first release

tools/audit_db.sql

@@ -0,0 +1,96 @@
+-- rond-with-you 数据一致性审计（PostgreSQL）
+-- 用法：psql "<dsn>" -f tools/audit_db.sql
+-- 各项期望值均为 0；非 0 说明有漂移或悬空引用，逐项排查。
+-- 注意：Windows + Git Bash 下 psql 不能直接吃中文命令行参数（会被按 GBK 传出去），
+-- 所以这个文件必须保存为 UTF-8 并用 -f 执行。
+\pset border 1
+
+\echo '=== 1. 聚合统计漂移（应为 0） ==='
+SELECT 'place.visit_count drift' AS chk, count(*) AS n
+  FROM places p
+ WHERE COALESCE(p.visit_count,0) <> (SELECT count(*) FROM visits v WHERE v.place_id=p.id)
+UNION ALL
+SELECT 'datasets.visit_count drift', count(*)
+  FROM datasets d
+ WHERE COALESCE(d.visit_count,0) <> (SELECT count(*) FROM visits v WHERE v.dataset_id=d.id)
+UNION ALL
+SELECT 'datasets.place_count drift', count(*)
+  FROM datasets d
+ WHERE COALESCE(d.place_count,0) <> (SELECT count(*) FROM places p WHERE p.dataset_id=d.id)
+UNION ALL
+SELECT 'duration_min wrong', count(*)
+  FROM visits WHERE departure IS NOT NULL
+   AND COALESCE(duration_min,-1) <> GREATEST(0, floor(extract(epoch FROM (departure-arrival))/60)::int)
+UNION ALL
+SELECT 'departure < arrival', count(*)
+  FROM visits WHERE departure IS NOT NULL AND departure < arrival;
+
+\echo '=== 2. 悬空引用（应为 0） ==='
+SELECT 'visits -> other dataset place' AS chk, count(*)
+  FROM visits v JOIN places p ON p.id=v.place_id WHERE p.dataset_id<>v.dataset_id
+UNION ALL
+SELECT 'movements.from_place', count(*)
+  FROM movements m WHERE m.from_place_id IS NOT NULL
+   AND NOT EXISTS (SELECT 1 FROM places p WHERE p.id=m.from_place_id)
+UNION ALL
+SELECT 'movements.to_place', count(*)
+  FROM movements m WHERE m.to_place_id IS NOT NULL
+   AND NOT EXISTS (SELECT 1 FROM places p WHERE p.id=m.to_place_id)
+UNION ALL
+SELECT 'movements.from_visit_src', count(*)
+  FROM movements m WHERE m.from_visit_src IS NOT NULL
+   AND NOT EXISTS (SELECT 1 FROM visits v WHERE v.dataset_id=m.dataset_id AND v.src_pk=m.from_visit_src)
+UNION ALL
+SELECT 'movements.to_visit_src', count(*)
+  FROM movements m WHERE m.to_visit_src IS NOT NULL
+   AND NOT EXISTS (SELECT 1 FROM visits v WHERE v.dataset_id=m.dataset_id AND v.src_pk=m.to_visit_src)
+UNION ALL
+SELECT 'weather.visit_src', count(*)
+  FROM weather w WHERE w.visit_src IS NOT NULL
+   AND NOT EXISTS (SELECT 1 FROM visits v WHERE v.dataset_id=w.dataset_id AND v.src_pk=w.visit_src)
+UNION ALL
+SELECT 'visit_tags dangling', count(*)
+  FROM visit_tags vt
+ WHERE NOT EXISTS (SELECT 1 FROM visits v WHERE v.id=vt.visit_id)
+    OR NOT EXISTS (SELECT 1 FROM tags t WHERE t.id=vt.tag_id)
+UNION ALL
+SELECT 'visit_tags cross-dataset', count(*)
+  FROM visit_tags vt
+  JOIN visits v ON v.id=vt.visit_id JOIN tags t ON t.id=vt.tag_id
+ WHERE v.dataset_id<>t.dataset_id
+UNION ALL
+SELECT 'movements end<start', count(*)
+  FROM movements WHERE started_at IS NOT NULL AND ended_at IS NOT NULL AND ended_at < started_at;
+
+\echo '=== 3. 留档与现役表的差集（skipped 行属正常，见下方明细） ==='
+SELECT r.entity, count(*) FILTER (WHERE r.skipped) AS skipped_ok, count(*) FILTER (WHERE NOT r.skipped) AS should_be_zero
+  FROM entity_raw r
+ WHERE (r.entity='ZLOCATION'      AND NOT EXISTS (SELECT 1 FROM places p     WHERE p.dataset_id=r.dataset_id AND p.src_pk=r.src_pk))
+    OR (r.entity='ZVISIT'         AND NOT EXISTS (SELECT 1 FROM visits v     WHERE v.dataset_id=r.dataset_id AND v.src_pk=r.src_pk))
