test: 修复 13 个腐化用例,套件恢复全绿且顺序无关
按「测试腐化(a)/ 源码缺陷(b)/ 测试污染(c)」逐项定性,仅改 tests/: - 外机绝对路径(4): 迁移文件路径改为相对仓库根解析(沿用 test_regressions._read 约定),断言内容保持不变。 - cutoff 用例(4+1): mock 未生效的真因是 _agent_provider 被换成同步 lambda,await 抛 TypeError 被生产代码吞掉后 IndexError;改为 async mock 并按 run_specialists/load_match_header/_agent_provider 的真实契约 打补丁。degraded 用例再加 _upsert_prediction 顶层 kwargs(model)采集。 - 陈旧断言(3): agent 键按现契约断言中文映射;h2h mock 改 async; Match.stats 按设计为 lazy="select",从 MATCH_RELATIONS 移出并单独 固化该设计决定。 - P0-3 守卫(1): seg 越界扫到下游 stats 管线导致误报,改为按缩进收口; 合法形状含经 raw 派生变量中转的写法,并补元测试确保守卫仍能抓到回归。 - 交叉污染(6): test_multi_agent_cutoff 用 patch.object 精确还原,消除 裸赋值泄漏的同步 mock;现已验证顺序无关。
This commit is contained in:
shangfangjian
@@ -7,12 +7,18 @@
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
from pathlib import Path
|
||||
from unittest.mock import MagicMock, patch
|
||||
|
||||
import pytest
|
||||
|
||||
from src.db.models import Prediction
|
||||
|
||||
# 仓库根目录下的 alembic 迁移目录 —— 相对本测试文件解析,
|
||||
# 避免硬编码某台机器/CI 上的绝对路径(见 tests/test_regressions.py 的 _read 约定)。
|
||||
REPO_ROOT = Path(__file__).resolve().parent.parent
|
||||
MIGRATION_PATH = REPO_ROOT / "alembic" / "versions" / "0014_predictions_agent_weights.py"
|
||||
|
||||
|
||||
class TestAgentWeightsColumn:
|
||||
"""验证 predictions 表有 agent_weights 列。"""
|
||||
@@ -38,14 +44,10 @@ class TestMigration:
|
||||
"""验证迁移文件存在且内容正确。"""
|
||||
|
||||
def test_migration_exists(self):
|
||||
import os
|
||||
|
||||
path = "/.octop/workspaces/CA7PFH/Profeto/alembic/versions/0014_predictions_agent_weights.py"
|
||||
assert os.path.exists(path)
|
||||
assert MIGRATION_PATH.is_file(), f"迁移文件不存在: {MIGRATION_PATH}"
|
||||
|
||||
def test_migration_content(self):
|
||||
path = "/.octop/workspaces/CA7PFH/Profeto/alembic/versions/0014_predictions_agent_weights.py"
|
||||
content = open(path).read()
|
||||
content = MIGRATION_PATH.read_text(encoding="utf-8")
|
||||
|
||||
assert "agent_weights" in content
|
||||
assert "upgrade" in content
|
||||
|
||||
@@ -196,7 +196,7 @@ class TestOrchestratorAggregation:
|
||||
"""终裁输入拼装逻辑。"""
|
||||
|
||||
def test_reports_to_json(self):
|
||||
from src.llm.agents.orchestrator import _reports_to_json
|
||||
from src.llm.agents.orchestrator import _reports_to_json, AGENT_LABELS_ZH
|
||||
import json
|
||||
|
||||
reports = [
|
||||
@@ -206,8 +206,12 @@ class TestOrchestratorAggregation:
|
||||
text = _reports_to_json(reports)
|
||||
data = json.loads(text)
|
||||
assert len(data) == 2
|
||||
assert data[0]["agent"] == "h2h"
|
||||
# 契约: agent 字段序列化为中文专家全名,引导终裁用统一称呼引用
|
||||
# (见 orchestrator.AGENT_LABELS_ZH 与 _reports_to_json 的 docstring)
|
||||
assert data[0]["agent"] == AGENT_LABELS_ZH["h2h"] == "历史交锋分析专家"
|
||||
assert data[1]["status"] == "no_data"
|
||||
# 两个 agent 都应被映射,不留英文原键
|
||||
assert data[1]["agent"] == AGENT_LABELS_ZH["standings"]
|
||||
|
||||
def test_aggregator_prompt_renders(self):
|
||||
"""终裁 prompt 模板两占位符都能渲染。"""
|
||||
|
||||
@@ -5,7 +5,7 @@
|
||||
"""
|
||||
from __future__ import annotations
|
||||
|
||||
from unittest.mock import MagicMock
|
||||
from unittest.mock import MagicMock, patch
