mirror of
https://github.com/ThearchyHelios/AgentLab.git
synced 2026-11-15 00:11:59 +00:00
ThearchyHelios
2d5e50fa88
配方语言（agentlab-recipe/2，新 recipe_types、recipe） - 封闭、不含正则、保存时静态校验：常量只能从表里的候选文字里选，占位符只能映射成空值，单位从固定词表里取 - 交叉表和列表两种版式；分段各自定位、不依赖先后顺序；合计按标签语义区间核对，关系恒等式、口径不同的声明都写在配方里 - 固定的解析器集合（parsers/2）：全角、各类横线、千分位、月日写法宽容归一，规范写法存库、原文另记 执行器（新 xlsx_cells、recipe_engine） - 两遍读取逐行对账，每个非空格都要有去向（格子账）；交叉表逆透视成长表，按统计期补年份并断言覆盖整期 - 合计行改作核对并另存一张表，不和明细混在一列；区域外文字、隐藏行列、合并单元格、空行之后的文字都单独报出 - STRICT 加主键写临时库；按锚点段记溯源，供后续证据下钻用；大表流式读取，扫描在 sheetData 处停止 校验与说明（新 recipe_checks、recipe_notes） - SQL 核对分通过、不一致、无法核对三种结果，失败分结构、数据质量、需确认三类；结构类直接拒收 - 数据质量类可以写理由接受，接受不改库、只改状态和说明 - 表的说明按核对结果生成、不含数字，经过数字检查；没核对过的不写「已核对」 起草（新 recipe_suggest、recipe_ai） - 先用规则起草：建议卡片加待确认问题（关系要不要登记、占位符表示什么） - 规则起草不完整时才提示 AI 兜底：界面先展示要发出的全部文字，用户同意才发；只发结构，数字、隐藏行列、遮罩列不发，公式先脱敏；同意绑定模型和服务方；AI 只写配方，有限次修订，记 token 和费用；截断单独识别 - 无论规则还是 AI 起草，都必须人工确认才能启用 导入流程（新 recipe_imports、recipe_confirm、recipe_diff、api/table_imports） - 暂存 → 起草 → 回答问题或改配方 → 试运行 → 勾确认项、写接受理由 → 确认并启用；接入期 1 的构建、导入记录、快照和运行固定版本 - 上传新一期按已确认的配方重放，不调用模型，出回执和差异卡，手动启用 - 改配方报破坏性变更（breaking:*）、单位变化、工作表改名的兜底，都要逐项勾选确认 前端导入向导（新 pages/import） - 原始网格（行列号、去向着色）、建议卡片、配方面板（表单，不露正则）、试运行回执、确认清单、差异卡、AI 起草的同意对话框
352 lines
17 KiB
Python
352 lines
17 KiB
Python
"""9.4 的数据变异（WP-7 验收）：在 XML 层改 D00 的字节（tests/fixtures/xlsx/mutate.py），按参考配方重放。
|
||
|
||
每个用例先把 D00（2026-08-01 起 31 天）按参考配方首次导入并启用，再把变异后的字节作为「上传新一期」重放：
|
||
上传新一期自动试运行，回来时已带状态、问题、核对、回执和差异卡。期望照 P2-SPEC 9.4 的表逐条断言，
|
||
另加 MUTATIONS 里的「外加」用例 PKDUP（同一分段里一行写 8-9、另一行写 8－9）。
|
||
|
||
拒收的用例另外断言：不能提交（409 trial_not_passed），当前快照没变（9.7 第 4 步「拒收的确认当前快照没变」）。
|
||
夹具全部合成、假名；不调用任何模型。
|
||
"""
|
||
from __future__ import annotations
|
||
|
||
import datetime as dt
|
||
import json
|
||
import re
|
||
import uuid
|
||
from pathlib import Path
|
||
from typing import Any
|
||
|
||
import pytest
|
||
from httpx import ASGITransport, AsyncClient
|
||
|
||
from app.core.config import settings
|
||
from app.data import table_versions
|
||
from app.data.engine import engines, run_query
|
||
from app.db.base import SessionLocal
|
||
from app.db.models import DataSource, SourceSnapshot
|
||
from app.main import app
|
||
from tests.fixtures.xlsx.flow import FLOW_EXPECT, flow_workbook
|
||
from tests.fixtures.xlsx.mutate import MUTATIONS, mutate
|
||
|
||
XLSX = "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"
|
||
FLOW_RECIPE = json.loads((Path(__file__).parent / "fixtures" / "recipes" / "flow_recipe.json").read_text("utf-8"))
|
||
AUG = dt.date(2026, 8, 1)
|
||
SHEET = "客流汇总"
|
||
TOTALS_TABLE = "时段客流_表内合计"
|
||
|
||
|
||
@pytest.fixture(autouse=True)
|
||
def store(monkeypatch, tmp_path):
|
||
"""每个测试一个独立的数据目录：上传目录、原件存档、试运行库、工件都在里面。"""
|
||
data = tmp_path / "data"
|
||
(data / "uploads").mkdir(parents=True)
|
||
monkeypatch.setattr(settings, "data_dir", data)
|
||
yield data
|
||
|
||
|
||
@pytest.fixture
|
||
async def client():
|
||
