"""E-M1 · 08. Cotejo de la implementación 1 (sql/0*_*) con la 2 (sql/impl2/), con la tolerancia del §7 del pre-registro: recuentos idénticos; b y ρ puntuales con diferencia relativa ≤ 0,5 %; extremos de los IC de ρ con diferencia ≤ 0,05; mismos veredictos. Solo biblioteca estándar. Salida: salidas/08_cotejo_impl2.md""" import csv R = "/em1/salidas" m1 = {(r["celda"], r["sens"]): r for r in csv.DictReader(open(f"{R}/04_carril_celdas.csv"))} m2 = list(csv.DictReader(open(f"{R}/impl2/carril.csv"))) MAPA = {"principal_SS_2019_2025": ("principal 2019-2025", ""), "obras_2019_2025": ("obras 2019-2025", ""), "SS_2020_2025": ("2020-2025 (sin tope)", ""), "SS_PLACSP_sola": ("solo PLACSP", ""), "sens_grado2": ("principal 2019-2025", "grado 2"), "sens_grado4": ("principal 2019-2025", "grado 4"), "sens_ajuste_7000": ("principal 2019-2025", "ajuste [7.000, 14.500)"), "sens_ajuste_10000": ("principal 2019-2025", "ajuste [10.000, 14.500)"), "sens_ventana_14000": ("principal 2019-2025", "ventana [14.000, 15.000)"), "sens_sin_ficticias": ("principal 2019-2025", "sin ficticias de redondeo"), "sens_sin_agente_secti": ("principal 2019-2025", "sin órganos agente_secti")} for a in range(2020, 2026): MAPA[f"SS_{a}"] = (f"año {a}", "") for t in ("age", "autonomica", "ayuntamiento", "empresa_fundacion_consorcio", "local_supramunicipal", "otro", "salud"): MAPA[f"SS_tipo_{t}"] = (f"tipo {t}", "") MAPA["SS_tipo_investigacion"] = ("tipo investigacion (contraste DA 54.ª)", "") out = ["# Cotejo implementación 1 frente a 2 (`08_cotejo_impl2.py`)\n\n## Test de Carril\n\n", "| test impl2 | celda impl1 | recuentos (9) idénticos | máx. dif. rel. b | dif. rel. ρ_b | dif. rel. ρ_c | máx. dif. abs. extremos IC ρ | veredicto 1 | veredicto 2 | ¿en tolerancia? |\n", "|---|---|---|---|---|---|---|---|---|---|\n"] fallos = 0 vistos = set() for r2 in m2: t = r2["test"] if t == "SS_2019": clave = ("2019 (con tope 118.3)", "") if t not in vistos else ("año 2019", "") vistos.add(t) elif t in MAPA: clave = MAPA[t] else: out.append(f"| {t} | (sin equivalente en impl1) | | | | | | | | |\n") continue r1 = m1[clave] cnt = all(int(float(r1[k1])) == int(float(r2[k2])) for k1, k2 in (("n_a", "n_a"), ("n_b", "n_b"), ("n_c", "n_c"), ("nv_a", "n_a_ventana"), ("nv_b", "n_b_ventana"), ("nv_c", "n_c_ventana"))) rel = lambda a, b: abs(float(a) - float(b)) / abs(float(b)) if float(b) != 0 else float("nan") db = max(rel(r1[k], r2[k]) for k in ("b_a", "b_b", "b_c")) insuf2 = r2["muestra_insuficiente"] == "True" sin_amont = r1["resultado"].startswith("sin amontonamiento") if sin_amont or insuf2: drb = drc = dci = float("nan") else: drb, drc = rel(r1["rho_b"], r2["rho_b"]), rel(r1["rho_c"], r2["rho_c"]) dci = max(abs(float(r1[a]) - float(r2[b])) for a, b in (("rho_b_lo", "rho_b_ic_lo"), ("rho_b_hi", "rho_b_ic_hi"), ("rho_c_lo", "rho_c_ic_lo"), ("rho_c_hi", "rho_c_ic_hi"))) v1 = r1["resultado"] v2 = f"{r2['resultado']} · {r2['firmeza']}" mismo = v1.replace(" · ", " ").replace("en el límite", "en_el_limite").split()[0] == r2["resultado"].split()[0] ok = cnt and db <= 0.005 and (sin_amont or (drb <= 0.005 and drc <= 0.005 and dci <= 0.05)) and mismo fallos += (not ok) f = lambda v, k=4: "—" if v != v else f"{v:.{k}f}".replace(".", ",") out.append(f"| {t} | {clave[0]}{' · ' + clave[1] if clave[1] else ''} | {'sí' if cnt else 'NO'} | {f(db, 6)} | {f(drb, 6)} | {f(drc, 6)} | {f(dci, 4)} | {v1} | {v2} | {'sí' if ok else '**NO**'} |\n") # Migración por celda (resto) c1 = {(r["source"], r["base"]): r for r in csv.DictReader(open(f"{R}/02_migracion_celdas.csv")) if r["clase"] == "resto" and r["dedup"] == "vista_unica"} c2 = {(r["source"], r["base"]): r for r in csv.DictReader(open(f"{R}/impl2/migracion_celdas.csv")) if r["clase"] == "resto"} out.append("\n## Migración por celda (resto, vista única)\n\n| fuente | base | n_norm antes / d2T (1 · 2) | banda vieja (1 · 2) | evaluable (1 · 2) | migra (1 · 2) | d2T/antes (1 · 2) | ¿igual? |\n|---|---|---|---|---|---|---|---|\n") b = lambda s: str(s).strip().lower() == "true" for k in sorted(set(c1) | set(c2)): a, z = c1.get(k), c2.get(k) if a is None or z is None: out.append(f"| {k[0]} | {k[1]} | {'solo impl2' if a is None else 'solo impl1'} | | | | | **NO** |\n") fallos += 1 continue bv1 = {"V": "V", "V1": "V1", "V2": "V2"}[a["banda_vieja"]] bv2 = z["banda_vieja"] r1 = 1 - float(a["caida_vieja"]) if a["caida_vieja"] not in ("", None) else float("nan") r2 = float(z["vieja_ratio_d2T_antes"]) if z["vieja_ratio_d2T_antes"] not in ("", None) else float("nan") igual = (int(a["n_norm_antes"]) == int(float(z["n_norm_antes"])) and int(a["n_norm_d2t"]) == int(float(z["n_norm_despues2T"])) and bv1 == bv2 and b(a["evaluable"]) == b(z["evaluable"]) and b(a["migracion"]) == b(z["migra"])) fallos += (not igual) f = lambda v: "—" if v != v else f"{v:.3f}".replace(".", ",") out.append(f"| {k[0]} | {k[1]} | {a['n_norm_antes']}/{a['n_norm_d2t']} · {int(float(z['n_norm_antes']))}/{int(float(z['n_norm_despues2T']))} | {bv1} · {bv2} | " f"{b(a['evaluable'])} · {b(z['evaluable'])} | {b(a['migracion'])} · {b(z['migra'])} | {f(r1)} · {f(r2)} | {'sí' if igual else '**NO**'} |\n") out.append(f"\n**Filas fuera de tolerancia: {fallos}.** (La caída de impl1 está redondeada a 3 decimales en el CSV.)\n") open(f"{R}/08_cotejo_impl2.md", "w").write("".join(out)) print("".join(out))