mirror of
https://github.com/ThearchyHelios/AgentLab.git
synced 2026-11-15 07:02:00 +00:00
双盲复算按需算任意快照的遮罩：按位置取自另一份快照遮罩列的输入不再当字面量交给盲算
- 预检以前只为 cell() 输入的快照算遮罩。口径卡上按位置取的一格（vars.res.rows[0][0]）若来自另一份快照，重算 认不出它在遮罩列上，指标照常盲算，遮罩值作为字面量写进了交给盲算模型的计算式。现在 _plan_metric 对这类输入 顺着取值链认出取自哪份快照的哪一列（recompute.input_origin，和证据面板、推演、可验证报告同一个认法），按需 算那份快照的遮罩交给重算的预检：取自遮罩列的按无法复算记缺口（原因同同一份快照上的那种）；那份快照读不出来、 数据源现在的遮罩又没遮这一列的，查询当时记下的遮罩无从核对，同样按无法复算记，原因说清。 - _Sources.hidden 改用证据面板那一套（api/evidence 的 _Masks、_merge_masks、_recorded_masks）：合并查询也按它 合并的几个数据源算，快照读不出来时按取值链上带着的数据源名算，结果按快照记下来不重复查。 - 两个遮罩分组用例换一对盲算 SQL（不分组、按下单日筛）：收窄的话说明和对照里会带出「region = '华东'」；照整条比 时说明是「时间列 + 另有未能归纳的差异」、对照里留着 GROUP BY，以此断言没有收窄，替换失效的「差异不止一条」。 去掉 _narrowed 的遮罩判断后三个遮罩分组用例都变红。 - 补用例：两份快照（cell() 取 A、按位置取 B 的遮罩列）不盲算、取值不进任务和产出，没设遮罩时照常盲算；快照读不 出来时的两种情况；时间列声明成文本的库，审批卡预告不记、真写跳过的原因和预告一字不差。
This commit is contained in:
ThearchyHelios
@@ -13,6 +13,11 @@
|
||||
连取值也不给；遮罩列不给。计算式只有一项来自查询、或者写了口径说明时不给列名；几项都来自查询又没写口径说明的，列名
|
||||
里看不出日期、口径的取舍时才给，好让它分得清哪一项是哪一项。
|
||||
|
||||
**遮罩列上的输入不复算。** cell() 取自遮罩列的输入不能单独复算；别的输入（计算式里写成当时代进去的值）里按位置取自
|
||||
查询结果的一格（vars.res.rows[0][2]），顺着取值链认出它取自哪份快照的哪一列（recompute.input_origin，证据面板、推演、
|
||||
可验证报告同一个认法），按需算那份快照要遮的列（和证据面板同一套，见 _Sources.hidden）——它不一定是 cell() 用到的
|
||||
那几份。取自遮罩列、遮没遮核对不了的，整个指标按无法复算记：遮罩值不能当字面量写进交给盲算的计算式。
|
||||
|
||||
**一致就放行，不一致交给人。** 不一致时逐个对不上的输入说清两条 SQL 差在哪（data/sql_semantics.explain_sql_diff）：
|
||||
两边先收窄到这一格（只留算出这一列的那一项；分组的按这一行的分组值改成等值条件，HAVING 一并去掉），一条查询算好几列、
|
||||
分组后取一行时才不会混进「选出的列」「分组」这类和这一格无关的差别。分组的列设了遮罩时不收窄：这一行的分组值写进
|
||||
@@ -326,11 +331,13 @@ def _skipped(ctx: NodeContext, reason: str) -> dict[str, Any]:
|
||||
|
||||
|
||||
class _Sources:
|
||||
"""一次复算里要读的查询快照和数据源：同一份只读一次。"""
|
||||
"""一次复算里要读的查询快照、数据源和快照要遮的列：同一份只读一次、只算一次。"""
|
||||
|
||||
def __init__(self) -> None:
|
||||
self.snapshots: dict[str, tuple[Any, str | None]] = {}
|
||||
self.rows: dict[str, DataSource | None] = {}
|
||||
self.masks: dict[tuple[str, str | None], frozenset[str]] = {}
|
||||
self._source_masks: Any = None
|
||||
|
||||
def snapshot(self, artifact: str) -> tuple[dict[str, Any] | None, str | None]:
|
||||
if artifact not in self.snapshots:
|
||||
@@ -351,18 +358,25 @@ class _Sources:
|
||||
select(DataSource).where(DataSource.name == name))).scalar_one_or_none()
|
||||
return self.rows[name]
|
||||
|
||||
async def hidden(self, artifact: str) -> frozenset[str]:
|
||||
"""这份快照要遮的列（小写）：查询当时记下的，加上数据源现在设的。宁可多遮。"""
|
||||
from app.data.engine import masked_columns
|
||||
async def hidden(self, artifact: str, source: str | None = None) -> frozenset[str]:
|
||||
"""这份快照要遮的列（小写），和证据面板同一套（api/evidence.py 的 _InputMasks._hidden）：查询当时记下的，加上