async with AsyncClient(transport=ASGITransport(app=app), base_url="http://test") as c:
|
||
yield c
|
||
|
||
|
||
# --------------------------------------------------------------------------
|
||
# 工具
|
||
# --------------------------------------------------------------------------
|
||
|
||
|
||
def _cell(ref: str) -> str:
|
||
return f"{SHEET}!{ref}"
|
||
|
||
|
||
def _confirm_ids(st: dict[str, Any]) -> list[str]:
|
||
return [i["id"] for i in st["trial"]["confirm_items"]]
|
||
|
||
|
||
def _problems(trial: dict[str, Any], code: str) -> list[dict[str, Any]]:
|
||
return [p for p in trial["problems"] if p["code"] == code]
|
||
|
||
|
||
def _check(trial: dict[str, Any], cid: str) -> dict[str, Any] | None:
|
||
return next((c for c in trial["checks"] if c["id"] == cid), None)
|
||
|
||
|
||
async def _baseline(client, seed: int) -> dict[str, Any]:
|
||
"""D00 按参考配方首次导入并启用：暂存 → PUT 参考配方 → 试运行（passed）→ 勾全部确认项提交。"""
|
||
raw, fn = flow_workbook(AUG, 31, seed=seed)
|
||
name = f"wp7mut_{uuid.uuid4().hex[:10]}"
|
||
resp = await client.post("/api/datasources/imports/stage", files={"file": (fn, raw, XLSX)}, data={"name": name})
|
||
assert resp.status_code == 201, resp.text
|
||
sid = resp.json()["id"]
|
||
resp = await client.put(f"/api/datasources/imports/{sid}/recipe", json={"recipe": FLOW_RECIPE})
|
||
assert resp.status_code == 200, resp.text
|
||
assert resp.json()["recipe_problems"] == []
|
||
resp = await client.post(f"/api/datasources/imports/{sid}/trial", json={})
|
||
assert resp.status_code == 200, resp.text
|
||
st = resp.json()
|
||
trial = st["trial"]
|
||
assert trial["status"] == "passed", trial["problems"]
|
||
assert {t["name"]: t["rows"] for t in trial["receipt"]["tables"]} == FLOW_EXPECT["rows"]
|
||
resp = await client.post(f"/api/datasources/imports/{sid}/commit",
|
||
json={"trial_id": trial["trial_id"], "confirmations": _confirm_ids(st)})
|
||
assert resp.status_code == 201, resp.text
|
||
out = resp.json()
|
||
return {"raw": raw, "fn": fn, "source_id": out["source"]["id"], "snapshot_id": out["snapshot_id"],
|
||
"import_id": out["import_id"], "table_hashes": trial["receipt"]["table_hashes"]}
|
||
|
||
|
||
async def _replay(client, base: dict[str, Any], case: str) -> dict[str, Any]:
|
||
"""变异后的 D00 作为「上传新一期」重放（文件名不变），返回自动试运行后的暂存区。"""
|
||
raw = mutate(base["raw"], case)
|
||
assert raw != base["raw"]
|
||
resp = await client.post(f"/api/datasources/{base['source_id']}/reupload",
|
||
files={"file": (base["fn"], raw, XLSX)})
|
||
assert resp.status_code == 201, resp.text
|
||
st = resp.json()
|
||
assert st["kind"] == "reupload" and st["trial"] is not None, st
|
||
return st
|
||
|
||
|
||
async def _current_snapshot(source_id: str) -> str | None:
|
||
async with SessionLocal() as session:
|
||
row = await session.get(DataSource, source_id)
|
||
return row.current_snapshot_id
|
||
|
||
|
||
