Function two_level_summary

benchmarks/src/postprocessing/overview.py:243–267

(
    df: pd.DataFrame,
    primary_grouping: Callable[[pd.DataFrame], Any],
    secondary_grouping: Callable[[pd.DataFrame], Any],
    file,
    print_total=False,
)

def two_level_summary(
    df: pd.DataFrame,
    primary_grouping: Callable[[pd.DataFrame], Any],
    secondary_grouping: Callable[[pd.DataFrame], Any],
    file,
    print_total=False,
):
    primary_group = primary_grouping(df)

    with pd_print_all():
        for group, data in sorted(primary_group, key=lambda item: item[0]):
            print(" ".join(group), file=file)

            secondary_group = secondary_grouping(data)
            for group, results in sorted(secondary_group, key=lambda item: item[0]):
                stats = results["duration"].describe()
                duration_str = f"{stats['mean']:.4f} s"
                if stats["count"] > 1:
                    duration_str += f" (min={stats['min']:.4f}, max={stats['max']:.4f}, std={stats['std']:.4f})"
                print(f"\t{' '.join(group)}: {duration_str}", file=file)
            if print_total:
                mean_duration = data["duration"].mean()
                print(f"(mean): {mean_duration:.4f} s", file=file)

            print(file=file)


def generate_comparison_html(benchmarks: List[str], database: Database, directory: Path):