|
||||
数据源现在设的；合并查询按它合并的那几个数据源算，换了名字的结果列按来历遮。快照读不出来时按 source（取值链上
|
||||
那份结果里带着的数据源名）现在设的算：遮罩不能因为快照坏了就失效。宁可多遮。"""
|
||||
from app.api.evidence import _Masks, _merge_masks, _recorded_masks # api.evidence 导入了 engine 里的模块
|
||||
from app.engine.merge_query import MERGE_SOURCE
|
||||
|
||||
snap, _ = self.snapshot(artifact)
|
||||
recorded = (snap or {}).get("mask_columns")
|
||||
out = {c.lower() for c in recorded if isinstance(c, str)} if isinstance(recorded, list) else set()
|
||||
name = (snap or {}).get("source")
|
||||
if isinstance(name, str) and name:
|
||||
row = await self.source(name)
|
||||
out |= {c.lower() for c in masked_columns(row.options if row is not None else None)}
|
||||
return frozenset(out)
|
||||
key = (artifact, source)
|
||||
if key not in self.masks:
|
||||
if self._source_masks is None:
|
||||
self._source_masks = _Masks()
|
||||
snap, _ = self.snapshot(artifact)
|
||||
if snap is not None and snap.get("source") == MERGE_SOURCE:
|
||||
hidden, _ = await _merge_masks(snap, self._source_masks)
|
||||
else:
|
||||
current, _ = await self._source_masks.of((snap or {}).get("source") or source)
|
||||
hidden = set(current) | (_recorded_masks(snap) if snap is not None else set())
|
||||
self.masks[key] = frozenset(hidden)
|
||||
return self.masks[key]
|
||||
|
||||
|
||||
def _loader(state: GraphState) -> tuple[Callable[[str], Any], dict[str, str]]:
|
||||
@@ -461,6 +475,8 @@ async def _plan_metric(state: GraphState, metric: dict[str, Any], options: _Opti
|
||||
problem: str | None = None
|
||||
for item in metric.get("inputs") or []:
|
||||
if not _from_query(item):
|
||||
if isinstance(item, dict):
|
||||
problem = problem or await _positional_masks(item, sources, load, outputs, masks)
|
||||
continue
|
||||
info, why = await _query_input(item, sources, aliases, tops)
|
||||
masks[item["artifact"]] = await sources.hidden(item["artifact"])
|
||||
@@ -495,6 +511,30 @@ async def _plan_metric(state: GraphState, metric: dict[str, Any], options: _Opti
|
||||
return row, jobs
|
||||
|
||||
|
||||
async def _positional_masks(item: dict[str, Any], sources: _Sources, load: Callable[[str], Any],
|
||||
outputs: dict[str, str], masks: dict[str, frozenset[str]]) -> str | None:
|
||||
"""不是 cell()、Agent 字段的输入里，按位置取自一份查询结果的一格（vars.res.rows[0][2]）：认出它取自哪份快照的哪一列
|
||||
（recompute.input_origin，证据面板、推演、可验证报告同一个认法），把那份快照要遮的列记进 masks，重算的预检据此认出它
|
||||
取自遮罩列、按遮罩说（同一份快照上的那种一样）。只算 cell() 用到的几份快照时，这一格落在另一份快照上就认不出，遮罩值
|
||||
会当字面量写进交给盲算的计算式。
|
||||
|
||||
那份快照读不出来、数据源现在设的遮罩又没遮这一列时，查询当时记下的遮罩无从核对：交回原因，按无法复算记，宁可多遮。
|
||||
取值链上的产出读不出来、那份结果没存快照的（遮没遮核对不了），重算的预检自己会拦下，这里不管。"""
|
||||
