# -*- coding: utf-8 -*-
"""五段加法二宿主复验 · Windows/IDBots · 真实(Twin) · 最终版
Wuhao 件: pin://dd342aa4e7110d22cddf964a8671fea39e7fc0e3370bdad05b8a51e0184d6c6ai0
模型: R = 35 + 骨架 + Σ条目 + max(k-1, 0)
  骨架 = 顶层 JSON 置空 truncation_points 后的序列化长度
  条目 = truncation_points[] 每枚单独序列化后逐个相加
  序列化: json.dumps(ensure_ascii=False, separators=(",", ":"))
字符口径: python len() = Unicode code point(Git Bash wc -m 按字节计,禁用)
"""
import json, io, os

SPILL = r"C:\Users\vinso\AppData\Roaming\IDBots\dsh-sessions\v0\spill"
PREFIX = "- RECEIPT read_metaweb_pins_batch: "
NUM_KEYS = ["requested", "readable_upstream", "returned", "omitted", "unreadable",
            "truncation_point_count", "bytes_written"]

def ser(obj):
    return json.dumps(obj, ensure_ascii=False, separators=(",", ":"))

rows = []
for root, dirs, files in os.walk(SPILL):
    for fn in files:
        p = os.path.join(root, fn)
        try:
            txt = io.open(p, encoding="utf-8", errors="strict").read()
        except Exception:
            continue
        for line in txt.splitlines():
            if not line.startswith(PREFIX):
                continue
            try:
                d = json.loads(line[len(PREFIX):])
            except Exception:
                continue
            if isinstance(d, dict) and "truncation_points" in d:
                rows.append((p, line, d))

print("n =", len(rows))
resid_zero = 0
print("\n[1] 五段加法残差(R 实测 vs 35+骨架+Σ条目+max(k-1,0))")
table = []
for p, line, d in rows:
    R = len(line)
    k = len(d["truncation_points"])
    top = dict(d); top["truncation_points"] = []
    sk = len(ser(top))
    ent = sum(len(ser(e)) for e in d["truncation_points"])
    pred = 35 + sk + ent + max(k - 1, 0)
    resid = R - pred
    if resid == 0: resid_zero += 1
    table.append((os.path.basename(p)[:26], R, k, sk, ent, pred, resid))
    print(f"  {os.path.basename(p)[:26]} | R={R} k={k:2} 骨架={sk} Σ条目={ent:4} 预测={pred} 残差={resid:+d}")
print(f"  残差零: {resid_zero}/{len(rows)}")

print("\n[2] 骨架结构常数(本席 12 键变体: 结构 = 212/213 + caliber(64) + Σ7数字位宽 + id项)")
for p, line, d in rows:
    top = dict(d); top["truncation_points"] = []
    sk = len(ser(top))
    calib_len = len(d["caliber"])
    widths = sum(len(str(d[kk])) - 1 for kk in NUM_KEYS)
    id_len = len(ser(d["omitted_ids"])) + len(ser(d["unreadable_ids"])) - 4
    base = 212 if d["truncated"] else 213
    recon = base + calib_len + widths + id_len
    print(f"  {os.path.basename(p)[:26]} | 骨架 {sk} = {base}+{calib_len}+{widths}+{id_len} = {recon} {'OK' if recon==sk else 'FAIL'}")

print("\n[3] 条目结构常数(entry base 99; total_bytes=null 变体 102)")
e0 = ser(dict(pin_id="", reason="", kept_chars=0, total_chars=0, cut_at_bytes=0, total_bytes=0, unit=""))
e0n = ser(dict(pin_id="", reason="", kept_chars=0, total_chars=0, cut_at_bytes=0, total_bytes=None, unit=""))
print(f"  base(all-zero)={len(e0)}  base(total_bytes=null)={len(e0n)}")
bad = 0; tot = 0
for p, line, d in rows:
    for e in d["truncation_points"]:
        tot += 1
        eb = dict(e); eb["pin_id"] = ""; eb["reason"] = ""
        eb["kept_chars"] = eb["total_chars"] = eb["cut_at_bytes"] = 0
        if "total_bytes" in eb: eb["total_bytes"] = 0 if eb["total_bytes"] is not None else None
        ebase = len(ser(eb))
        add = (len(e["pin_id"]) - 2) + (len(e["reason"]) - 2)
        add += sum(len(str(e[kk])) - 1 for kk in ("kept_chars", "total_chars", "cut_at_bytes"))
        if e.get("total_bytes") is not None:
            add += len(str(e["total_bytes"])) - 1
        else:
            add += 3  # null(4) 相对 0(1)
        if len(ser(e)) != ebase + add: bad += 1
print(f"  条目逐枚零残差: {tot-bad}/{tot}")

print("\n[4] 容器字符/字节与 slack(饱和帧 = truncated 且 tp==N)")
sat_chars = []
for p, line, d in rows:
    txt = io.open(p, encoding="utf-8").read()
    chars = len(txt); byts = os.path.getsize(p); slack = 18000 - chars
    sat = d["truncated"] and d["truncation_point_count"] == d["requested"]
    if sat: sat_chars.append(chars)
    print(f"  {os.path.basename(p)[:26]} | sat={sat} | chars={chars} bytes={byts} slack={slack}")
if sat_chars:
    import statistics
    print(f"  饱和帧字符: {sat_chars}  mean={statistics.mean(sat_chars):.1f}  dev_max={max(abs(c-17984) for c in sat_chars)} ({max(abs(c-17984) for c in sat_chars)/17984*100:.3f}%)")

print("\n[5] 三枚必红臂(先证明会红)")
# 取一枚 k=12 饱和帧做臂
target = next((p, line, d) for p, line, d in rows if len(d["truncation_points"]) == 12)
p, line, d = target
R = len(line); k = len(d["truncation_points"])
one_entry = len(ser(d["truncation_points"][0]))
pred_missing = 35 + len(ser(dict(d, truncation_points=[]))) + sum(len(ser(e)) for e in d["truncation_points"][1:]) + max(k - 2, 0)
print(f"  A1 删一枚条目对象: 实测 R={R}, 按少一枚预测={pred_missing}, 偏差={R-pred_missing} (≈条目{one_entry}+逗号1) -> {'RED-OK' if R-pred_missing >= one_entry else 'NO-RED'}")
d2 = dict(d); d2["x_injected"] = 0
R2 = len(PREFIX + ser(d2))
print(f"  A2 骨架塞字段: 原 R={R}, 注入后实测={R2}, 结构常数偏移={R2-R} -> {'RED-OK' if R2-R > 0 else 'NO-RED'}")
txt = io.open(p, encoding="utf-8").read()
chars = len(txt); byts = os.path.getsize(p)
print(f"  A3 字节当字符: 容器字符={chars}<=18000 而字节={byts}>18000 -> 字节口径假超帽 {'RED-OK' if byts > 18000 >= chars else 'NO-RED'}")