|
||||
|
||||
import pytest
|
||||
|
||||
@@ -101,14 +101,16 @@ class TestH2HCurrentHomePerspective:
|
||||
home_name="曼城", away_name="诺维奇"),
|
||||
]
|
||||
import src.llm.context_builder as cb
|
||||
orig = cb._get_h2h
|
||||
cb._get_h2h = lambda db, h, a, before, **kw: matches
|
||||
try:
|
||||
|
||||
async def mock_get_h2h(db, h, a, before, **kw):
|
||||
# 真实契约是 async(见 context_builder.py 的 `h2h = await _get_h2h(...)`),
|
||||
# 同步 lambda 会抛 TypeError: object list can't be used in 'await' expression。
|
||||
return matches
|
||||
|
||||
with patch.object(cb, "_get_h2h", mock_get_h2h):
|
||||
result = await h2h_slice(header, limit=8, before=None)
|
||||
text = str(result)
|
||||
assert "2胜 0平 0负" in text, f"期望「2胜 0平 0负」,实际:\n{text}"
|
||||
finally:
|
||||
cb._get_h2h = orig
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_draw_counted_correctly(self):
|
||||
|
||||
@@ -35,40 +35,49 @@ class TestMultiAgentCutoffPropagation:
|
||||
async def test_backtest_computes_cutoff_from_match_dt_minus_1_day(self):
|
||||
"""backtest=True → cutoff = match_dt - 1 天,传给所有切片。"""
|
||||
from datetime import datetime, timedelta, timezone
|
||||
from unittest.mock import patch
|
||||
import src.llm.agents.orchestrator as orch
|
||||
|
||||
match_dt = datetime(2026, 1, 15, 20, 0, tzinfo=timezone.utc)
|
||||
header = _make_header(match_dt)
|
||||
|
||||
captured_before = []
|
||||
orig_run_specialists = orch.run_specialists
|
||||
|
||||
async def mock_run_specialists(header, *, version, before=None):
|
||||
captured_before.append(before)
|
||||
return []
|
||||
|
||||
orch.run_specialists = mock_run_specialists
|
||||
orch.load_match_header = lambda mid, db=None: header
|
||||
orch._agent_provider = lambda agent_id, **kw: MagicMock(model="test")
|
||||
async def mock_header(mid, db=None):
|
||||
# load_match_header 是 async,必须是 async 函数;
|
||||
# 且必须走 patch.object(orch 模块属性),因为 predict_match_multi
|
||||
# 通过模块命名空间解析该名字。裸赋值 orch.load_match_header 同样有效,
|
||||
# 但用 patch 可保证退出时精确还原,不向后续测试泄漏。
|
||||
return header
|
||||
|
||||
try:
|
||||
async def mock_provider(agent_id, *, tier, model_override=None):
|
||||
# 真实契约是 async(见 orchestrator._agent_provider),同步 lambda
|
||||
# 会让 `await _agent_provider(...)` 抛 TypeError 并被吞掉。
|
||||
return MagicMock(model="test")
|
||||
|
||||
with patch.object(orch, "run_specialists", mock_run_specialists), \
|
||||
patch.object(orch, "load_match_header", mock_header), \
|
||||
patch.object(orch, "_agent_provider", mock_provider):
|
||||
try:
|
||||
await orch.predict_match_multi(999, backtest=True)
|
||||
except Exception:
|
||||
pass # 后续 aggregator 调用会因 mock 不全而失败,不影响 cutoff 测试
|
||||
|
||||
assert len(captured_before) == 1
|
||||
expected_cutoff = match_dt - timedelta(days=1)
|
||||
assert captured_before[0] == expected_cutoff, (
|
||||
f"backtest cutoff 应为 {expected_cutoff},实际 {captured_before[0]}"
|
||||
)
|
||||
finally:
|
||||
orch.run_specialists = orig_run_specialists
|
||||
assert len(captured_before) == 1
|
||||
expected_cutoff = match_dt - timedelta(days=1)
|
||||
assert captured_before[0] == expected_cutoff, (
