Subtask of #56. Lock down the end-to-end behavior with realistic pipelines as test fixtures. These also double as the canonical examples in docs/python/expressions.mdx.
Four pipelines, each in examples/pipelines/<name>.py and exercised by python/tests/v0/test_pipeline_<name>.py:
txn_fraud.py — Pipeline 1 from the capstone discussion (~22 features). Exercises Core / Sketch / Velocity / Recency / Decay / Geo families with @bv.expr-derived predicates.pageview_behavior.py — Pipeline 2 (~20 features). Exercises source-level filter, with_columns, the full operator-family span.sensor_anomaly.py — IoT telemetry: trend, trend_residual, z_score, outlier_count, seasonal_deviation, streak.playback_behavior.py — Content/video: event_type filtering, burst_count over filtered events, decayed_count, entropy.
Each test:
- Registers the pipeline against a real embedded server.
- Pushes a fixed sequence of N events.
- Calls
app.get(...) for each derived feature.
- Asserts exact computed values against a golden file.
Done when all four pipelines register, push, and read back the expected feature values. ~400 LOC tests + ~300 LOC pipeline definitions.
Subtask of #56. Lock down the end-to-end behavior with realistic pipelines as test fixtures. These also double as the canonical examples in
docs/python/expressions.mdx.Four pipelines, each in
examples/pipelines/<name>.pyand exercised bypython/tests/v0/test_pipeline_<name>.py:txn_fraud.py— Pipeline 1 from the capstone discussion (~22 features). Exercises Core / Sketch / Velocity / Recency / Decay / Geo families with@bv.expr-derived predicates.pageview_behavior.py— Pipeline 2 (~20 features). Exercises source-levelfilter,with_columns, the full operator-family span.sensor_anomaly.py— IoT telemetry: trend, trend_residual, z_score, outlier_count, seasonal_deviation, streak.playback_behavior.py— Content/video: event_type filtering, burst_count over filtered events, decayed_count, entropy.Each test:
app.get(...)for each derived feature.Done when all four pipelines register, push, and read back the expected feature values. ~400 LOC tests + ~300 LOC pipeline definitions.