Files
AgentLab/backend/tests/test_metric_id_duplicate.py
T
ThearchyHelios 2f90f0eb3a 口径台账:固定执行、验证审阅发布与增量重导入
- QueryCore 固定执行(K09):按已发布的不可变口径跑一次,交回类型化标量、发布/manifest 引用与逐查询证据
- 依赖闭包(K08):按选中的指标算出要拷的取数节点、要的入口字段,接进 derive / compose / metrics
- SQL 模板参数完整性(K01):试跑、组合问数、派生图、运行时取数共用一套语义,缺参数在渲染前就拒
- 受限字段变更服务(K03)与口径领域边界(K01):改动打在整图深拷贝上,身份与证据归属不可改
- 版本化受控模板绑定(K04):业务口径按显式绑定落到 SQL,对不上就报 unsupported / needs_update
- 验证记录(K05)与最小审阅:样例匹配不等于正式批准,审批人只从服务端身份取
- 发布依据清单(K07):发布时按不可变图物化内容寻址清单,正式用途按它判
- 增量重导入(K11)与 B/L/R 三方合并、受限自然语言修改提案(K12):只产候选,落图仍走既有写路径
- 接口与前端:口径草稿页接上字段变更、绑定状态、验证与重导入,新增 caliber-drafts / studio 检查脚本

验证:前端 tsc -b 通过。
2026-11-10 15:05:18 +08:00

175 lines
9.2 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
"""指标 ID 收口:同一张口径卡上的指标 ID 要唯一(K08 的重复身份)。
下游 `nodes.<卡>.metrics[i]`、出具契约、比例契约、报告引用都按指标 ID 与位置取数:同一张卡上两个指标同名就分不清是哪一条。
所以**画布 / 发布前检查**(engine/schema)和**运行时**(engine/nodes/metrics)都要拦住,不能取第一条静默执行——那会把另一条的
数顶替过来。闭合解析(ledger/closure.selection_index)遇到同名只留第一次,运行时不能顺着它把第一条算出来,必须在过滤前
就响亮失败(两边都不静默,不冲突)。
空 ID(没写声明的)不算重复:缺 ID 另有报法(画布 metrics.id_missing、运行时「指标定义不完整」)。旧正常图(ID 唯一)
行为一字不变。数据是虚构店铺库。
"""
from __future__ import annotations
import pytest
from app.engine.nodes.metrics import SelectionRefused, duplicate_metric_ids, require_selection
from app.engine.runner import run_manager
from app.engine.schema import GraphSpec, validate_graph
from tests import ledger_samples as S
INPUTS = {"start": "2026-09-01", "end": "2026-09-08"}
@pytest.fixture(autouse=True)
async def engine_up(tmp_path):
await run_manager.setup()
await S.fresh_source(tmp_path / "shop.db")
yield
await run_manager.shutdown()
# ==========================================================================
# 纯函数:重复的认定
# ==========================================================================
def test_duplicate_metric_ids_ignores_blank_and_reports_each_id_once():
defs = [{"id": "a"}, {"id": "b"}, {"id": "a"}, {"id": " a "}, {"id": ""}, {"x": 1}, {"id": "b"}, {"id": "b"}]
assert duplicate_metric_ids(defs) == ["a", "b"]
# 空 ID / 不是字典 / 没有 id 的都不算重复(另有报法)
assert duplicate_metric_ids([{"id": ""}, {"id": " "}, {"x": 1}, {}, "nope"]) == []
assert duplicate_metric_ids([]) == []
def test_require_selection_refuses_duplicates_instead_of_taking_the_first():
"""闭合解析对同名只留第一次;运行时不能顺着它算第一条——要在过滤前就响亮失败。"""
defs = [{"id": "gmv", "expression": "cell(nodes.fetch, 0, 'gmv')"},
{"id": "gmv", "expression": "cell(nodes.fetch, 0, 'other')"}]
with pytest.raises(SelectionRefused, match="重复"):
require_selection(defs, ["gmv"])
def test_closure_and_runtime_share_the_duplicate_verdict():
"""闭包把重复记成未决、运行时响亮失败:判定只有一处(ledger.closure.duplicate_metric_ids),不冲突。"""
from app.ledger import closure
graph = S.chain(S.node("start", "input"),
S.node("fetch", "tool", tool=S.tool(), args={"sql": S.TEMPLATE}),
S.node("card", "metrics", metrics=[{"id": "gmv", "expression": "cell(nodes.fetch, 0, 'gmv')"},
{"id": "gmv", "expression": "cell(nodes.fetch, 0, 'other')"}]))
metrics = graph["nodes"][2]["data"]["config"]["metrics"]
assert closure.DUPLICATE_METRIC_ID in closure.CODES
assert closure.duplicate_metric_ids(metrics) == duplicate_metric_ids(metrics) == ["gmv"]
# 闭包:这张卡本身未决;运行时:过滤前就拒(绝不取第一条)
assert closure.DUPLICATE_METRIC_ID in [p["code"] for p in closure.build(graph, "wf-dup", 1, "card").unresolved]
with pytest.raises(SelectionRefused, match="重复"):
require_selection(metrics, ["gmv"])
# ==========================================================================
# 画布 / 发布前检查:schema 报错
# ==========================================================================
def duplicate_card(*, with_caliber_from: bool = False) -> dict:
metrics = [{"id": "gmv", "name": "销售额", "unit": "元", "expression": "cell(nodes.fetch, 0, 'gmv')"},
{"id": "gmv", "name": "销售额(重复)", "unit": "元",
"expression": "cell(nodes.fetch, 0, 'gmv')"}]
card_config: dict = {"metrics": metrics}
if with_caliber_from:
card_config["caliber_from"] = {"workflow_id": "wf", "workflow_version": 1, "node_id": "card"}
return S.chain(S.node("start", "input", fields=[{"name": "start", "required": True}, {"name": "end", "required": True}]),