async def _assert_rejected(client, base: dict[str, Any], st: dict[str, Any], codes: tuple[str, ...]) -> None:
|
||
"""拒收：状态 rejected、期望的问题 code 都在、提交不了、当前快照没变。"""
|
||
trial = st["trial"]
|
||
assert trial["status"] == "rejected", trial["problems"]
|
||
assert st["status"] == "rejected"
|
||
got = {p["code"] for p in trial["problems"]} | {c["id"] for c in trial["checks"] if c["status"] != "passed"}
|
||
for code in codes:
|
||
assert code in got, (code, trial["problems"], trial["checks"])
|
||
# 拒收时不算差异卡和确认项（7.6）
|
||
assert trial["diff"] is None and trial["confirm_items"] == []
|
||
resp = await client.post(f"/api/datasources/imports/{st['id']}/commit",
|
||
json={"trial_id": trial["trial_id"], "confirmations": []})
|
||
assert resp.status_code == 409 and resp.json()["code"] == "trial_not_passed", resp.text
|
||
assert await _current_snapshot(base["source_id"]) == base["snapshot_id"]
|
||
|
||
|
||
async def _scalar(source_id: str, sql: str) -> Any:
|
||
async with SessionLocal() as session:
|
||
row = await session.get(DataSource, source_id)
|
||
view = await table_versions.resolve_source(session, row)
|
||
try:
|
||
return (await run_query(view, sql)).rows[0][0]
|
||
finally:
|
||
await engines.invalidate(view.id)
|
||
|
||
|
||
def _table_meta(schema_cache: dict[str, Any], table: str) -> dict[str, Any]:
|
||
"""schema_cache 里某张表的结构（introspect 的形状：tables = {表名: {comment, columns: [...]}}）。"""
|
||
tables = schema_cache.get("tables") or {}
|
||
assert table in tables, sorted(tables)
|
||
return tables[table]
|
||
|
||
|
||
# --------------------------------------------------------------------------
|
||
# 夹具本身：MUTATIONS 照 9.4 抄录
|
||
# --------------------------------------------------------------------------
|
||
|
||
|
||
def test_mutation_table_matches_spec():
|
||
expect = {
|
||
"P1": ("passed", ()), "P1b": ("needs_decision", ("K1",)), "P2": ("rejected", ("value_blank",)),
|
||
"P6": ("rejected", ("axis_duplicate",)), "P9": ("rejected", ("K1", "G1")),
|
||
"P20": ("rejected", ("row_unclaimed",)), "P23": ("passed", ()), "P24": ("passed", ()),
|
||
"P25": ("rejected", ("axis_extra_cells",)), "P26": ("rejected", ("axis_coverage",)),
|
||
"PKDUP": ("rejected", ("pk_duplicate", "label_duplicate")),
|
||
}
|
||
assert {k: v[1:] for k, v in MUTATIONS.items()} == expect
|
||
|
||
|
||
# --------------------------------------------------------------------------
|
||
# 能通过的变异
|
||
# --------------------------------------------------------------------------
|
||
|
||
|
||
async def test_p1_totals_without_formula_pass_by_label_range(client):
|
||
"""P1：第 28–30 行去掉公式、保留缓存值 → passed（K1 按标签区间核对通过，derived_form=literal）。"""
|
||
base = await _baseline(client, seed=21)
|
||
st = await _replay(client, base, "P1")
|
||
trial = st["trial"]
|
||
assert trial["status"] == "passed", trial["problems"]
|
||
assert [p for p in trial["problems"] if p["category"] in ("structure", "recipe", "input")] == []
|
||
