intended workflow. Keep partial completion and coercion visible if supported, and verify they are communicated rather than hidden behind a generally successful preview.
What to carry forward
Accept when sample placement cannot conceal an unsupported value behind a false readiness claim and execution follows the documented contract. Record which exception types were exercised. Leave untested inference categories explicit; one late text value does not verify every structural, length, or field-specific validation boundary.
Technical background: DuckDB documentation.
Keep the decision with the work.
Use a Work Item in Aglet to record the problem, the evidence you have, and the next decision. Add an owner and priority, then keep updates in the discussion so the next person can follow the reasoning.
Create an account See the product workflow