+    OR (r.entity='ZMOVEMENT'      AND NOT EXISTS (SELECT 1 FROM movements m  WHERE m.dataset_id=r.dataset_id AND m.src_pk=r.src_pk))
+    OR (r.entity='ZHOURLYWEATHER' AND NOT EXISTS (SELECT 1 FROM weather w    WHERE w.dataset_id=r.dataset_id AND w.src_pk=r.src_pk))
+    OR (r.entity='ZACTIVITY'      AND NOT EXISTS (SELECT 1 FROM activities a WHERE a.dataset_id=r.dataset_id AND a.src_pk=r.src_pk))
+    OR (r.entity='ZTAG'           AND NOT EXISTS (SELECT 1 FROM tags t       WHERE t.dataset_id=r.dataset_id AND t.src_pk=r.src_pk))
+ GROUP BY r.entity;
+
+\echo '=== 4. 笔记本/投票按 src_pk 关联，是否落在当前数据集（>0 即跨数据集错位） ==='
+SELECT 'notes not in active dataset' AS chk, count(*) FROM place_notes n
+ WHERE NOT EXISTS (SELECT 1 FROM places p WHERE p.dataset_id=(SELECT id FROM datasets WHERE is_active ORDER BY id DESC LIMIT 1) AND p.src_pk=n.src_pk)
+UNION ALL
+SELECT 'votes not in active dataset', count(*)
+  FROM place_votes v
+ WHERE NOT EXISTS (SELECT 1 FROM places p WHERE p.dataset_id=(SELECT id FROM datasets WHERE is_active ORDER BY id DESC LIMIT 1) AND p.src_pk=v.src_pk);
+
+\echo '=== 5. 数据特征（非缺陷，仅供了解） ==='
+SELECT 'places 同名同坐标重复组数' AS item, count(*)::text AS v FROM (
+  SELECT 1 FROM places GROUP BY name, round(lat::numeric,5), round(lon::numeric,5) HAVING count(*)>1) t
+UNION ALL
+SELECT 'visits 零时长', count(*)::text FROM visits WHERE duration_min=0
+UNION ALL
+SELECT 'visits 超 30 天', count(*)::text FROM visits WHERE duration_min>43200
+UNION ALL
+SELECT 'visits 无活动', count(*)::text FROM visits WHERE activity_id IS NULL
+UNION ALL
+SELECT 'places 无类别', count(*)::text FROM places WHERE COALESCE(poi_category,'')=''
+UNION ALL
+SELECT 'places 无坐标', count(*)::text FROM places WHERE COALESCE(lat,0)=0 AND COALESCE(lon,0)=0;

tools/photo_coord_audit.py

@@ -0,0 +1,354 @@
+#!/usr/bin/env python3
+"""照片导入的坐标体检与批量换算。
+
+为什么需要它：导入页把「用户选的坐标系」交给 /api/regeo（coord 字段），服务端只对 wgs84
+做换算，坐标再 toFixed(6) 回填表单落库。选错一次，显示坐标就一字不差等于 EXIF 原值，在
+GCJ-02 底图上偏 350~680m。库里没有 created_at，也没记录导入时的坐标系选择，所以判据只有
+原图 EXIF：
+
+    库内坐标 == round(EXIF, 6)        -> 没转换，需要修
+    库内坐标 == gcj(round(EXIF, 6))   -> 已转换，正确
+    （实测 24/24 命中，0 反例）
+
+没有原图时退化为「sp > rond 基线上限 = 站点新建」，只能靠 --include-unverified 才动。
+
+用法：
+    python tools/photo_coord_audit.py --site https://life.shiroko.one --cookie "uid|tok" \
+        --photos D:/photos/libo --photos D:/photos/foshan --out fix.csv
+    # 确认清单后加 --apply
+"""
+import argparse
+import csv
+import html as htmllib
+import io
+import json
+import math
+import os
+import re
+import socket
+import sys
+import time
+import urllib.error
+import urllib.parse
+import urllib.request
+
+# 生产机只监听 IPv4：默认解析出 IPv6 会先撞在超时上，而 urllib 不会自动回退
+_gai = socket.getaddrinfo
+
+
+def _gai_prefer_v4(host, port, family=0, type=0, proto=0, flags=0):
+    infos = _gai(host, port, family, type, proto, flags)
+    v4 = [i for i in infos if i[0] == socket.AF_INET]
+    return v4 or infos
+
+
+socket.getaddrinfo = _gai_prefer_v4
+
+UA = "Mozilla/5.0 (Windows NT 10.0; Win64; x64) rond-coord-audit"
+FORM_RE = re.compile(r'<form method="post" action="/admin/edit/place/update".*?</form>', re.S)
+ORIG_RE = re.compile(r'导入时的坐标：<span class="mono">([-0-9.]+),\s*([-0-9.]+)</span>')
+
+
+class _NoRedirect(urllib.request.HTTPRedirectHandler):
