#!/usr/bin/env python3 """Oracle and analysis for the rotation differential test. Oracle: majority / pairwise agreement among the implementations on the pool index shown at each query instant. There is no independent ground truth; the Dart app's reconciliation is treated as the reference behaviour because the app republishes the widget whenever it is opened, so any native answer that differs from it is a visible jump. Every disagreement is assigned a root-cause class by explicit, mechanical rules (below) and the residue is reported as 'unexplained'. """ import json, sys, glob, os, collections from datetime import datetime, timezone, timedelta from zoneinfo import ZoneInfo R = sys.argv[1] OUT = os.path.join(R, "out") scen = {s["id"]: s for s in json.load(open(os.path.join(OUT, "scenarios.json")))["scenarios"]} zones = json.load(open(os.path.join(OUT, "scenarios.json")))["zones"] def minute_status(zone, q_ms, minute): """Is local minute `minute` on q's local date a gap, an overlap or normal?""" tz = ZoneInfo(zone) d = datetime.fromtimestamp(q_ms / 1000, tz).date() a = datetime(d.year, d.month, d.day, minute // 60, minute % 60, tzinfo=tz, fold=0) b = a.replace(fold=1) ua, ub = a.astimezone(timezone.utc), b.astimezone(timezone.utc) if ua.astimezone(tz).replace(tzinfo=None) != a.replace(tzinfo=None): return "gap" if ua != ub: return "overlap" return "normal" tot = collections.Counter() # (kind) -> queries agree = collections.Counter() # (kind, pair) -> agreements cls = collections.Counter() # root-cause class -> query count cls_scen = collections.defaultdict(set) cls_zone = collections.defaultdict(set) examples = collections.defaultdict(list) per_zone = collections.defaultdict(lambda: collections.Counter()) timing = collections.defaultdict(dict) runtimes = {} PAIRS = ["app=swift", "app=kotlin", "swift=kotlin", "all3", "cpi=app", "swiftShown=swift", "kotlinShown=kotlin", "swiftShown=kotlinShown"] for zone in zones: f = zone.replace("/", "_") d = json.load(open(os.path.join(OUT, f"dart_{f}.json"))) s = json.load(open(os.path.join(OUT, f"swift_{f}.json"))) k = json.load(open(os.path.join(OUT, f"kotlin_{f}.json"))) assert s["tzCurrent"] in (zone, "GMT") and k["tzDefault"] == zone, (zone, s["tzCurrent"], k["tzDefault"]) runtimes["dart"] = d["runtime"]; runtimes["swift_tz"] = s["tzDataVersion"]; runtimes["kotlin"] = k["runtime"]; runtimes["kotlin_tz"] = k["tzDataVersion"] timing[zone] = {"dart_us": d["elapsedMicros"], "swift_us": s["elapsedMicros"], "kotlin_us": k["elapsedMicros"], "queries": d["queries"]} assert s["ids"] == [r["id"] for r in d["results"]] == k["ids"] for i, r in enumerate(d["results"]): sc = scen[r["id"]] kind = r["kind"] p = r["payload"] for j, q in enumerate(r["queries"]): a, c = r["dart_app"][j], r["dart_cpi"][j] si, ki = s["instant"][i][j], k["instant"][i][j] ss, ks = s["shown"][i][j], k["shown"][i][j] tot[kind] += 1; tot["ALL"] += 1 per_zone[zone]["queries"] += 1 res = {"app=swift": a == si, "app=kotlin": a == ki, "swift=kotlin": si == ki, "all3": a == si == ki, "cpi=app": c == a, "swiftShown=swift": ss < 0 or ss == si, "kotlinShown=kotlin": ks < 0 or ks == ki, "swiftShown=kotlinShown": ss < 0 or ks < 0 or ss == ks} for pr, ok in res.items(): if ok: agree[(kind, pr)] += 1; agree[("ALL", pr)] += 