unis-oa/backend/scripts/check_completed_dedup.py

219 lines
10 KiB
Python
Raw Permalink Blame History

This file contains ambiguous Unicode characters!

This file contains ambiguous Unicode characters that may be confused with others in your current locale. If your use case is intentional and legitimate, you can safely ignore this warning. Use the Escape button to highlight these characters.

#!/usr/bin/env python3
"""守「我的已办按单折叠」:一个人办过几次的流程,列表里永远只摆一条。
已办表一次表态一行,退回重办 / 撤回再办 / 终止前先退一回,都会在同一张单上留下
好几行。以前「我的已办」把这些行原样摆出来,同一条流程摆两条,看着就像系统发重了
单子。现在改成按单折叠(``app/service/completed.py``),这条脚本盯着折叠的六件事:
1. **一条流程一条**:列表里每张单只出现一次,摆的是最近那次表态;底下几次记在
``handle_count``,导出的 CSV 记在「处理次数」那一列;
2. **页眉那个数对得上**:``total`` 数的是单不是行,翻页不重不漏;
3. **同一秒办完两条也要稳定**:并列时按主键定一个,两次查询给同一个答案;
4. **筛选条件不打架**:关键词、流程、日期筛的是行,折叠之后组内条件与回捞行用的
是同一套,条数与内容都按当前筛选口径来;
5. **别人的行不漏进来**:只数本人身份名下的表态;
6. **各处口径一个样**:首页角标(``portal/home`` 的 counts)与工作台
(``statistics/workbench``)那个「我的已办」也数单,跟列表一致。
库里不留东西:全程在一个未提交的事务里跑,收尾 ``rollback``::
PYTHONPATH=. python3 scripts/check_completed_dedup.py
"""
from __future__ import annotations
import csv
import io
from datetime import datetime, timedelta
from sqlalchemy import delete, select
from app import models
from app.api.export import export_tasks
from app.api.portal import home_counts
from app.api.runtime import list_completed
from app.api.support import workbench
from app.db import SessionLocal
from app.deps import CurrentUser
STAMP = datetime.now().strftime("%m%d%H%M%S")
ME = f"__dedup_me_{STAMP}"
OTHER = f"__dedup_other_{STAMP}"
PROCESS_A = f"dedup_p1_{STAMP}"
PROCESS_B = f"dedup_p2_{STAMP}"
BASE = datetime(2026, 1, 1, 9, 0, 0)
FAILS: list[str] = []
PASSED = 0
def check(label: str, condition, detail: str = "") -> None:
global PASSED
if condition:
PASSED += 1
print(f"[通过] {label}")
else:
FAILS.append(label)
print(f"[FAIL] {label}" + (f" -> {detail}" if detail else ""))
def stamp_of(moment: datetime) -> str:
"""序列化后的时间写法,断言里省得每处都抄一遍格式。"""
return moment.strftime("%Y-%m-%d %H:%M:%S")
def page(db, user, **query) -> dict:
"""打一次「我的已办」,回 data 那一层。"""
return list_completed(
user=user,
db=db,
keyword=query.get("keyword", ""),
process=query.get("process", ""),
category=query.get("category", ""),
start_date=query.get("start_date", ""),
end_date=query.get("end_date", ""),
page=query.get("page", 1),
page_size=query.get("page_size", 20),
)["data"]
def shape(items: list[dict]) -> list[tuple[str, int, datetime]]:
"""把结果收成 ``(work, 处理次数, 代表行时间)``,断言只看这三样。"""
return [(item["work"], int(item["handle_count"]), item["completed_time"]) for item in items]
def seed(db) -> dict[str, datetime]:
"""造四张单:办过 3 次(含同一秒并列)、1 次、2 次(其中一次才撞关键词),以及别人的 1 次。"""
marks = {
"A": BASE + timedelta(hours=5), # 14:00 —— 这张单的最后一次表态
"B": BASE + timedelta(hours=4), # 13:00
"C": BASE + timedelta(hours=3), # 12:00
"C_FA": BASE + timedelta(hours=2), # 11:00 —— C 上唯一撞「法务」关键词的那次
"NOON": BASE + timedelta(hours=3), # 12:00 —— A / C 在这一刻并列
}
def add(work: str, minutes: int, activity: str, *, process: str = PROCESS_A, identity: str = ME,
latest: bool = False) -> None:
moment = BASE + timedelta(minutes=minutes)
db.add(
models.TaskCompleted(
task=f"task-{work}-{minutes}",
job=f"job-{work}-{STAMP}",
work=work,
title=f"折叠校验单-{work}",
process=process,
process_name=f"折叠校验流程-{process}",
serial=f"DEDUP{work}{STAMP}",
activity=f"act-{activity}",
activity_name=activity,
person=identity,
identity=identity,
identity_name="折叠校验账号",
start_time=moment - timedelta(minutes=1),
completed_time=moment,
latest=latest,
)
)
# A:部门主管退回 → 14:00 与 14:00 整并列(终止 / 补录)
add("A", 180, "部门主管")
add("A", 300, "发起人", latest=True)