+    # 保存成功是 303 + Location 里带 ok=/err=，跟过去就看不到结论了；
+    # 会话失效也是 303 跳登录页，跟着走会静默拿到登录页 HTML 当成空数据
+    def redirect_request(self, req, fp, code, msg, headers, newurl):
+        return None
+
+
+_opener = urllib.request.build_opener(_NoRedirect)
+
+
+def http(url, cookie="", data=None, timeout=60, tries=3):
+    # 走 Cloudflare 时长页面偶发读超时，重试比直接失败划算；写库的重试由调用方控
+    last = None
+    for i in range(tries):
+        req = urllib.request.Request(url, data=data)
+        req.add_header("User-Agent", UA)
+        if cookie:
+            req.add_header("Cookie", "rond_session=" + cookie)
+        if data is not None:
+            req.add_header("Content-Type", "application/x-www-form-urlencoded")
+        try:
+            with _opener.open(req, timeout=timeout) as r:
+                return r.status, dict(r.headers), r.read()
+        except urllib.error.HTTPError as e:
+            return e.code, dict(e.headers), e.read()
+        except Exception as e:            # 超时 / 连接被掐
+            last = e
+            if i + 1 < tries:
+                time.sleep(1.5 * (i + 1))
+    raise RuntimeError("请求失败 %s：%s" % (url, last))
+
+
+def field(block, name):
+    m = re.search(r'name="%s"[^>]*?value="([^"]*)"' % re.escape(name), block)
+    return htmllib.unescape(m.group(1)) if m else ""
+
+
+def parse_places(page_html):
+    out = []
+    for block in FORM_RE.findall(page_html):
+        src = field(block, "src_pk")
+        if not src.isdigit():
+            continue
+        cat = ""
+        sel = re.search(r'<select name="category".*?</select>', block, re.S)
+        if sel:
+            cur = re.search(r'<option value="([^"]*)" selected>', sel.group(0))
+            cat = htmllib.unescape(cur.group(1)) if cur else ""
+        p = {
+            "src_pk": int(src),
+            "name": field(block, "name"),
+            "category": cat,
+            "lat": field(block, "lat"),
+            "lon": field(block, "lon"),
+            "province": field(block, "province"),
+            "city": field(block, "city"),
+            "district": field(block, "district"),
+        }
+        mo = ORIG_RE.search(block)
+        p["orig"] = (mo.group(1) + "," + mo.group(2)) if mo else ""
+        out.append(p)
+    return out
+
+
+def fetch_places(site, cookie, verbose=True):
+    """翻页读完整个地点列表（含被区域黑名单/迷雾遮罩挡住、公开接口看不到的点）。"""
+    places, seen, page = [], set(), 1
+    while page <= 200:
+        status, _, body = http("%s/admin/edit?page=%d" % (site, page), cookie)
+        if status != 200:
+            raise SystemExit("读 /admin/edit?page=%d 失败：HTTP %d（会话过期？）" % (page, status))
+        chunk = parse_places(body.decode("utf-8", "replace"))
+        chunk = [p for p in chunk if p["src_pk"] not in seen]
+        if not chunk:
+            break
+        for p in chunk:
+            seen.add(p["src_pk"])
+        places.extend(chunk)
+        if verbose:
+            sys.stderr.write("\r读取地点列表… 第 %d 页，累计 %d 个" % (page, len(places)))
+            sys.stderr.flush()
+        page += 1
+    if verbose:
+        sys.stderr.write("\n")
+    return places
+
+
+def wgs2gcj(lat, lon):
+    a, ee = 6378245.0, 0.00669342162296594323
+
+    def tl(x, y):
+        r = -100.0 + 2.0 * x + 3.0 * y + 0.2 * y * y + 0.1 * x * y + 0.2 * math.sqrt(abs(x))
