"""虚构的大库:3,000 张表(next-round 03 的大库验收夹具)。名字、注释、需求一律虚构。 - 2,990 张「同一种写法」的表:20 个业务域 × 15 种表 × 10 个分区(华东、华南……),表名 {域}_{表}_{分区},注释 「{域中文}{表中文}({分区中文})」。它们彼此只差几个字,正好是挑表时最容易混淆的那一类。 - 10 张目标表(TARGETS):名字和注释各不相同,每张配一句脚本化的需求。前 5 张给数据目录(中文名),有目录的表一律在探查 范围里(pinned);后 5 张在探查范围外,只有对象索引里的注释——SQLite 没有表注释,测试里直接写进对象索引的 comment, 模拟 MySQL 的 TABLE_COMMENT、PostgreSQL 的 obj_description。 每张表 5 列(id 主键、业务日期、金额、状态、所属域的编号),没有数据:大库的验收看的是对象索引、范围、挑表,不查数。 """ from __future__ import annotations import sqlite3 from dataclasses import dataclass from pathlib import Path DOMAINS = [ ("ord", "订单"), ("inv", "库存"), ("pur", "采购"), ("sal", "销售"), ("fin", "财务"), ("hrm", "人事"), ("crm", "客户"), ("lgs", "物流"), ("prd", "商品"), ("mkt", "营销"), ("svc", "客服"), ("whs", "仓储"), ("pay", "支付"), ("rtn", "退货"), ("sto", "门店"), ("emp", "员工"), ("sup", "供应商"), ("prm", "促销"), ("mbr", "会员"), ("aud", "审计"), ] ENTITIES = [ ("hdr", "表头"), ("line", "明细"), ("log", "日志"), ("snap", "快照"), ("daily", "日汇总"), ("monthly", "月汇总"), ("cfg", "配置"), ("map", "映射"), ("hist", "历史"), ("tmp", "临时表"), ("bak", "备份"), ("stat", "统计"), ("dim", "维度"), ("fact", "事实"), ("queue", "队列"), ] REGIONS = [ ("east", "华东"), ("south", "华南"), ("north", "华北"), ("west", "西部"), ("central", "华中"), ("ne", "东北"), ("arch", "历史归档"), ("test", "测试"), ("y2019", "2019年"), ("y2020", "2020年"), ] @dataclass(frozen=True) class Target: name: str comment: str need: str #: 有数据目录(中文名)、一律在探查范围里 cataloged: bool TARGETS = [ Target("member_points_ledger", "会员积分流水", "会员积分每个月发放了多少、消耗了多少", True), Target("store_traffic_counter", "门店每日客流计数", "各门店上周每天的客流量", True), Target("refund_reason_dict", "退货原因字典", "按退货原因统计退货单数", True), Target("coupon_redemption", "优惠券核销记录", "优惠券的核销率", True), Target("supplier_delivery_sla", "供应商到货时效", "供应商的到货准时率", True), Target("warehouse_turnover_snap", "仓库库存周转快照", "各仓库的库存周转天数", False), Target("shift_roster", "员工排班表", "下周员工的排班情况", False), Target("complaint_ticket", "客户投诉工单", "客户投诉工单的平均处理时长", False), Target("online_payment_flow", "线上支付流水", "线上支付的成功率", False), Target("waybill_track", "物流运单签收轨迹", "物流运单从发货到签收用了多久", False), ] TOTAL = len(DOMAINS) * len(ENTITIES) * len(REGIONS) - len(TARGETS) + len(TARGETS) def generic_names() -> list[tuple[str, str]]: """2,990 张同一种写法的表 (表名, 注释):按域、表、分区的顺序,最后 10 个组合让给目标表。""" out: list[tuple[str, str]] = [] for code, cn in DOMAINS: for ent, ent_cn in ENTITIES: for reg, reg_cn in REGIONS: out.append((f"{code}_{ent}_{reg}", f"{cn}{ent_cn}({reg_cn})")) return out[: len(out) - len(TARGETS)] def comments() -> dict[str, str]: """全部 3,000 张表的注释(写进对象索引,模拟有表注释的数据库)。""" out = dict(generic_names()) out.update({t.name: t.comment for t in TARGETS}) return out def build(path: Path | str) -> Path: path = Path(path) db = sqlite3.connect(path) for name, _comment in generic_names(): code = name.split("_", 1)[0] db.execute(f"CREATE TABLE {name} (id INTEGER PRIMARY KEY, biz_date TEXT, amount REAL, status TEXT, " f"{code}_id INTEGER)") for t in TARGETS: db.execute(f"CREATE TABLE {t.name} (id INTEGER PRIMARY KEY, biz_date TEXT, amount REAL, status TEXT, " "ref_id INTEGER)") db.commit() db.close() return path __all__ = ["DOMAINS", "ENTITIES", "REGIONS", "TARGETS", "TOTAL", "Target", "build", "comments", "generic_names"]