je_load_density.utils.regression.diff

Cross-run regression diff.

Loads two runs from a SQLite persistence database and compares per-name latency percentiles + failure rate. Returns a structured diff so CI can gate on regressions above a configurable tolerance.

Functions

diff_runs(database_path, baseline_run_id, ...)	Compare two runs in database_path.
summarise_records(records)	Aggregate raw records into {name: {count, failures, p50, p95, p99, mean}}.

je_load_density.utils.regression.diff.diff_runs(database_path: str, baseline_run_id: int, current_run_id: int, tolerance: float = 0.1) → Dict[str, Any]

Compare two runs in database_path. tolerance is the fractional increase allowed (e.g. 0.10 = 10 %). Metrics that breach the tolerance are listed under regressions.

je_load_density.utils.regression.diff.summarise_records(records: Iterable[Dict[str, Any]]) → Dict[str, Dict[str, float]]

Aggregate raw records into {name: {count, failures, p50, p95, p99, mean}}.