auto_control/core/ledger.py

"""账号台账：一台设备上登录着哪些账号。

**它解决什么**（用户场景）：一台手机登好几个抖音号、十几台手机四五十个号，
账号信息散在一张电子表格里 —— 换 IP、加号、想查"这台登的是哪几个号"都得翻表。
台账落进平台后有三个月牙：

  1. web「账号」页 —— 列表 / 增删改 / **从表格粘贴导入**（`web/devices_api.py` 的 `/api/ledger*`）
  2. 任务的「条件判断」取号 —— `if_el` 的 `cmp_source=device|all|group`
     （比对运算符用「包含」时，纯号 `35377983067` 能匹配元素原文 `抖音号：35377983067`）
  3. 手机端 Agent 的「身份大字页」 —— 平台推 `accounts_b64`（`doc/DEVICE_AGENT.md` §5.2）

⚠ **台账里的抖音号是纯号，绝不能拿去当去重身份**：
`done_mark.identity` 存的是**元素原文**、逐字算 key，格式不一致（`35377983067`
vs `抖音号：35377983067`）会让去重**静默失效**。取号只用于条件判断的比对值。

DB 访问自建 app context（照 `core/dedup.py`），调用方（任务线程 / Web 线程）不用关心。
"""
import re
import time
import uuid

from core.logger import get_logger
from core.models import db, Device, DeviceAccount, DeviceGroup

_log = get_logger("core.ledger")

# 可编辑字段（接口 / 导入 / 前端共用一份清单，避免某处漏字段）
FIELDS = ("device_name", "phone", "nickname", "douyin_id", "registered_at",
          "sim_in_device", "can_post_video", "bio", "note")
# 这两列是布尔（"空=否"）
BOOL_FIELDS = ("sim_in_device", "can_post_video")

# 粘贴导入：表头列名 → 字段（允许乱序/缺列/多列；认不出的列忽略）
HEADER_ALIASES = {
    "device_name": ("设备号", "设备", "设备名", "机器", "机型"),
    "phone": ("手机号", "手机", "号码", "手机号码"),
    "nickname": ("账号名称", "账号名", "昵称", "账号", "名称"),
    "douyin_id": ("抖音号", "抖音", "抖音id", "douyin"),
    "registered_at": ("注册时间", "注册日期", "注册"),
    "sim_in_device": ("卡在机内", "卡在机", "卡在", "机内"),
    "can_post_video": ("可发视频", "可发", "发视频"),
    "bio": ("简介", "签名", "个人简介"),
    "note": ("备注", "说明"),
}
# 没有表头时按这个顺序解析（= 用户表格的列顺序，界面上要写明）
DEFAULT_ORDER = FIELDS

_TRUE_WORDS = ("是", "有", "对", "在", "✔", "√", "✓", "1", "true", "y", "yes", "t")
_FALSE_WORDS = ("否", "无", "没有", "不在", "不", "0", "false", "n", "no", "f", "")
# 抖音号形态：纯数字。不拦（少数号可能带字母），只在导入结果里留痕 ——
# 候选值一旦是错的，任务会"正常跑完但永远不命中"，最难查
_DOUYIN_RE = re.compile(r"^\d{6,20}$")

_app = None


def init_app(app):
    """web_server 启动时调用：绑 app（后台线程/任务线程访问 db 要自推 context）。"""
    global _app
    _app = app


def _ctx():
    if _app is None:
        raise RuntimeError("ledger 未关联 Flask app（web_server 启动时调用 init_app）")
    return _app.app_context()


def _now():
    return time.strftime("%Y-%m-%d %H:%M:%S")


def _norm_bool(v):
    """是/否 → (True/False, 是否认得)。认不出的按"否"处理，但**留痕**。"""
    if isinstance(v, bool):
        return v, True
    s = str(v or "").strip().lower()
    if s in _TRUE_WORDS:
        return True, True
    if s in _FALSE_WORDS:
        return False, True
    return False, False


def _clean(fields):
    """把接口来的字段清洗成可落库的值。只处理传进来的键。返回 (values, error)。"""
    out = {}
    for k in FIELDS:
        if k not in (fields or {}):
            continue
        v = fields.get(k)
        out[k] = _norm_bool(v)[0] if k in BOOL_FIELDS else str(v or "").strip()
    if "device_name" in out and not out["device_name"]:
        return {}, "设备号不能为空"
    if "douyin_id" in out and not out["douyin_id"]:
        return {}, "抖音号不能为空"
    return out, ""