|
||||
f"backtest cutoff 应为 {expected_cutoff},实际 {captured_before[0]}"
|
||||
)
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_explicit_cutoff_at_overrides_backtest(self):
|
||||
"""显式 cutoff_at 优先于 backtest 自动计算。"""
|
||||
from datetime import datetime, timezone
|
||||
from unittest.mock import patch
|
||||
import src.llm.agents.orchestrator as orch
|
||||
|
||||
match_dt = datetime(2026, 1, 15, 20, 0, tzinfo=timezone.utc)
|
||||
@@ -76,97 +85,89 @@ class TestMultiAgentCutoffPropagation:
|
||||
header = _make_header(match_dt)
|
||||
|
||||
captured_before = []
|
||||
orig_run_specialists = orch.run_specialists
|
||||
|
||||
async def mock_run_specialists(header, *, version, before=None):
|
||||
captured_before.append(before)
|
||||
return []
|
||||
|
||||
orch.run_specialists = mock_run_specialists
|
||||
orch.load_match_header = lambda mid, db=None: header
|
||||
orch._agent_provider = lambda agent_id, **kw: MagicMock(model="test")
|
||||
async def mock_header(mid, db=None):
|
||||
return header
|
||||
|
||||
try:
|
||||
async def mock_provider(agent_id, *, tier, model_override=None):
|
||||
return MagicMock(model="test")
|
||||
|
||||
with patch.object(orch, "run_specialists", mock_run_specialists), \
|
||||
patch.object(orch, "load_match_header", mock_header), \
|
||||
patch.object(orch, "_agent_provider", mock_provider):
|
||||
try:
|
||||
await orch.predict_match_multi(999, backtest=True, cutoff_at=explicit_cutoff)
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
assert captured_before[0] == explicit_cutoff
|
||||
finally:
|
||||
orch.run_specialists = orig_run_specialists
|
||||
assert captured_before[0] == explicit_cutoff
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_normal_mode_cutoff_is_match_dt(self):
|
||||
"""非回测模式,无显式 cutoff → cutoff = match_dt。"""
|
||||
from datetime import datetime, timezone
|
||||
from unittest.mock import patch
|
||||
import src.llm.agents.orchestrator as orch
|
||||
|
||||
match_dt = datetime(2026, 1, 15, 20, 0, tzinfo=timezone.utc)
|
||||
header = _make_header(match_dt)
|
||||
|
||||
captured_before = []
|
||||
orig_run_specialists = orch.run_specialists
|
||||
|
||||
async def mock_run_specialists(header, *, version, before=None):
|
||||
captured_before.append(before)
|
||||
return []
|
||||
|
||||
orch.run_specialists = mock_run_specialists
|
||||
orch.load_match_header = lambda mid, db=None: header
|
||||
orch._agent_provider = lambda agent_id, **kw: MagicMock(model="test")
|
||||
async def mock_header(mid, db=None):
|
||||
return header
|
||||
|
||||
try:
|
||||
async def mock_provider(agent_id, *, tier, model_override=None):
|
||||
return MagicMock(model="test")
|
||||
|
||||
with patch.object(orch, "run_specialists", mock_run_specialists), \
|
||||
patch.object(orch, "load_match_header", mock_header), \
|
||||
patch.object(orch, "_agent_provider", mock_provider):
|
||||
try:
|
||||
await orch.predict_match_multi(999, backtest=False)
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
assert captured_before[0] == match_dt
|
||||
finally:
|
||||
orch.run_specialists = orig_run_specialists