+        r += (20.0 * math.sin(6 * x * math.pi) + 20.0 * math.sin(2 * x * math.pi)) * 2 / 3
+        r += (20.0 * math.sin(y * math.pi) + 40.0 * math.sin(y / 3 * math.pi)) * 2 / 3
+        r += (160.0 * math.sin(y / 12 * math.pi) + 320 * math.sin(y * math.pi / 30)) * 2 / 3
+        return r
+
+    def tg(x, y):
+        r = 300.0 + x + 2.0 * y + 0.1 * x * x + 0.1 * x * y + 0.1 * math.sqrt(abs(x))
+        r += (20.0 * math.sin(6 * x * math.pi) + 20.0 * math.sin(2 * x * math.pi)) * 2 / 3
+        r += (20.0 * math.sin(x * math.pi) + 40.0 * math.sin(x / 3 * math.pi)) * 2 / 3
+        r += (150.0 * math.sin(x / 12 * math.pi) + 300.0 * math.sin(x / 30 * math.pi)) * 2 / 3
+        return r
+
+    dl, dn = tl(lon - 105.0, lat - 35.0), tg(lon - 105.0, lat - 35.0)
+    rl = lat / 180 * math.pi
+    m = math.sin(rl)
+    m = 1 - ee * m * m
+    sm = math.sqrt(m)
+    return (round(lat + (dl * 180) / ((a * (1 - ee)) / (m * sm) * math.pi), 6),
+            round(lon + (dn * 180) / (a / sm * math.cos(rl) * math.pi), 6))
+
+
+def haversine(a, b, c, d):
+    R = 6371.0088
+    p1, p2 = math.radians(a), math.radians(c)
+    dp, dl = math.radians(c - a), math.radians(d - b)
+    x = math.sin(dp / 2) ** 2 + math.cos(p1) * math.cos(p2) * math.sin(dl / 2) ** 2
+    return 2 * R * math.asin(math.sqrt(x)) * 1000
+
+
+def read_exif(dirs):
+    """读出照片目录里所有带 GPS 的坐标（四舍五入到 6 位，与导入页 toFixed(6) 对齐）。"""
+    try:
+        from PIL import Image
+    except ImportError:
+        raise SystemExit("需要 Pillow：pip install pillow")
+    try:
+        import pillow_heif
+        pillow_heif.register_heif_opener()
+    except ImportError:
+        pass
+    pts, skipped = [], 0
+    for root_dir in dirs:
+        for root, _, files in os.walk(root_dir):
+            for fn in sorted(files):
+                if not fn.lower().endswith((".jpg", ".jpeg", ".heic", ".heif", ".png")):
+                    continue
+                path = os.path.join(root, fn)
+                try:
+                    ex = Image.open(path).getexif()
+                    g = ex.get_ifd(34853)
+                    la, lo = g.get(2), g.get(4)
+                    if not la or not lo:
+                        skipped += 1
+                        continue
+                    lat = float(la[0]) + float(la[1]) / 60 + float(la[2]) / 3600
+                    lon = float(lo[0]) + float(lo[1]) / 60 + float(lo[2]) / 3600
+                    if str(g.get(1)).upper().startswith("S"):
+                        lat = -lat
+                    if str(g.get(3)).upper().startswith("W"):
+                        lon = -lon
+                    pts.append({"file": fn, "lat": round(lat, 6), "lon": round(lon, 6),
+                                "taken": (ex.get(36867) or ex.get(306) or "")})
+                except Exception:
+                    skipped += 1
+    return pts, skipped
+
+
+def classify(place, photos):
+    try:
+        lat, lon = round(float(place["lat"]), 6), round(float(place["lon"]), 6)
+    except ValueError:
+        return "invalid", None
+    for ph in photos:
+        if ph["lat"] == lat and ph["lon"] == lon:
+            return "raw", ph
+    for ph in photos:
+        if wgs2gcj(ph["lat"], ph["lon"]) == (lat, lon):