add("A", 300, "归档确认")
# B:一次办完
add("B", 240, "部门主管")
# C:法务核验(关键词只撞这一条)+ 部门主管
add("C", 120, "法务核验", process=PROCESS_B)
add("C", 180, "部门主管", process=PROCESS_B)
# D:别人在同一个环节办过的单,一条都不该漏进本人列表
add("D", 200, "部门主管", identity=OTHER)
db.flush()
return marks
def main() -> int:
with SessionLocal() as db:
person = db.execute(select(models.User).limit(1)).scalars().first()
if person is None:
print("[FAIL] 库里没有任何 users 行,造不出 CurrentUser")
return 1
user = CurrentUser(person=person, ref=ME, is_admin=True)
try:
marks = seed(db)
data = page(db, user)
got = shape(data["items"])
check("一张单只摆一条,四张单里本人办过的只有三张",
[row[0] for row in got] == ["A", "B", "C"], str(got))
check("总数数的是单不是行", data["total"] == 3, f"total={data['total']}")
check("摆出来的是最近那次表态", got[0][2] == stamp_of(marks["A"]), str(got))
check("处理次数如实带出来", [row[1] for row in got] == [3, 1, 2], str(got))
check("同一秒并列的两次表态不撑出第二条",
len({item["work"] for item in data["items"]}) == len(data["items"]))
again = shape(page(db, user)["items"])
check("同样的查询跑两遍答案一模一样(并列行的顺序不是随机的)", again == got, str(again))
walked = []
for index in range(1, 4):
one = page(db, user, page=index, page_size=1)
if one["total"] != 3:
check(f"第 {index} 页的总数还是 3", False, f"total={one['total']}")
walked.extend(row[0] for row in shape(one["items"]))
check("一页一条翻到底:不重不漏", walked == ["A", "B", "C"], str(walked))
check("翻过界的页给空表", page(db, user, page=9, page_size=1)["items"] == [])
hit = shape(page(db, user, keyword="法务")["items"])
check("关键词只撞其中一次表态时,这条单仍只摆一条,摆的是撞上的那次",
hit == [("C", 1, stamp_of(marks["C_FA"]))], str(hit))
check("按当前筛选数条数", page(db, user, keyword="法务")["total"] == 1)
check("关键词撞不到就是空", page(db, user, keyword="查无此单")["items"] == [])
dated = shape(page(db, user, start_date="2026-01-01T12:00:00", end_date="2026-01-01T14:00:00")["items"])
# 窗口 [12:00, 14:00):A 的两次 14:00 被筛掉,只剩 12:00 那次;A 与 C 在 12:00 上并列,
# 排序按 work 倒序补全,所以 C 排在 A 前——这一条同时在验"顺序是确定的"。
check("日期窗口筛的是行:窗口外那几次表态不算进次数",
[(row[0], row[1]) for row in dated] == [("B", 1), ("C", 1), ("A", 1)], str(dated))
check("窗口外的表态不再冒充最近一次",
all(row[2] < stamp_of(marks["A"]) for row in dated)
and all(row[2] == stamp_of(marks["NOON"]) or row[2] == stamp_of(marks["B"]) for row in dated),
str(dated))
scoped = page(db, user, process=PROCESS_A)
check("按流程筛仍然按单折叠",
[row[0] for row in shape(scoped["items"])] == ["A", "B"] and scoped["total"] == 2,
str(shape(scoped["items"])))
body = export_tasks(
scope="completed", keyword="", process="", category="", expired="", start_date="", end_date="",
db=db, user=user,
).body.decode("utf-8-sig")
rows = list(csv.reader(io.StringIO(body)))
header, lines = rows[0], rows[1:]
check("导出多带一列「处理次数」", "处理次数" in header, str(header))
check("导出第一列说清这一行是审批还是抄送处理", header[0] == "类型", str(header))
check("导出也一张单一行", len(lines) == 3, str(len(lines)))
check("纯审批的这一沓,「类型」那一列全是审批办理",
{line[header.index("类型")] for line in lines} == {"审批办理"}, str(lines))
serial_at, times_at = header.index("单号"), header.index("处理次数")
counts = {line[serial_at]: line[times_at] for line in lines}
check("导出那一列跟页面同一个次数",
counts == {f"DEDUP{w}{STAMP}": str(n) for w, n in (("A", 3), ("B", 1), ("C", 2))}, str(counts))
check("首页角标数单不数行", home_counts(db, user)["task_completed"] == 3,
str(home_counts(db, user)["task_completed"]))
check("工作台那个「我的已办」与列表同一个口径",
workbench(user=user, db=db)["data"]["completed"] == 3)
finally:
db.rollback()
with SessionLocal() as cleaner:
cleaner.execute(delete(models.TaskCompleted).where(models.TaskCompleted.identity.in_([ME, OTHER])))
cleaner.commit()
print()
print(f"通过 {PASSED} 项,失败 {len(FAILS)} 项")
return 1 if FAILS else 0
if __name__ == "__main__":
raise SystemExit(main())