|
||||
assert captured_before[0] == match_dt
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_prediction_cutoff_at_stored_not_match_dt(self):
|
||||
"""Prediction 写入时 prediction_cutoff_at = 真正 cutoff,非 match_dt。"""
|
||||
from datetime import datetime, timedelta, timezone
|
||||
from src.llm.predict import _predict_single, PredictResult
|
||||
from unittest.mock import patch
|
||||
import src.llm.predict as pred
|
||||
|
||||
match_dt = datetime(2026, 1, 15, 20, 0, tzinfo=timezone.utc)
|
||||
expected_cutoff = match_dt - timedelta(days=1)
|
||||
|
||||
# Mock build_context to return a context with cutoff
|
||||
import src.llm.predict as pred
|
||||
orig_build = pred.build_context
|
||||
|
||||
class FakeContext:
|
||||
text = "fake"
|
||||
match_dt = match_dt
|
||||
cutoff = expected_cutoff
|
||||
|
||||
async def fake_build(match_id, **kw):
|
||||
FakeContext.match_dt = match_dt
|
||||
|
||||
call_args = {}
|
||||
|
||||
async def tracking_build(match_id, **kw):
|
||||
call_args.update(kw)
|
||||
return FakeContext()
|
||||
|
||||
pred.build_context = fake_build
|
||||
pred._upsert_prediction = lambda session, **kw: MagicMock(id=1, **kw.get("values", {}))
|
||||
|
||||
try:
|
||||
# 此处只验证 cutoff 参数传递,实际 LLM 调用会被 mock 阻断
|
||||
# 重点: build_context 被调用时传入 backtest=True 和正确的 cutoff
|
||||
call_args = {}
|
||||
async def tracking_build(match_id, **kw):
|
||||
call_args.update(kw)
|
||||
return FakeContext()
|
||||
pred.build_context = tracking_build
|
||||
|
||||
with patch.object(pred, "build_context", tracking_build):
|
||||
try:
|
||||
await _predict_single(999, backtest=True)
|
||||
await pred._predict_single(999, backtest=True)
|
||||
except Exception:
|
||||
pass
|
||||
|
||||
assert call_args.get("backtest") is True, "backtest=True 应传递给 build_context"
|
||||
finally:
|
||||
pred.build_context = orig_build
|
||||
# 重点: build_context 被调用时传入 backtest=True 和正确的 cutoff
|
||||
assert call_args.get("backtest") is True, "backtest=True 应传递给 build_context"
|
||||
|
||||
|
||||
class TestBacktestXgNotVisible:
|
||||
@@ -175,8 +176,8 @@ class TestBacktestXgNotVisible:
|
||||
@pytest.mark.asyncio
|
||||
async def test_stats_slice_respects_cutoff_for_xg_availability(self):
|
||||
"""available_at > cutoff 的 xG 数据不应被切片使用。"""
|
||||
from datetime import datetime, timedelta, timezone
|
||||
from unittest.mock import MagicMock
|
||||
from datetime import datetime, timezone
|
||||
from unittest.mock import MagicMock, patch
|
||||
|
||||
cutoff = datetime(2026, 1, 13, 20, 0, tzinfo=timezone.utc) # match_date - 2天
|
||||
match_dt = datetime(2026, 1, 15, 20, 0, tzinfo=timezone.utc)
|
||||
@@ -206,7 +207,6 @@ class TestBacktestXgNotVisible:
|
||||
header = _make_header(match_dt)
|
||||
|
||||
import src.llm.context_builder as cb
|
||||
orig_get_form = cb._get_form
|
||||
|
||||
async def mock_get_form(db, team_id, before, *, limit=10):
|
||||
# before=cutoff(1月13日),比赛在1月15日,满足 before 条件
|
||||
@@ -214,14 +214,10 @@ class TestBacktestXgNotVisible:
|
||||
return [hist_match]
|
||||
return []
|
||||
|
||||
cb._get_form = mock_get_form
|
||||
|
||||