# ================== 查 ==================
def get_by_id(acc_id):
    """取一条（dict）或 None。

    **跨边界一律返回 dict**：commit 之后 ORM 行的属性会过期，出了 app context
    再读属性会抛 DetachedInstanceError（接口层 `row.to_dict()` 必炸）。
    """
    with _ctx():
        row = DeviceAccount.query.get(acc_id)
        return row.to_dict() if row else None


def find_by_douyin(douyin_id, exclude_id=""):
    """按抖音号找已存在的那条（唯一性判据；空号不参与）。返回 (id, 设备号) 或 None。

    只返回两个纯值，同样是为了不让 ORM 行溜出 context。
    """
    d = str(douyin_id or "").strip()
    if not d:
        return None
    with _ctx():
        q = DeviceAccount.query.filter(DeviceAccount.douyin_id == d)
        if exclude_id:
            q = q.filter(DeviceAccount.id != exclude_id)
        row = q.first()
        return (row.id, row.device_name or "") if row else None


def list_accounts(device_name="", q="", limit=500):
    """列表 + 统计（「账号」页用）。返回 (rows, stats)。"""
    with _ctx():
        query = DeviceAccount.query
        if device_name:
            query = query.filter(DeviceAccount.device_name == device_name)
        if q:
            like = f"%{q.strip()}%"
            query = query.filter(db.or_(DeviceAccount.douyin_id.like(like),
                                        DeviceAccount.phone.like(like),
                                        DeviceAccount.nickname.like(like)))
        rows = (query.order_by(DeviceAccount.device_name, DeviceAccount.id)
                .limit(max(1, min(int(limit or 500), 2000))).all())
        all_rows = DeviceAccount.query.all()
        stats = {
            "total": len(all_rows),
            "devices": len({r.device_name for r in all_rows if r.device_name}),
            "can_post_video": len([r for r in all_rows if r.can_post_video]),
            "shown": len(rows),
        }
        return [r.to_dict() for r in rows], stats


def device_names():
    """台账里出现过的设备号（筛选下拉用）。"""
    try:
        with _ctx():
            return sorted({r.device_name for r in DeviceAccount.query.all() if r.device_name})
    except Exception:
        return []


def for_device(serial="", device_name=""):
    """这台设备的账号。serial 优先，再按设备名（换 IP / 改名都找得到）。

    `serial` 是**录入时的快照**：设备换 IP 后台账里那条还是旧地址，
    所以两边都查一遍并按抖音号去重。
    """
    with _ctx():
        rows, seen = [], set()
        if serial:
            for r in DeviceAccount.query.filter(DeviceAccount.serial == serial).all():
                rows.append(r)
                seen.add(r.douyin_id)
        name = device_name
        if not name and serial:
            dev = Device.query.get(serial)
            name = (dev.name if dev else "") or ""
        if name:
            for r in DeviceAccount.query.filter(DeviceAccount.device_name == name).all():
                if r.douyin_id not in seen:
                    rows.append(r)
                    seen.add(r.douyin_id)
        return [r.to_dict() for r in rows]


def resolve_device(serial="", device_name=""):
    """serial/设备名 → (设备号, 账号行, 说明)。

    "本机是哪台设备"这条链只在这一个地方定义，任务侧和设备端都走它 ——
    两套规则必然分叉（那会变成"任务取到 3 个号、页面显示 0 个号"这种鬼故事）。
    """
    name = (device_name or "").strip()
    if not name and serial:
        with _ctx():
            dev = Device.query.get(serial)
            name = (dev.name if dev else "") or ""
    rows = for_device(serial=serial, device_name=name)
    # 兜底：台账里有人把"设备号"直接填成地址（手填习惯），按 serial 再认一次
    if not rows and serial:
        with _ctx():
            hit = DeviceAccount.query.filter(DeviceAccount.device_name == serial).all()
            if hit:
                name = serial
                rows = [r.to_dict() for r in hit]
    if not rows:
        note = (f"设备名『{name}』在台账里没有账号" if name
                else f"设备 {serial or '（无地址）'} 认不出设备名（设备池里没登记或还没刷新）")
        if serial and name:
            note += f"，地址 {serial} 也没对上台账"
        return name, [], note
    return name, rows, f"设备『{name}』登记 {len(rows)} 个号"


