dce_003 · Deployment E2E: real empty BQ source — no crash, empty result
Status: ✓ Passing · Last run: 2026-06-09T06:39:34 · Pipeline: deployment_e2e
· History: 1/1 runs passed (100%)
What was tested
Source BigQuery table has zero rows (typical: upstream pipeline hasn't run yet today). The Python section's df.to_dataframe() returns an empty DataFrame; the code branches to an explicit empty-result handler. Expect: section runs clean, 0 predictions, no errors. Production-side equivalent of dc_003 (empty_batch) — exercising the real empty-DataFrame round-trip through BigQuery's pandas adapter.
What we planted in the data
- 500 entities
- Signal strength: 0.50 (sigmoid slope multiplier)
n_new_rows=50
What the system did
- Training rows built: 500
- Features used in model: 3
- Model selected:
DeploymentE2ERunner
What we checked — all assertions passed
| Status | Assertion | Detail |
|---|---|---|
| ✓ | e2e_section_completed_cleanly |
clean run (no pipeline or section errors) |
| ✓ | e2e_scored_row_count_eq |
scored 0 rows, expected 0 |
| ✓ | e2e_predictions_in_band |
empty source — no predictions to validate (vacuously OK) |
How this could have gone wrong
(no assertion descriptions provided)
Why this case matters
An empty source is a normal operational state ("nothing happened since the last refresh"). The deployment must produce a clean, empty live_artifact — not a stack trace or a fake "0 predictions with NaN" row.
Reproducing
# from auto_insight_api/
python -m validation.v2 run dce_003 --pipeline deployment_e2e -v
- Case config:
validation/v2/cases/dce_003_e2e_empty_source.yaml - Data shape:
deployment_e2e_empty_source - Analytics type:
deployment_e2e