+            return "gcj", ph
+    return "unknown", None
+
+
+def api_coords(site, cookie, lat, lon):
+    q = urllib.parse.urlencode({"mode": "wgs2gcj", "lat": lat, "lon": lon})
+    status, _, body = http("%s/api/coords?%s" % (site, q), cookie)
+    if status != 200:
+        raise RuntimeError("换算接口 HTTP %d" % status)
+    d = json.loads(body.decode("utf-8"))
+    if d.get("error"):
+        raise RuntimeError(d["error"])
+    return d
+
+
+def save_place(site, cookie, p, lat, lon):
+    form = urllib.parse.urlencode({
+        "src_pk": p["src_pk"], "name": p["name"], "category": p["category"],
+        "lat": "%.6f" % lat, "lon": "%.6f" % lon,
+        "province": p["province"], "city": p["city"], "district": p["district"],
+    }).encode("utf-8")
+    status, headers, _ = http("%s/admin/edit/place/update" % site, cookie, data=form, timeout=60)
+    loc = headers.get("Location", "")
+    if status not in (302, 303):
+        return False, "HTTP %d" % status
+    if "err=" in loc:
+        return False, urllib.parse.unquote(loc.split("err=")[-1])
+    return True, urllib.parse.unquote(loc.split("ok=")[-1]) if "ok=" in loc else "ok"
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--site", required=True, help="站点根地址，如 https://life.shiroko.one")
+    ap.add_argument("--cookie", default=os.environ.get("ROND_COOKIE", ""),
+                    help="rond_session 的值（也可用环境变量 ROND_COOKIE）")
+    ap.add_argument("--photos", action="append", default=[], help="原图目录，可重复")
+    ap.add_argument("--baseline-max-sp", type=int, default=323,
+                    help="rond 基线里 ZLOCATION 的 Z_PK 上限，大于它的都是站点新建")
+    ap.add_argument("--out", default="photo-coord-fix.csv")
+    ap.add_argument("--apply", action="store_true", help="真的写库（不加则只出清单）")
+    ap.add_argument("--include-unverified", action="store_true",
+                    help="连没有原图可核对的点一起换算（风险自负）")
+    ap.add_argument("--only", default="", help="只处理这些 src_pk，逗号分隔")
+    ap.add_argument("--sleep", type=float, default=0.15, help="每次写库之间的间隔秒数")
+    args = ap.parse_args()
+
+    site = args.site.rstrip("/")
+    if not args.cookie:
+        raise SystemExit("缺少 --cookie（浏览器登录后复制 rond_session 的值）")
+
+    places = fetch_places(site, args.cookie)
+    created = [p for p in places if p["src_pk"] > args.baseline_max_sp]
+    print("地点总数 %d，其中站点新建（sp > %d）%d 个"
+          % (len(places), args.baseline_max_sp, len(created)))
+
+    photos = []
+    if args.photos:
+        photos, skipped = read_exif(args.photos)
+        print("原图：读到 %d 张带 GPS 的照片（%d 张无 GPS/读不了）" % (len(photos), skipped))
+
+    rows = []
+    for p in created:
+        verdict, ph = classify(p, photos)
+        new = wgs2gcj(round(float(p["lat"]), 6), round(float(p["lon"]), 6))
+        moved = haversine(round(float(p["lat"]), 6), round(float(p["lon"]), 6), new[0], new[1])
+        rows.append({**p, "verdict": verdict, "photo": ph["file"] if ph else "",
+                     "new_lat": new[0], "new_lon": new[1], "moved": moved})
+
+    todo = [r for r in rows if r["verdict"] == "raw"]
+    if args.include_unverified:
+        todo += [r for r in rows if r["verdict"] == "unknown"]
+    if args.only:
+        want = {int(x) for x in args.only.replace(" ", "").split(",") if x}