def output(node_id: Any) -> Any:
|
||||
key = outputs.get(str(node_id or ""))
|
||||
return load(key) if key else None
|
||||
|
||||
origin = recompute.input_origin(item, output)
|
||||
if origin is None or origin.unverified is not None or not origin.artifact:
|
||||
return None
|
||||
hidden = await sources.hidden(origin.artifact, origin.source)
|
||||
masks[origin.artifact] = hidden
|
||||
if any(c.lower() in hidden for c in origin.columns):
|
||||
return None
|
||||
_, why = sources.snapshot(origin.artifact)
|
||||
return f"输入 {item.get('path')} 所在的{why}，核对不了这一格是否取自设置了遮罩的列" if why else None
|
||||
|
||||
|
||||
async def _query_input(item: dict[str, Any], sources: _Sources, aliases: dict[str, str], tops: list[str]
|
||||
) -> tuple[dict[str, Any], str | None]:
|
||||
"""一个来自查询快照的输入：(盲算要用的信息, None) 或 (_, 为什么不能单独复算)。"""
|
||||
|
||||
@@ -7,13 +7,13 @@
|
||||
- 不一致：挂起，审批 payload 和对照工件按约定，payload 另带实际生效的 on_reject、摘要里每个指标带选 a、选 b 各会往
|
||||
数据目录里写什么的预告；a / b / neither 三种选择的产出和缺口，没选的照实记，驳回、on_reject 为 fail 时失败
|
||||
- 差异说明只比这一格所在的那一列（一条查询算好几列、分组后取一行的，收窄了再比，HAVING 一并去掉）；分组的列设了
|
||||
遮罩时（改了名再分组的也算）不收窄，遮罩列的取值不进审批卡、对照工件、产出和事件
|
||||
遮罩时（改了名再分组的也算）不收窄、照整条比，遮罩列的取值不进审批卡、对照工件、产出和事件
|
||||
- 只差一个方面时写数据目录（业务日期、有效记录条件），差了不止一处不写，人工确认过的不同说法不覆盖，指标自己的
|
||||
筛选范围（取值在运行输入里、用到分组的列）不写，两条 SQL 规范化后一致的不写；人工驳回过的条件不会借着和别的
|
||||
指标并成一项写回去，真写的和审批卡上的预告一致
|
||||
指标并成一项写回去，真写的和审批卡上的预告一致；时间列在表结构里声明成文本的不记成业务日期，预告照实说
|
||||
- 全局关闭直通、节点覆盖；设置按发起时的快照；重放时盲算不再执行、事件不重复；报告的写作目录不含盲算的查询
|
||||
- 触顶、分组结果的维度（只给取值、遮罩列不给）、带着取舍的列名不交给盲算、两个指标用到同一格只盲算一次、遮罩列上的
|
||||
盲算结果不采用
|
||||
盲算结果不采用；按位置取自另一份快照遮罩列的输入（cell() 取 A、vars.res.rows[0][0] 取 B）按无法复算记，不盲算
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
@@ -31,6 +31,7 @@ from app.db.base import SessionLocal
|
||||
from app.db.models import Approval, Run, RunEvent, Setting
|
||||
from app.engine import spend
|
||||
from app.engine.launch import launch_run
|
||||
from app.engine.nodes.dual_check import GAP_UNCHECKED
|
||||
from app.engine.runner import run_manager
|
||||
from app.main import app
|
||||
from tests.fixtures.dual_check_shop import (
|
||||
@@ -567,6 +568,37 @@ async def test_the_same_choice_twice_writes_the_catalog_once(monkeypatch, shop,
|
||||
# --------------------------------------------------------------------------
|
||||
|
||||
|
||||
@pytest.fixture
|
||||
async def text_dated_shop(tmp_path):
|
||||
"""同一个库，只是下单日、支付日两列声明成文本（SQLite 常这样存日期）。"""
|
||||
name, source_id = await add_source(build(tmp_path / "text_dated.db", date_type="TEXT"))
|
||||
yield name, source_id
|
||||
await drop_source(source_id)
|
||||
|
||||
|
||||