k1 = _check(trial, "K1")
|
||
assert k1 is not None and k1["status"] == "passed" and k1["checked"] == 93, k1
|
||
# 写死的合计没有公式：不出 G1（9.2 D07 同理）
|
||
assert _check(trial, "G1") is None
|
||
assert trial["receipt"]["derived_form"] == {"夜间合计": "literal"}
|
||
# 值没变：三张表的哈希与 D00 相同
|
||
assert trial["receipt"]["table_hashes"] == base["table_hashes"]
|
||
# 合计从公式变成写死的数：差异卡要求确认（同 9.2 D07 的 derived_form；G1 消失记 checks）
|
||
kinds = {d["kind"] for d in trial["diff"]}
|
||
assert kinds == {"checks", "derived_form"}, trial["diff"]
|
||
assert "diff:derived_form:夜间合计" in _confirm_ids(st)
|
||
|
||
|
||
async def test_p1b_uncached_totals_need_decision_then_commit_with_null_totals(client):
|
||
"""P1b：第 28–30 行保留公式、去掉缓存值 → needs_decision：K1 unverifiable（uncached: 93，可接受）、
|
||
G1 passed；写理由接受后提交，合计表的客流全为 NULL、表结构保留，说明写「未能核对」、不写「一致」。"""
|
||
base = await _baseline(client, seed=22)
|
||
st = await _replay(client, base, "P1b")
|
||
trial = st["trial"]
|
||
assert trial["status"] == "needs_decision", trial["problems"]
|
||
k1 = _check(trial, "K1")
|
||
assert k1 is not None and k1["status"] == "unverifiable", k1
|
||
assert k1["reasons"] == {"uncached": 93} and k1["acceptable"] is True
|
||
g1 = _check(trial, "G1")
|
||
assert g1 is not None and g1["status"] == "passed", g1
|
||
assert trial["acceptable"] == ["K1"]
|
||
# 试运行时的说明预览（还没接受）也不能说「一致」
|
||
preview = trial["notes"][TOTALS_TABLE]["comment"]
|
||
assert "一致" not in preview, preview
|
||
|
||
# 不带接受 → 422 acceptance_required；理由是空白 → 同样 422
|
||
confirms = _confirm_ids(st)
|
||
resp = await client.post(f"/api/datasources/imports/{st['id']}/commit",
|
||
json={"trial_id": trial["trial_id"], "confirmations": confirms})
|
||
assert resp.status_code == 422 and resp.json()["code"] == "acceptance_required", resp.text
|
||
assert "K1" in resp.json()["detail"]
|
||
resp = await client.post(f"/api/datasources/imports/{st['id']}/commit",
|
||
json={"trial_id": trial["trial_id"], "confirmations": confirms,
|
||
"acceptances": [{"check_id": "K1", "reason": " "}]})
|
||
assert resp.status_code == 422 and resp.json()["code"] == "acceptance_required", resp.text
|
||
assert await _current_snapshot(base["source_id"]) == base["snapshot_id"]
|
||
|
||
resp = await client.post(f"/api/datasources/imports/{st['id']}/commit",
|
||
json={"trial_id": trial["trial_id"], "confirmations": confirms,
|
||
"acceptances": [{"check_id": "K1", "reason": "合成理由：导出工具未保存计算结果"}]})
|
||
assert resp.status_code == 201, resp.text
|
||
out = resp.json()
|
||
assert out["snapshot_id"] != base["snapshot_id"]
|
||
sid = base["source_id"]
|
||
# 合计表：行数照旧 93，客流全为 NULL；明细表不受影响
|
||
assert await _scalar(sid, f'SELECT COUNT(*) FROM "{TOTALS_TABLE}"') == 93
|
||
assert await _scalar(sid, f'SELECT COUNT("客流") FROM "{TOTALS_TABLE}"') == 0
|
||
assert await _scalar(sid, 'SELECT COUNT(*) FROM "时段客流"') == FLOW_EXPECT["rows"]["时段客流"]
|
||
async with SessionLocal() as session:
|