S.node("fetch", "tool", tool=S.tool(), args={"sql": S.TEMPLATE}),
S.node("card", "metrics", caliber="重复口径", caliber_version="v1", **card_config),
S.node("out", "output", fields=[{"name": "r", "value": "{{ nodes.card.text }}"}]))
def issues(graph: dict) -> list:
return list(validate_graph(GraphSpec.model_validate(graph)).issues)
def test_schema_rejects_a_duplicate_metric_id():
found = [i for i in issues(duplicate_card()) if i.code == "metrics.id_duplicate"]
assert found and [i.level for i in found] == ["error"]
assert found[0].node_id == "card" and "gmv" in found[0].message
assert any("唯一" in i.message for i in found)
def test_schema_flags_duplicates_even_on_a_caliber_from_card():
"""转抄卡本地写的 metrics 不生效,但同一张卡上的 ID 照样要唯一(现状是给一条「不会生效」的警告)。"""
found = [i for i in issues(duplicate_card(with_caliber_from=True)) if i.code == "metrics.id_duplicate"]
assert found
def test_schema_leaves_unique_ids_and_blank_ids_alone():
unique = S.chain(S.node("start", "input"), S.node("fetch", "tool", tool=S.tool(), args={"sql": S.TEMPLATE}),
S.node("card", "metrics", metrics=[
{"id": "a", "name": "A", "expression": "cell(nodes.fetch, 0, 'gmv')"},
{"id": "b", "name": "B", "expression": "cell(nodes.fetch, 0, 'gmv')"}]))
assert not [i for i in issues(unique) if i.code == "metrics.id_duplicate"]
# 两条都没有 ID:是「缺少 ID」,不是「重复」
blank = S.chain(S.node("start", "input"), S.node("fetch", "tool", tool=S.tool(), args={"sql": S.TEMPLATE}),
S.node("card", "metrics", metrics=[
{"id": "", "name": "A", "expression": "cell(nodes.fetch, 0, 'gmv')"},
{"name": "B", "expression": "cell(nodes.fetch, 0, 'gmv')"}]))
codes = [i.code for i in issues(blank)]
assert "metrics.id_missing" in codes and "metrics.id_duplicate" not in codes
# ==========================================================================
# 运行时:源卡读取(本地卡 / 转抄卡)也响亮失败
# ==========================================================================
async def run(graph: dict):
started = await run_manager.start(graph=graph, input_payload=dict(INPUTS))
return await S.wait(started.id)
async def upstream_with_duplicates() -> str:
graph = S.chain(
S.node("start", "input", fields=[{"name": "start", "required": True}, {"name": "end", "required": True}]),
S.node("fetch", "tool", tool=S.tool(), args={"sql": S.TEMPLATE}),
S.node("card", "metrics", caliber="重复口径", caliber_version="v1", metrics=[
{"id": "gmv", "name": "销售额", "unit": "元", "expression": "cell(nodes.fetch, 0, 'gmv')"},
{"id": "gmv", "name": "销售额(重复)", "unit": "元", "expression": "cell(nodes.fetch, 0, 'gmv')"}]))
return await S.publish(graph=graph, name="K08 重复上游")
def consumer(workflow_id: str, *, selection: list[str] | None = None) -> dict:
ref: dict = {"workflow_id": workflow_id, "workflow_version": 1, "node_id": "card"}
if selection is not None:
ref["metrics"] = list(selection)
return S.chain(
S.node("start", "input", fields=[{"name": "start", "required": True}, {"name": "end", "required": True}]),
S.node("fetch", "tool", tool=S.tool(), args={"sql": S.TEMPLATE}),
S.node("card", "metrics", caliber_from=ref),
S.node("out", "output", fields=[{"name": "r", "value": "{{ nodes.card.text }}"}]))
async def test_runtime_refuses_duplicate_ids_on_a_pinned_card():
"""钉住的那一版卡上 ID 重复:整卡读取也要响亮失败,不能取第一条静默执行。"""
workflow_id = await upstream_with_duplicates()
row = await run(consumer(workflow_id))
assert row.status == "failed" and "重复" in str(row.error)
assert "gmv" in str(row.error)
async def test_runtime_refuses_a_selection_over_a_duplicate_card():
"""带选择点名那个重复的 ID:闭合解析只留第一次,运行时不能顺着它算第一条——在过滤前就拦住。"""
workflow_id = await upstream_with_duplicates()
row = await run(consumer(workflow_id, selection=["gmv"]))
assert row.status == "failed" and "重复" in str(row.error)
assert "gmv" in str(row.error)
async def test_runtime_leaves_a_unique_card_alone():
row = await run(consumer(await S.publish(name="K08 唯一上游")))
assert row.status == "succeeded", row.error
async def test_publish_precheck_blocks_a_duplicate_local_card():
"""本地重复 ID 的卡过不了画布 / 发布前检查:run_manager.start 直接拒,不会跑到运行时。"""
with pytest.raises(ValueError, match="重复"):
await run_manager.start(graph=duplicate_card(), input_payload=dict(INPUTS))