async def test_time_columns_declared_as_text_are_not_written_as_the_business_date(monkeypatch, text_dated_shop, client):
|
||||
"""两边按不同的列筛时间，可这两列在表结构里声明成文本：认不准存的是不是日期，选哪边都不记成业务日期。审批卡上的
|
||||
预告照实说（认不出可记的事实，不带写到哪），裁决后时间线上跳过的原因和预告一字不差。声明成 DATE 的同一个库照常记
|
||||
（test_choosing_the_blind_result_degrades_and_writes_the_catalog）。"""
|
||||
name, source_id = text_dated_shop
|
||||
Blind(monkeypatch, name, {"销售额": (RIGHT, "gmv")})
|
||||
run = await start(graph(name, BY_ORDER_DATE))
|
||||
assert run.status == "interrupted", run.error
|
||||
[item] = (await pending(run.id)).payload["summary"]
|
||||
assert len(item["diff"]) == 1 and "order_date" in item["diff"][0] and "paid_date" in item["diff"][0]
|
||||
said = item["catalog"]
|
||||
# 两列都是文本：各说被采纳的那一边的列
|
||||
for side, label, column in (("a", "原流程", "order_date"), ("b", "盲算", "paid_date")):
|
||||
assert set(said[side]) == {"write", "reason"} and said[side]["write"] is False, said[side]
|
||||
assert f"{label}按时间筛的列「orders.{column}」在表结构里声明为 TEXT，无法确认是日期 / 时间类型" \
|
||||
in said[side]["reason"]
|
||||
done = await decide(client, run.id, {"gmv": "b"})
|
||||
assert done.status == "succeeded", done.error
|
||||
[wrote] = logs(await events(run.id), "dual_check_catalog")
|
||||
assert wrote["written"] == [] and [s["reason"] for s in wrote["skipped"]] == [said["b"]["reason"]]
|
||||
assert await notes_of(source_id) is None
|
||||
|
||||
|
||||
async def test_a_refund_filter_becomes_the_valid_row_filter(monkeypatch, shop, client):
|
||||
name, source_id = shop
|
||||
Blind(monkeypatch, name, {"销售额": (RIGHT, "gmv")})
|
||||
@@ -969,42 +1001,60 @@ async def test_masked_group_columns_are_not_shown_to_the_blind_agent(monkeypatch
|
||||
|
||||
async def test_masked_group_values_stay_out_of_the_card_and_the_catalog_log(monkeypatch, masked_shop, client):
|
||||
"""原流程按遮罩列（region）分组、口径卡取华东那一行：差异说明收窄到这一行时，不能拿这一行的分组值补等值条件——
|
||||
那样「region = '华东'」就进了审批卡、对照工件、事件和目录日志。遮罩的分组列上不收窄，照整条比。"""
|
||||
那样「region = '华东'」就进了审批卡、对照工件、事件和目录日志。遮罩的分组列上不收窄，照整条比。
|
||||
|
||||
盲算没分组、按下单日筛：收窄了的话，两边差在时间列和原流程多出来的「region = '华东'」，说明和对照里都带着这个取值；
|
||||
照整条比，时间列照样说得出，原流程的分组、多选的列归纳不了（_compared_whole）。"""
|
||||
name, _ = masked_shop
|
||||
first = {**GMV, "id": "first", "name": "首组销售额", "expression": "cell(nodes.fetch, 0, 'gmv')"}
|
||||
Blind(monkeypatch, name, {"首组销售额": (RIGHT, "gmv")})
|
||||
Blind(monkeypatch, name, {"首组销售额": (BY_ORDER_DATE, "gmv")})
|
||||
run = await start(graph(name, BY_REGION, metrics=[first], dual={"definitions": {}}))
|
||||
assert run.status == "interrupted", run.error
|
||||
payload = (await pending(run.id)).payload
|
||||
[item] = payload["summary"]
|
||||