||
snap = await session.get(SourceSnapshot, out["snapshot_id"])
|
||
cache = snap.schema_cache
|
||
meta = _table_meta(cache, TOTALS_TABLE)
|
||
cols = [c["name"] for c in meta["columns"]]
|
||
assert cols == ["日期", "合计项", "起始小时", "结束小时", "客流"], cols
|
||
comment = meta.get("comment") or ""
|
||
assert "本期原表未保存计算结果，合计值为空，未能核对" in comment, comment
|
||
assert "一致" not in comment, comment
|
||
for c in meta["columns"]:
|
||
assert "一致" not in (c.get("comment") or ""), c
|
||
|
||
|
||
async def test_p23_dimension_cut_rows_still_reads_everything(client):
|
||
"""P23：<dimension> 截到第 27 行 → passed，table_hashes 与 D00 完全相同。"""
|
||
base = await _baseline(client, seed=23)
|
||
st = await _replay(client, base, "P23")
|
||
trial = st["trial"]
|
||
assert trial["status"] == "passed", trial["problems"]
|
||
assert trial["receipt"]["table_hashes"] == base["table_hashes"]
|
||
assert {t["name"]: t["rows"] for t in trial["receipt"]["tables"]} == FLOW_EXPECT["rows"]
|
||
# 陈旧的 <dimension> 不影响任何东西：内容与上一期逐格相同，没有差异
|
||
assert trial["diff"] == [], trial["diff"]
|
||
|
||
|
||
async def test_p24_dimension_cut_columns_still_reads_everything(client):
|
||
"""P24：<dimension> 截到倒数第二个日期列 → passed，table_hashes 与 D00 完全相同。"""
|
||
base = await _baseline(client, seed=24)
|
||
st = await _replay(client, base, "P24")
|
||
trial = st["trial"]
|
||
assert trial["status"] == "passed", trial["problems"]
|
||
assert trial["receipt"]["table_hashes"] == base["table_hashes"]
|
||
assert {t["name"]: t["rows"] for t in trial["receipt"]["tables"]} == FLOW_EXPECT["rows"]
|
||
assert trial["diff"] == [], trial["diff"]
|
||
|
||
|
||
# --------------------------------------------------------------------------
|
||
# 拒收的变异
|
||
# --------------------------------------------------------------------------
|
||
|
||
|
||
async def test_p2_blank_value_is_rejected_at_d15(client):
|
||
"""P2：D15 改成空 → rejected value_blank（D15）。"""
|
||
base = await _baseline(client, seed=25)
|
||
st = await _replay(client, base, "P2")
|
||
await _assert_rejected(client, base, st, ("value_blank",))
|
||
hits = _problems(st["trial"], "value_blank")
|
||
assert len(hits) == 1 and hits[0]["category"] == "structure"
|
||
assert hits[0]["cells"] == [_cell("D15")], hits[0]
|
||
|
||
|
||
async def test_p6_duplicate_axis_date_is_rejected(client):
|
||
"""P6：R4 改成「8月15日」（与 Q4 重复）→ rejected axis_duplicate。"""
|
||
base = await _baseline(client, seed=26)
|
||
st = await _replay(client, base, "P6")
|
||
await _assert_rejected(client, base, st, ("axis_duplicate",))
|
||
hit = _problems(st["trial"], "axis_duplicate")[0]
|
||
assert hit["category"] == "structure"
|
||
assert {_cell("Q4"), _cell("R4")} <= set(hit["cells"]), hit
|
||
|
||
|
||
async def test_p9_wrong_formula_range_fails_k1_and_g1(client):
|
||
"""P9：C28 改成 =SUM(C22:C24)，缓存值自洽 → rejected：K1 mismatch（C28）且 G1 mismatch。"""
|
||
base = await _baseline(client, seed=27)
|