1 if not res["all3"]: per_zone[zone]["instantDisagree"] += 1 # ---- root-cause rules ---- labels = [] if not res["all3"]: if si == ki and a != si: t0 = sc["t0"]; I = p["adelante_refresh_interval_seconds"] * 1000 if p["adelante_schedule_dates"] == "[]" and I > 0: nb = (t0 // 1000) * 1000; ph_n = (q - nb) % I; ph_a = (q - t0) % I labels.append("D1 sub-second anchor truncation" if ph_n < ph_a else "app-vs-native other (interval)") elif p["adelante_initial_quote_only"]: labels.append("D2 Today-override +24h across DST (app)") else: labels.append("D7 Dart elapsed count vs its own published list (gap)") elif a == si and ki != si: st = minute_status(zone, q, sc["minute"]) if sc["minute"] is not None else "n/a" labels.append(f"D3 Kotlin tail boundary ({st})") elif a == ki and si != ki: dates = json.loads(p["adelante_schedule_dates"]) last_day = datetime.fromtimestamp(dates[-1], ZoneInfo(zone)).date() if dates else None if last_day and minute_status(zone, int(datetime(last_day.year, last_day.month, last_day.day, 12, tzinfo=ZoneInfo(zone)).timestamp() * 1000), 0) == "gap": labels.append("D4b Swift tail day count after a midnight transition") else: st = minute_status(zone, q, sc["minute"]) if sc["minute"] is not None else "n/a" labels.append(f"D4a Swift tail boundary ({st})") else: labels.append("three-way disagreement") if not res["swiftShown=swift"]: st = minute_status(zone, q, sc["minute"]) if sc["minute"] is not None else "n/a" labels.append("D5 Swift timeline shows stale entry") if not res["kotlinShown=kotlin"]: st = minute_status(zone, q, sc["minute"]) if sc["minute"] is not None else "n/a" labels.append("D6 Kotlin alarm shows stale render") if not res["cpi=app"]: labels.append("C currentPoolIndex != app" + (" (daily-time)" if p["adelante_schedule_dates"] != "[]" else " (interval)")) for L in labels: cls[L] += 1; cls_scen[L].add(r["id"]); cls_zone[L].add(zone) if len(examples[L]) < 6: tz = ZoneInfo(zone) examples[L].append({"scenario": r["id"], "zone": zone, "kind": kind, "minute": sc["minute"], "mode": sc["mode"], "n": r["n"], "initialOnly": p["adelante_initial_quote_only"], "t0_local": datetime.fromtimestamp(sc["t0"] / 1000, tz).isoformat(), "q_ms": q, "q_local": datetime.fromtimestamp(q / 1000, tz).isoformat(timespec="milliseconds"), "dart_app": a, "dart_cpi": c, "swift": si, "kotlin": ki, "swift_shown": ss, "kotlin_shown": ks}) summary = { "runtimes": runtimes, "queries": dict(tot), "scenarios": len(scen), "agreement": {f"{kd}|{pr}": {"agree": agree[(kd, pr)], "total": tot[kd], "rate": round(agree[(kd, pr)] / tot[kd], 6)} for kd in tot for pr in PAIRS}, "classes": {L: {"queries": cls[L], "scenarios": len(cls_scen[L]), "zones": sorted(cls_zone[L])} for L in sorted(cls)}, "examples": examples, "perZone": {z: dict(v) for z, v in per_zone.items()}, "timing": timing, } json.dump(summary, open(os.path.join(OUT, "summary.json"), "w"), indent=1) print(json.dumps({k: summary[k] for k in ("runtimes", "queries", "scenarios")}, indent=1)) for kd in ["ALL", "interval", "daily", "daily-tail", "long", "leap"]: print(kd, {pr: summary["agreement"][f"{kd}|{pr}"]["rate"] for pr in PAIRS}) for L, v in summary["classes"].items(): print(f"{L}: {v['queries']} queries, {v['scenarios']} scenarios, {len(v['zones'])} zones")