assert item["original_rendered"].startswith("300") and item["blind_rendered"].startswith("310")
|
||||
card = artifact_store.load(payload["card_artifact"])
|
||||
done = await decide(client, run.id, {"first": "a"})
|
||||
assert done.status == "succeeded", done.error
|
||||
evs = await events(run.id)
|
||||
_no_masked_value(payload, card, done, evs)
|
||||
# 照整条比：原流程分组、盲算没分组都列出来，不止差一个方面，选了也不写目录
|
||||
assert len(payload["summary"][0]["diff"]) > 1
|
||||
_compared_whole(payload, card)
|
||||
# 差异说明里有未能归纳的部分：选了也不写目录
|
||||
[wrote] = logs(evs, "dual_check_catalog")
|
||||
assert wrote["written"] == [] and "不止差一个方面" in wrote["skipped"][0]["reason"]
|
||||
|
||||
|
||||
async def test_a_renamed_masked_group_column_is_not_narrowed_either(monkeypatch, masked_shop, client):
|
||||
"""原流程把遮罩列改了名再分组（region AS area）：结果列名 area 不在遮罩里，分组值却还是 region 的取值。收窄时看的
|
||||
是分组键用到的列，一样不拿这一行的分组值补等值条件，「华东」不进审批卡、对照工件、产出和事件。"""
|
||||
是分组键用到的列，一样不拿这一行的分组值补等值条件，「华东」不进审批卡、对照工件、产出和事件，照整条比。"""
|
||||
name, _ = masked_shop
|
||||
renamed = (f"SELECT region AS area, SUM(amount) AS gmv FROM orders WHERE status <> 'refunded' "
|
||||
f"AND paid_date {WEEK} GROUP BY region ORDER BY region")
|
||||
first = {**GMV, "id": "first", "name": "首组销售额", "expression": "cell(nodes.fetch, 0, 'gmv')"}
|
||||
Blind(monkeypatch, name, {"首组销售额": (RIGHT, "gmv")})
|
||||
Blind(monkeypatch, name, {"首组销售额": (BY_ORDER_DATE, "gmv")})
|
||||
run = await start(graph(name, renamed, metrics=[first], dual={"definitions": {}}))
|
||||
assert run.status == "interrupted", run.error
|
||||
payload = (await pending(run.id)).payload
|
||||
assert payload["summary"][0]["original_rendered"].startswith("300")
|
||||
[item] = payload["summary"]
|
||||
assert item["original_rendered"].startswith("300") and item["blind_rendered"].startswith("310")
|
||||
card = artifact_store.load(payload["card_artifact"])
|
||||
done = await decide(client, run.id, {"first": "a"})
|
||||
assert done.status == "succeeded", done.error
|
||||
_no_masked_value(payload, card, done, await events(run.id))
|
||||
# 照整条比：原流程分组、盲算没分组都列出来
|
||||
assert len(payload["summary"][0]["diff"]) > 1
|
||||
_compared_whole(payload, card)
|
||||
|
||||
|
||||
def _compared_whole(payload: dict, card: dict) -> None:
|
||||
"""差异说明照整条比的痕迹（原流程按地区分组取一行，盲算没分组、按下单日筛）：时间列照样说得出；原流程的分组、多选的
|
||||
地区列归纳不了，另起一条「另有未能归纳的差异」；规范化后的对照里还留着原流程的 GROUP BY。收窄了的话分组去掉、换成
|
||||
等值条件，两边都是简单形状，说明里就是时间列和多出来的那个条件，这几样都不会有。"""
|
||||
[metric] = card["metrics"]
|
||||
diff = metric["diff"]
|
||||
assert [i["aspect"] for i in diff["items"]] == ["time", "other"], diff["items"]
|
||||
assert diff["fallback"] is True and any("GROUP BY" in line for line in diff["text_diff"]), diff["text_diff"]
