Export PCB Quality Summaries from a Consistent Data Snapshot

A quality report may run several queries while new test results are arriving. If each query sees a different state, the reported total, failure count, and detailed rows can disagree even though each query is correct on its own. Define a consistent reporting boundary before exporting the package.

SQLite’s isolation documentation explains how transactions and journal modes affect what readers observe, including snapshot behavior in WAL mode. Use the guarantees of the actual database configuration rather than assuming that separate queries automatically share one view.

Choose the reporting boundary

State the included production period, product revisions, stations, and test-attempt policy. Decide how to handle attempts still in progress at the cutoff. A report based on completed attempts answers a different question from one that counts every started board.

Use a consistent snapshot or another deliberate export mechanism supported by the system. Record the boundary identity or retrieval context needed to reproduce the report. A timestamp written on the cover page is insufficient if the underlying queries each selected “current” data independently.

Reconcile the report components

Account for late-arriving results and corrected records. Decide whether they belong in a revised report or a later reporting period. Do not quietly update one worksheet while leaving the summary page and exported PDF from an earlier state.

Verify export behavior under activity

1. Prepare a known dataset with identifiable records.
2. Begin the report export through the production path.
3. Introduce controlled new or updated records during the export.
4. Confirm that all report sections follow the defined boundary.
5. Store the report identity, export configuration, and reconciliation evidence.

Database consistency does not automatically include files stored elsewhere. If the report links to photographs or waveforms, ensure the export process captures the intended versions and detects unavailable attachments. Treat the database and external-file relationship explicitly.

Keep revised reports distinguishable from their predecessors. Record why a report changed, who reviewed the correction, and which data were added or corrected. A stable filename that is repeatedly overwritten makes it difficult to determine which version supported an earlier engineering decision.

The objective is a coherent evidence set with one population definition. That allows differences between successive reports to reflect actual new information instead of accidental timing between queries.