def serial_of(device_name):
    """设备名 → **当前**地址（拿不到返回空串）。

    ⚠ 为什么不能直接用台账/计划行里的 `serial` 快照：设备地址会变（换 IP、重连）。
    快照是"录入时"的值，久了就是错的 —— 要下发操作（发布视频等）时必须现查。
    优先设备池的**内存快照**（零 DB、已经是刷新过的），再退回 DB。
    """
    name = (device_name or "").strip()
    if not name:
        return ""
    try:
        from core import device_pool
        for d in device_pool.list_devices():
            if (d.get("name") or "") == name:
                return d.get("serial") or ""
    except Exception:
        pass
    try:
        with _ctx():
            dev = Device.query.filter(Device.name == name).first()
            return (dev.serial if dev else "") or ""
    except Exception:
        return ""


def douyin_ids(scope, serial="", device_name="", group=""):
    """按范围取抖音号（任务「条件判断」取号用）。返回 (去重后的纯号列表, 说明)。

    说明是给日志/步骤明细看的**人话解释** —— 取不到号时它是唯一线索
    （取不到候选值的后果是"这一步永远不命中"，但任务本身会正常跑完）。
    scope: `device`=本机台账 / `all`=全部 / `group`=某设备分组 / 其他=不取
    """
    scope = (scope or "").strip()
    if scope in ("", "manual", "none"):
        return [], ""
    try:
        with _ctx():
            if scope == "device":
                _, rows, note = resolve_device(serial=serial, device_name=device_name)
                return _uniq([r.get("douyin_id") for r in rows]), note
            if scope == "all":
                ids = _uniq([r.douyin_id for r in DeviceAccount.query.all()])
                return ids, f"全部台账共 {len(ids)} 个号"
            if scope == "group":
                names = _device_names_of_group(group)
                if not names:
                    return [], f"设备分组『{group}』没有设备（或分组不存在）"
                rows = DeviceAccount.query.filter(
                    DeviceAccount.device_name.in_(list(names))).all()
                ids = _uniq([r.douyin_id for r in rows])
                return ids, f"设备分组『{group}』（{len(names)} 台设备）共 {len(ids)} 个号"
            return [], f"未知的取号范围 {scope!r}"
    except Exception as e:
        # 取号失败绝不拦住任务：返回空，由调用方回落到手填值
        _log.warning(f"台账取号失败（回落手填值）: {e}")
        return [], f"台账取号失败: {e}"


def _uniq(ids):
    out, seen = [], set()
    for i in ids or []:
        s = str(i or "").strip()
        if s and s not in seen:
            seen.add(s)
            out.append(s)
    return out


def _device_names_of_group(group_name):
    """设备分组 → 设备名集合（分组里存的是 serial，这里换成台账里的设备号）。"""
    g = DeviceGroup.query.filter(DeviceGroup.name == (group_name or "").strip()).first()
    if not g:
        return set()
    serials = g.get_serials() or []
    if not serials:
        return set()
    devs = Device.query.filter(Device.serial.in_(list(serials))).all()
    return {d.name for d in devs if d.name}


def counts_by_device():
    """{设备名: 号数}（设备池显示"这台几个号"用，一次拿全避免 N 次请求）。"""
    try:
        with _ctx():
            out = {}
            for r in DeviceAccount.query.all():
                if r.device_name:
                    out[r.device_name] = out.get(r.device_name, 0) + 1
            return out
    except Exception as e:
        _log.warning(f"台账统计失败: {e}")
        return {}


def by_device():
    """{设备名: {count, accounts[]}}（设备维度查看弹窗用）。"""
    groups = {}
    try:
        with _ctx():
            for r in DeviceAccount.query.order_by(DeviceAccount.douyin_id).all():
                g = groups.setdefault(r.device_name or "", {"count": 0, "accounts": []})
                g["count"] += 1
                g["accounts"].append(r.to_dict())
    except Exception as e:
        _log.warning(f"台账按设备分组失败: {e}")
    return groups


def agent_accounts(serial="", device_name="", limit=50, include_phone=False):
    """设备端 Agent 的账号块（身份大字页用）。

    · **默认不带手机号**：大字页是放在机器旁的公开屏幕，手机号不该默认上屏
    · 超量**按整条丢**（`truncated=True`），绝不按字节切 —— 切了就是坏 JSON，
      设备端只能整块丢掉
    · 任何异常都返回空块（不是抛）：台账坏了不能把"显示身份"这个动作弄坏
    """
    try:
        name, rows, note = resolve_device(serial=serial, device_name=device_name)
        out = []
        for r in rows[:max(1, int(limit or 50))]:
            item = {"name": r.get("nickname") or "", "douyin_id": r.get("douyin_id") or "",
                    "device_name": r.get("device_name") or "", "device_no": name,
                    "sim_in_device": bool(r.get("sim_in_device")),
                    "can_post_video": bool(r.get("can_post_video"))}
            if include_phone:
                item["phone"] = r.get("phone") or ""
            out.append(item)
        return {"v": 1, "total": len(rows), "shown": len(out),
                "truncated": len(out) < len(rows), "accounts": out, "note": note}
    except Exception as e:
        _log.warning(f"设备端账号块生成失败: {e}")
        return {"v": 1, "total": 0, "shown": 0, "truncated": False, "accounts": [],
                "note": f"读取失败: {e}"}