|
||||
[item] = payload["summary"]
|
||||
assert len(item["diff"]) == 2 and "order_date" in item["diff"][0] and "另有未能归纳的差异" in item["diff"][1], item
|
||||
assert item["fallback"] is False
|
||||
|
||||
|
||||
async def test_a_masked_group_column_of_the_blind_query_is_not_narrowed_either(monkeypatch, masked_shop, client):
|
||||
@@ -1021,6 +1071,81 @@ async def test_a_masked_group_column_of_the_blind_query_is_not_narrowed_either(m
|
||||
_no_masked_value(payload, card, done, await events(run.id))
|
||||
|
||||
|
||||
#: 销售额（第一份快照的一格）加上首组地区名的字数（第二份快照按位置取的一格）
|
||||
MIX = {"id": "mix", "name": "销售额加首组名称字数", "unit": "元", "decimals": 1, "format": "thousands",
|
||||
"expression": "cell(nodes.fetch, 0, 'gmv') + len(vars.res.rows[0][0])"}
|
||||
|
||||
|
||||
def with_regions(spec: dict, source: str) -> dict:
|
||||
"""取数之后另查一次按地区分组的结果（第二份快照），数据整形把它解析进 vars.res：口径卡按位置取它的格。"""
|
||||
nodes = list(spec["nodes"])
|
||||
at = next(i for i, n in enumerate(nodes) if n["id"] == "fetch") + 1
|
||||
nodes[at:at] = [node("regions", "tool", tool=f"db_query__{source}", args={"sql": BY_REGION}),
|
||||
node("parse", "transform", mode="json", template="{{ nodes.regions }}", assign_to="res")]
|
||||
return {"nodes": nodes, "edges": [{"source": a["id"], "target": b["id"]} for a, b in zip(nodes, nodes[1:])]}
|
||||
|
||||
|
||||
async def test_a_masked_cell_read_by_position_from_another_snapshot_is_not_blind_computed(monkeypatch, masked_shop):
|
||||
"""口径卡上一个指标：cell() 取第一次查询（A，不碰遮罩列）的销售额，另按位置取第二次查询（B，按遮罩列 region 分组）的
|
||||
首组名称（vars.res.rows[0][0]）。B 不是 cell() 用到的快照：以前预检只为 A 算遮罩，认不出按位置取的那一格在遮罩列上，
|
||||
指标照常盲算，「华东」作为字面量写进了交给盲算模型的计算式。现在按需算 B 的遮罩：这个指标按无法复算记缺口、不盲算，
|
||||
遮罩列的取值不进盲算的任务、节点产出和节点发的事件。"""
|
||||
name, _ = masked_shop
|
||||
blind = Blind(monkeypatch, name, {"销售额加首组名称字数": (RIGHT, "gmv")})
|
||||
run = await start(with_regions(graph(name, RIGHT, metrics=[MIX], dual={"definitions": {}}), name))
|
||||
assert run.status == "succeeded", run.error
|
||||
evs = await events(run.id)
|
||||
out = dual_output(run, evs)
|
||||
[row] = out["metrics"]
|
||||
assert row["metric_id"] == "mix" and row["status"] == "unchecked", row
|
||||
assert "遮罩" in row["reason"] and "推演" not in row["reason"], row["reason"]
|
||||
assert out["status"] == "unchecked" and out["issuance_gaps"] == [GAP_UNCHECKED.format(name=MIX["name"],
|
||||
reason=row["reason"])]
|
||||
assert blind.tasks == [] and blind.runs == 0, blind.tasks
|
||||
assert "华东" not in json.dumps(out, ensure_ascii=False)
|
||||
assert not [e.type for e in evs if e.node_id == "dual" and not e.type.startswith(("tool.", "llm."))
