#!/usr/bin/env python3
"""Aggregate equally sized benchmark trials from the recorded Ubuntu 24.04 runs."""

import json
import statistics
from pathlib import Path

root = Path(__file__).resolve().parent
raw = root / "raw"
groups = {
    "gate": {
        "old": ["old-gate.json", "old-gate2.json"],
        "new": ["new-gate.json", "new-gate2.json"],
    },
    "db": {
        "old": ["old-db.json", "old-db2.json", "old-db3.json"],
        "new": ["new-db.json", "new-db2.json", "new-db3.json"],
    },
    "incidents": {
        "old": ["old-incidents.json", "old-incidents2.json"],
        "new": ["new-incidents.json", "new-incidents2.json"],
    },
    "internal": {
        "old": ["old.json", "old2.json"],
        "new": ["new.json", "new2.json"],
    },
}


def collect(files):
    cases = {}
    for name in files:
        document = json.loads((raw / name).read_text())
        for case, record in document["cases"].items():
            trials = record if isinstance(record, list) else record["trials"]
            iterations = document.get("samples_per_trial", record.get("iterations_per_trial") if isinstance(record, dict) else None)
            if iterations is None:
                raise ValueError(f"Missing iteration count in {name}:{case}")
            entry = cases.setdefault(case, {"iterations_per_trial": iterations, "trials": []})
            if entry["iterations_per_trial"] != iterations:
                raise ValueError(f"Iteration count drift in {name}:{case}")
            entry["trials"].extend(trials)
    return cases


def summarize(entry, group):
    trials = entry["trials"]
    result = {
        "trial_count": len(trials),
        "samples": len(trials) * entry["iterations_per_trial"],
        "mean_ms": statistics.mean(t["mean_ms"] for t in trials),
        "median_ms": statistics.median(t["median_ms"] for t in trials),
        "median_trial_p95_ms": statistics.median(t["p95_ms"] for t in trials),
        "max_ms": max(t["max_ms"] for t in trials),
    }
    if "queries_min" in trials[0]:
        result["queries_min"] = min(t["queries_min"] for t in trials)
        result["queries_max"] = max(t["queries_max"] for t in trials)
    elif group == "gate":
        result["queries_min"] = result["queries_max"] = 1
    return result


summary = {
    "before_revision": "b9affcc8601eca978f28c640e62e305d487b6ef0",
    "after_revision": "d899eef4aad5e50e1297c66c130af5e85de28394",
    "environment": "Ubuntu 24.04.4 VM 96276; PHP 8.4.26; MariaDB 10.11.14",
    "units": "milliseconds per treatment",
    "groups": {},
}
for group, versions in groups.items():
    old = collect(versions["old"])
    new = collect(versions["new"])
    if set(old) != set(new):
        raise ValueError(f"Case mismatch in {group}")
    values = {}
    for case in old:
        before = summarize(old[case], group)
        after = summarize(new[case], group)
        if before["samples"] != after["samples"]:
            raise ValueError(f"Sample mismatch in {group}:{case}")
        values[case] = {
            "before": before,
            "after": after,
            "median_delta_percent": (after["median_ms"] / before["median_ms"] - 1) * 100,
        }
    summary["groups"][group] = values

(root / "comparison.json").write_text(json.dumps(summary, ensure_ascii=False, indent=2) + "\n")
print(f"Wrote {root / 'comparison.json'} with {sum(map(len, summary['groups'].values()))} cases")