# ================== 增 / 改 / 删 ==================
def add(fields):
    """新增一条。返回 (dict, error)。"""
    values, err = _clean(fields)
    if err:
        return None, err
    if not values.get("device_name"):
        return None, "设备号不能为空"
    if not values.get("douyin_id"):
        return None, "抖音号不能为空"
    # try 必须在 `with _ctx()` 里面：否则 rollback 自己会在 context 外抛
    # RuntimeError，把真正的错误盖掉（这条踩过一次，别挪出去）
    with _ctx():
        try:
            if find_by_douyin(values["douyin_id"]):
                return None, f"抖音号 {values['douyin_id']} 已在台账里（一个号只能有一条）"
            row = DeviceAccount(id=uuid.uuid4().hex[:8], created_at=_now(), updated_at=_now())
            for k in FIELDS:
                # 没传的布尔列必须是 False，不能是 ""（Boolean 列收到空串会直接报
                # TypeError: Not a boolean value，整条插不进去）
                setattr(row, k, values.get(k, False if k in BOOL_FIELDS else ""))
            _fill_serial(row)
            db.session.add(row)
            db.session.commit()
            _log.info(f"账号台账新增：{row.device_name} / {row.nickname} / {row.douyin_id}")
            return row.to_dict(), ""       # 出 context 前转成纯值
        except Exception as e:
            db.session.rollback()
            return None, f"保存失败: {e}"


def update(acc_id, fields):
    """改一条（只改传进来的键）。返回 (dict, error)。"""
    with _ctx():
        try:
            row = DeviceAccount.query.get(acc_id)
            if not row:
                return None, "记录不存在"
            values, err = _clean(fields)
            if err:
                return None, err
            new_id = values.get("douyin_id", row.douyin_id)
            dup = find_by_douyin(new_id, exclude_id=acc_id)
            if dup:
                return None, f"抖音号 {new_id} 已被 {dup[1]} 占用（一个号只能有一条）"
            for k, v in values.items():
                setattr(row, k, v)
            row.updated_at = _now()
            _fill_serial(row)
            db.session.commit()
            return row.to_dict(), ""
        except Exception as e:
            db.session.rollback()
            return None, f"保存失败: {e}"


def delete(acc_id):
    """删一条（删掉之后任务就取不到这个号了）。"""
    with _ctx():
        try:
            row = DeviceAccount.query.get(acc_id)
            if not row:
                return False
            db.session.delete(row)
            db.session.commit()
            _log.info(f"账号台账删除：{row.device_name} / {row.nickname} / {row.douyin_id}")
            return True
        except Exception as e:
            db.session.rollback()
            _log.warning(f"账号台账删除失败: {e}")
            return False


def _fill_serial(row):
    """按设备名补 serial 快照（台账只填设备号时，自动关联设备池那条）。"""
    if row.serial:
        return
    if not row.device_name:
        return
    dev = Device.query.filter(Device.name == row.device_name).first()
    if dev:
        row.serial = dev.serial


# ================== 粘贴导入 ==================
def parse_paste(text, delimiter="\t"):
    """把粘贴的表格文本解析成台账行。返回 (rows, errors)。

    rows 里每行是 `{"_line": 3, "device_name": "...", ..., "_warns": [...]}`
    （`_line` = 粘贴文本里的行号，界面把错误指回原行；`_warns` 是"能导但要注意"的提示）。
    errors 是 `[{"line": 3, "reason": "..."}]`（这些行会被跳过）。