+        todo = [r for r in todo if r["src_pk"] in want]
+
+    n_raw = sum(1 for r in rows if r["verdict"] == "raw")
+    n_gcj = sum(1 for r in rows if r["verdict"] == "gcj")
+    n_unk = sum(1 for r in rows if r["verdict"] == "unknown")
+    print("判定：没转换 %d ｜ 已正确 %d ｜ 无原图可核 %d" % (n_raw, n_gcj, n_unk))
+
+    with io.open(args.out, "w", encoding="utf-8-sig", newline="") as f:
+        w = csv.writer(f)
+        w.writerow(["src_pk", "名称", "城市", "判定", "原图", "现在纬度", "现在经度",
+                    "换算后纬度", "换算后经度", "移动(米)"])
+        for r in sorted(rows, key=lambda x: x["src_pk"]):
+            w.writerow([r["src_pk"], r["name"], r["city"],
+                        {"raw": "没转换", "gcj": "已正确", "unknown": "无原图", "invalid": "坐标无效"}[r["verdict"]],
+                        r["photo"], r["lat"], r["lon"],
+                        "%.6f" % r["new_lat"], "%.6f" % r["new_lon"], "%.0f" % r["moved"]])
+    print("清单已写：%s" % args.out)
+    for r in todo[:10]:
+        print("  sp=%-5s %-20s %s,%s → %.6f,%.6f  (%.0fm)"
+              % (r["src_pk"], r["name"][:18], r["lat"], r["lon"], r["new_lat"], r["new_lon"], r["moved"]))
+    if len(todo) > 10:
+        print("  …共 %d 条" % len(todo))
+
+    if not args.apply:
+        print("\n[dry-run] 加 --apply 才会写库。")
+        return
+    if not todo:
+        print("\n没有需要处理的点。")
+        return
+
+    applied, failed = [], []
+    for i, r in enumerate(todo, 1):
+        try:
+            d = api_coords(site, args.cookie, r["lat"], r["lon"])
+            ok, msg = save_place(site, args.cookie, r, d["lat"], d["lon"])
+        except Exception as e:
+            ok, msg, d = False, str(e), None
+        if ok:
+            applied.append({**r, "applied_lat": "%.6f" % d["lat"], "applied_lon": "%.6f" % d["lon"]})
+            sys.stderr.write("\r已改 %d/%d  sp=%s %s        " % (i, len(todo), r["src_pk"], msg[:40]))
+        else:
+            failed.append({**r, "err": msg})
+            sys.stderr.write("\r失败 %d/%d  sp=%s %s        " % (i, len(todo), r["src_pk"], msg[:40]))
+        sys.stderr.flush()
+        time.sleep(args.sleep)
+    print()
+
+    log = args.out + ".applied.csv"
+    with io.open(log, "w", encoding="utf-8-sig", newline="") as f:
+        w = csv.writer(f)
+        w.writerow(["src_pk", "名称", "原纬度", "原经度", "新纬度", "新经度", "回退用经度", "回退用纬度"])
+        for r in applied:
+            w.writerow([r["src_pk"], r["name"], r["lat"], r["lon"],
+                        r["applied_lat"], r["applied_lon"], r["lat"], r["lon"]])
+    print("成功 %d 条，失败 %d 条；回退清单：%s" % (len(applied), len(failed), log))
+    for r in failed:
+        print("  失败 sp=%s %s：%s" % (r["src_pk"], r["name"][:16], r["err"]))
+
+
+if __name__ == "__main__":
+    main()

tools/verify_rondbackup.py

@@ -0,0 +1,116 @@
+import zipfile, sqlite3, os, shutil, sys, tempfile
+
+A = sys.argv[1] if len(sys.argv) > 1 else r"C:/OtherFiles/Github/rond-with-you/data/20260914_151649.rondbackup"
+B = sys.argv[2] if len(sys.argv) > 2 else r"C:/OtherFiles/Github/rond-with-you/storage/export-final.rondbackup"
+
+MANAGED = ["ZLOCATION", "ZVISIT", "ZMOVEMENT", "ZHOURLYWEATHER", "ZACTIVITY", "ZTAG", "ZTAGGROUP", "ZTRANSPORT"]
+UNMANAGED = ["ZRAWVISIT", "ZKEYWORD", "ZPOICATEGORY", "ZACTIVITYGROUP"]
+
+tmp = tempfile.mkdtemp(prefix="rondcmp")