|
||||
and "华东" in json.dumps(e.data, ensure_ascii=False)]
|
||||
|
||||
|
||||
async def test_an_unmasked_cell_read_by_position_from_another_snapshot_is_blind_computed(monkeypatch, shop):
|
||||
"""同样的图、数据源没设遮罩：预检照常通过，指标进入盲算（按位置取的那一格作为字面量写进计算式），两边一致。拦下上面
|
||||
那个指标的确实是遮罩。"""
|
||||
name, _ = shop
|
||||
blind = Blind(monkeypatch, name, {"销售额加首组名称字数": (RIGHT, "gmv")})
|
||||
run = await start(with_regions(graph(name, RIGHT, metrics=[MIX], dual={"definitions": {}}), name))
|
||||
assert run.status == "succeeded", run.error
|
||||
out = dual_output(run, await events(run.id))
|
||||
assert out["status"] == "agreed" and out["metrics"][0]["status"] == "agreed", out
|
||||
[task] = blind.tasks
|
||||
assert "len('华东')" in task
|
||||
|
||||
|
||||
#: 按位置取的那一格：数据整形解析好的第二份查询结果（首组地区名），快照 id 是编的
|
||||
POSITIONAL = {"path": "vars.res.rows[0][0]", "value": "华东", "node_id": "parse", "via": "transform", "status": "ok"}
|
||||
LOST = "cd" * 32
|
||||
|
||||
|
||||
def parsed_regions(source: str) -> dict:
|
||||
return {"columns": ["region", "gmv"], "rows": [["华东", 300.0], ["华北", 30.0]], "artifact": LOST, "source": source}
|
||||
|
||||
|
||||
async def test_a_cell_read_by_position_from_an_unreadable_snapshot_is_not_trusted(masked_shop):
|
||||
"""按位置取的那一格所在的快照读不出来（被改过、文件不在了）：查询当时记下的遮罩无从核对。数据源现在设的遮罩遮着这一列
|
||||
的，照遮罩算（交给重算的预检按遮罩说）；没遮的也不当成没遮，按无法复算记、原因说清。"""
|
||||
from app.engine.nodes.dual_check import _positional_masks, _Sources
|
||||
|
||||
name, _ = masked_shop
|
||||
for source, expected in ((name, None), ("dc_no_such_source", "输入 vars.res.rows[0][0] 所在的查询快照与哈希不一致，"
|
||||
"疑似被修改，核对不了这一格是否取自设置了遮罩的列")):
|
||||
sources = _Sources()
|
||||
sources.snapshots[LOST] = (None, "查询快照与哈希不一致，疑似被修改")
|
||||
masks: dict = {}
|
||||
state = {"state:parse": parsed_regions(source)}
|
||||
why = await _positional_masks(POSITIONAL, sources, state.get, {"parse": "state:parse"}, masks)
|
||||
assert why == expected, source
|
||||
assert ("region" in masks[LOST]) is (source == name), masks
|
||||
|
||||
|
||||
def _no_masked_value(payload: dict, card: dict, done: Run, evs: list[RunEvent], value: str = "华东") -> None:
|
||||
"""遮罩列上的取值不出现在审批 payload、对照工件、节点产出和节点发的事件里。查询本身的 tool.end、模型的往来按
|
||||
设计带着原始结果（证据面板显示时再遮），不在此列。"""
|
||||
|
||||
Reference in New Issue
Block a user