    分隔符默认 Tab（Excel 直接粘贴就是 Tab）。**不做"多个空格当分隔"**——
    账号名里本来就有双空格（实测 `有牛奶面包  你吃吗？`），按空格切会切坏。
    带表头时按列名映射（顺序随意、缺列留空）；不带表头则按 DEFAULT_ORDER 的固定顺序。
    """
    if delimiter not in ("\t", "|", ","):
        delimiter = "\t"
    text = (text or "").replace("\r\n", "\n").replace("\r", "\n")
    lines = [(i + 1, ln) for i, ln in enumerate(text.split("\n")) if ln.strip()]
    if not lines:
        return [], [{"line": 0, "reason": "没有可解析的内容"}]

    first_cells = [c.strip() for c in lines[0][1].split(delimiter)]
    mapped = _map_header(first_cells)
    if mapped:
        body = lines[1:]
    else:
        mapped = {name: i for i, name in enumerate(DEFAULT_ORDER)}
        body = lines
    if not body:
        return [], [{"line": lines[0][0], "reason": "只有表头，没有数据行"}]

    rows, errors = [], []
    for lineno, line in body:
        cells = line.split(delimiter)
        row = {"_line": lineno, "_warns": []}
        for f in FIELDS:                       # 先把所有字段填上，缺列留空
            row[f] = ""
        for field, idx in mapped.items():
            if field in FIELDS:
                row[field] = cells[idx].strip() if idx < len(cells) else ""
        if not row["device_name"] and not row["douyin_id"]:
            errors.append({"line": lineno, "reason": "设备号与抖音号都是空的，已跳过"})
            continue
        if not row["device_name"]:
            errors.append({"line": lineno, "reason": "缺设备号，已跳过"})
            continue
        if not row["douyin_id"]:
            errors.append({"line": lineno, "reason": "缺抖音号，已跳过"})
            continue
        for f in BOOL_FIELDS:
            raw = row[f]
            val, known = _norm_bool(raw)
            row[f] = val
            if not known:
                row["_warns"].append(f"『{raw}』认不出，按「否」处理")
        if not _DOUYIN_RE.match(row["douyin_id"]):
            row["_warns"].append(f"抖音号『{row['douyin_id']}』不是纯数字 —— "
                                 f"任务比对时可能匹配不上")
        rows.append(row)
    return rows, errors


def _map_header(cells):
    """列名 → 列下标（认不出"设备号+抖音号"时返回 {}，按固定顺序解析）。"""
    mapping = {}
    for idx, cell in enumerate(cells):
        c = re.sub(r"\s+", "", cell or "")
        if not c:
            continue
        for field, aliases in HEADER_ALIASES.items():
            if field in mapping:
                continue
            if any(c == a or c.startswith(a) for a in aliases):
                mapping[field] = idx
                break
    if "device_name" in mapping and "douyin_id" in mapping:
        return mapping
    return {}


def import_rows(rows, mode="skip", dry_run=False):
    """把解析好的行落库（或只预览）。返回逐行结果 + 汇总。

    mode=`skip`（默认）：抖音号已在台账里 → 跳过（不动现有那条）
    mode=`overwrite`：抖音号已在台账里 → 用粘贴的这行**整行覆盖**它
    dry_run=True：只判重、不写库（界面上"先预览后导入"）
    """
    mode = mode if mode in ("skip", "overwrite") else "skip"
    counts = {"total": 0, "added": 0, "updated": 0, "skipped": 0, "failed": 0}
    out_rows = []
    for row in rows or []:
        line = row.get("_line", 0)
        fields = {k: v for k, v in row.items() if k not in ("_line", "_warns")}
        counts["total"] += 1
        existing = find_by_douyin(fields.get("douyin_id"))
        item = {"line": line, "device_name": fields.get("device_name", ""),
                "douyin_id": fields.get("douyin_id", ""),
                "nickname": fields.get("nickname", ""),
                "warns": list(row.get("_warns") or []), "error": ""}
        if existing:
            if mode == "overwrite":
                item["action"] = "update"
                if not dry_run:
                    _, err = update(existing[0], fields)
                    if err:
                        item["action"], item["error"] = "fail", err
                counts["updated" if item["action"] == "update" else "failed"] += 1
            else:
                item["action"] = "skip"
                item["error"] = f"已存在（{existing[1]}），跳过"
                counts["skipped"] += 1
        else:
            item["action"] = "add"
            if not dry_run:
                _, err = add(fields)
                if err:
                    item["action"], item["error"] = "fail", err
            counts["added" if item["action"] == "add" else "failed"] += 1
        out_rows.append(item)
    if not dry_run:
        _log.info(f"账号台账导入：新增 {counts['added']}、覆盖 {counts['updated']}、"
                  f"跳过 {counts['skipped']}、失败 {counts['failed']}")
    return {"counts": counts, "rows": out_rows, "mode": mode, "dry_run": bool(dry_run)}