+
+
+def extract(path, tag):
+    d = os.path.join(tmp, tag)
+    os.makedirs(d, exist_ok=True)
+    with zipfile.ZipFile(path) as z:
+        z.extractall(d)
+    for f in os.listdir(d):
+        if f.endswith(".sqlite"):
+            return os.path.join(d, f)
+    for root, dirs, files in os.walk(d):
+        for f in files:
+            if f.endswith(".sqlite"):
+                return os.path.join(root, f)
+    raise SystemExit("no sqlite in " + path)
+
+
+sa = extract(A, "orig")
+sb = extract(B, "gen")
+
+ca = sqlite3.connect(sa)
+cb = sqlite3.connect(sb)
+
+diff_total = 0
+
+
+def rows(c, table):
+    cur = c.execute('SELECT * FROM "%s"' % table)
+    cols = [d[0] for d in cur.description]
+    out = {}
+    for r in cur.fetchall():
+        out[r[0]] = dict(zip(cols, r))
+    return out
+
+
+def norm(v):
+    if isinstance(v, float):
+        return round(v, 6)
+    if isinstance(v, bytes):
+        return v
+    return v
+
+
+def cmp(table):
+    global diff_total
+    ra = rows(ca, table)
+    rb = rows(cb, table)
+    ca_ = sqlite3.connect(sa).execute('PRAGMA table_info("%s")' % table).fetchall()
+    cols = [x[1] for x in ca_]
+    miss = [k for k in ra if k not in rb]
+    extra = [k for k in rb if k not in ra]
+    vdiff = []
+    for k in ra:
+        if k not in rb:
+            continue
+        d = []
+        for c in cols:
+            va, vb = norm(ra[k].get(c)), norm(rb[k].get(c))
+            if isinstance(va, bytes) or isinstance(vb, bytes):
+                if va != vb:
+                    d.append((c, type(va).__name__, type(vb).__name__))
+                continue
+            if va != vb:
+                d.append((c, va, vb))
+        if d:
+            vdiff.append((k, d))
+    n = len(miss) + len(extra) + len(vdiff)
+    diff_total += n
+    print("[%s] rows orig=%d gen=%d  missing=%d extra=%d valuediff=%d" % (table, len(ra), len(rb), len(miss), len(extra), len(vdiff)))
+    if miss:
+        print("   missing PKs:", miss[:10])
+    if extra:
+        print("   extra PKs:", extra[:10])
+    for k, d in vdiff[:3]:
+        print("   PK %s differs: %s" % (k, d[:8]))
+
+
+print("=== MANAGED ===")
+for t in MANAGED:
+    cmp(t)
+print("=== UNMANAGED ===")
+for t in UNMANAGED:
+    cmp(t)
+
+# 结构对比
+def struct(c):
+    t = sorted(x[0] for x in c.execute("SELECT name FROM sqlite_master WHERE type='table'").fetchall())
+    i = sorted(x[0] for x in c.execute("SELECT name FROM sqlite_master WHERE type='index'").fetchall())
+    return t, i
+
+ta, ia = struct(ca)
+tb, ib = struct(cb)
+print("tables: %d vs %d  equal=%s" % (len(ta), len(tb), ta == tb))
+if ta != tb:
+    print("  only-orig:", [x for x in ta if x not in tb])
+    print("  only-gen :", [x for x in tb if x not in ta])
+print("indexes: %d vs %d  equal=%s" % (len(ia), len(ib), ia == ib))
+if ia != ib:
+    print("  only-orig:", [x for x in ia if x not in ib])
+    print("  only-gen :", [x for x in ib if x not in ia])
+
+ma = ca.execute("SELECT Z_PLIST FROM Z_METADATA").fetchone()[0]
+mb = cb.execute("SELECT Z_PLIST FROM Z_METADATA").fetchone()[0]
+print("Z_METADATA plist: %d vs %d bytes equal=%s" % (len(ma), len(mb), ma == mb))
+
+print("总差异 =", diff_total)
+shutil.rmtree(tmp, ignore_errors=True)