||
st = await _replay(client, base, "P9")
|
||
await _assert_rejected(client, base, st, ("K1", "G1"))
|
||
trial = st["trial"]
|
||
k1, g1 = _check(trial, "K1"), _check(trial, "G1")
|
||
for c in (k1, g1):
|
||
assert c["status"] == "mismatch" and c["category"] == "structure" and c["acceptable"] is False, c
|
||
assert c["cells"] == [_cell("C28")], c
|
||
assert c["failed"] == 1, c
|
||
# 只有核对没过：执行器本身没有结构问题
|
||
assert [p for p in trial["problems"] if p["category"] in ("structure", "recipe")] == []
|
||
|
||
|
||
async def test_p20_unclaimed_data_row_32_is_rejected(client):
|
||
"""P20：第 32 行加「备用分区（人次）」和两个数，<dimension> 不更新 → rejected row_unclaimed（第 32 行）。"""
|
||
base = await _baseline(client, seed=28)
|
||
st = await _replay(client, base, "P20")
|
||
await _assert_rejected(client, base, st, ("row_unclaimed",))
|
||
hits = _problems(st["trial"], "row_unclaimed")
|
||
assert len(hits) == 1, hits
|
||
assert hits[0]["category"] == "structure"
|
||
assert all(re.fullmatch(rf"{SHEET}![A-Z]+32", c) for c in hits[0]["cells"]), hits[0]
|
||
assert _cell("B32") in hits[0]["cells"]
|
||
|
||
|
||
async def test_p25_extra_total_column_right_of_axis_is_rejected(client):
|
||
"""P25：AH4「合计」、AH 列各数据行写 999，<dimension> 不更新 → rejected axis_extra_cells（AH4）。"""
|
||
base = await _baseline(client, seed=29)
|
||
st = await _replay(client, base, "P25")
|
||
await _assert_rejected(client, base, st, ("axis_extra_cells",))
|
||
hit = _problems(st["trial"], "axis_extra_cells")[0]
|
||
assert hit["category"] == "structure" and _cell("AH4") in hit["cells"], hit
|
||
assert hit["fix"] == "ignore_cells"
|
||
|
||
|
||
async def test_p26_missing_last_date_column_fails_coverage(client):
|
||
"""P26：删掉 AG 列所有格，B2 仍写到 31 日 → rejected axis_coverage。"""
|
||
base = await _baseline(client, seed=30)
|
||
st = await _replay(client, base, "P26")
|
||
await _assert_rejected(client, base, st, ("axis_coverage",))
|
||
hit = _problems(st["trial"], "axis_coverage")[0]
|
||
assert hit["category"] == "structure"
|
||
assert "30" in hit["message"] and "31" in hit["message"], hit
|
||
|
||
|
||
async def test_pkdup_mixed_dash_labels_report_pk_and_label_duplicates(client):
|
||
"""9.4 外加：同一分段里一行写 8-9、另一行写 8－9 → rejected pk_duplicate（两处坐标），label_duplicate
|
||
也出现（4.2 第 3 步：定位失败的分段跳过，其余照常，一次报全）。"""
|
||
base = await _baseline(client, seed=31)
|
||
st = await _replay(client, base, "PKDUP")
|
||
await _assert_rejected(client, base, st, ("pk_duplicate", "label_duplicate"))
|
||
trial = st["trial"]
|
||
pk = _problems(trial, "pk_duplicate")
|
||
assert len(pk) == 1 and pk[0]["category"] == "structure", pk
|
||
# 两处坐标：B11（8-9）和 B12（8－9）那两行同一日期的格子
|
||
assert {_cell("C11"), _cell("C12")} <= set(pk[0]["cells"]), pk[0]
|
||
assert "C11" in pk[0]["message"] and "C12" in pk[0]["message"], pk[0]["message"]
|
||
dup = _problems(trial, "label_duplicate")
|
||
assert dup and _cell("B12") in dup[0]["cells"], dup
|