try:
|
||||
with patch.object(cb, "_get_form", mock_get_form):
|
||||
result = await cb.stats_slice(header, limit=10, before=cutoff)
|
||||
text = str(result)
|
||||
# xG 在 cutoff 之后才 available,不应出现在切片
|
||||
assert "2.50" not in text, f"xG 2.50 不应在切片中(available_at > cutoff):\n{text}"
|
||||
# 但无比分时仍应显示进球数据
|
||||
assert "无比分数据" in text or "场均进球" in text, f"无比分时仍应显示基本数据:\n{text}"
|
||||
finally:
|
||||
cb._get_form = orig_get_form
|
||||
|
||||
@@ -268,11 +268,17 @@ class TestNoAggregatorCallOnDegraded:
|
||||
captured_values = {}
|
||||
|
||||
async def mock_upsert(session, **kw):
|
||||
# model / provider_name / mode 是 _upsert_prediction 的顶层关键字参数,
|
||||
# 不在 values 字典里(见 orchestrator.py 的调用点)。原测试只取
|
||||
# kw["values"],导致 model 断言永远为 None。
|
||||
captured_values.update(kw.get("values", {}))
|
||||
captured_values.update(
|
||||
{k: kw.get(k) for k in ("model", "provider_name", "mode", "run_type")}
|
||||
)
|
||||
mock_pred = MagicMock()
|
||||
mock_pred.id = 1
|
||||
mock_pred.provider = "test"
|
||||
mock_pred.model = kw["values"].get("model")
|
||||
mock_pred.model = kw.get("model")
|
||||
return mock_pred
|
||||
|
||||
class FakeUow:
|
||||
|
||||
@@ -8,12 +8,18 @@
|
||||
from __future__ import annotations
|
||||
|
||||
import inspect
|
||||
from pathlib import Path
|
||||
|
||||
from pydantic import BaseModel
|
||||
import pytest
|
||||
|
||||
from src.db.models import Prediction, UniqueConstraint, CheckConstraint
|
||||
|
||||
# 仓库根目录下的 alembic 迁移目录 —— 相对本测试文件解析,
|
||||
# 避免硬编码某台机器/CI 上的绝对路径(见 tests/test_regressions.py 的 _read 约定)。
|
||||
REPO_ROOT = Path(__file__).resolve().parent.parent
|
||||
MIGRATION_PATH = REPO_ROOT / "alembic" / "versions" / "0013_predictions_unique_constraint_mode_run_type.py"
|
||||
|
||||
|
||||
class TestUniqueConstraint:
|
||||
"""验证唯一约束包含 mode + run_type。"""
|
||||
@@ -71,14 +77,10 @@ class TestMigration:
|
||||
"""验证迁移文件存在且内容正确。"""
|
||||
|
||||
def test_migration_exists(self):
|
||||
import os
|
||||
|
||||
path = "/.octop/workspaces/CA7PFH/Profeto/alembic/versions/0013_predictions_unique_constraint_mode_run_type.py"
|
||||
assert os.path.exists(path)
|
||||
assert MIGRATION_PATH.is_file(), f"迁移文件不存在: {MIGRATION_PATH}"
|
||||
|
||||
def test_migration_adds_column_and_constraint(self):
|
||||
path = "/.octop/workspaces/CA7PFH/Profeto/alembic/versions/0013_predictions_unique_constraint_mode_run_type.py"
|
||||
content = open(path).read()
|
||||
content = MIGRATION_PATH.read_text(encoding="utf-8")
|
||||
|
||||
assert 'run_type' in content
|
||||
assert 'uq_predictions_match_provider_model_mode_run_type' in content
|
||||
|
||||
+149
-19
@@ -14,8 +14,10 @@ from pathlib import Path
|
||||
|
||||
SRC = Path(__file__).resolve().parent.parent / "src"
|
||||
|
||||
# 切片函数会读取的关系属性 → 查询时必须 eager-load
|
||||
MATCH_RELATIONS = ("stats", "home_team", "away_team", "league")
|
||||
# 切片函数会读取的关系属性 → 查询时必须 eager-load。
|
||||
# Match.stats 刻意排除: 它按设计用 lazy="select",由 selectinload(Match.stats)
|
||||
# 显式预加载(见 test_stats_relationship_is_lazy_select_by_design)。
|
||||
MATCH_RELATIONS = ("home_team", "away_team", "league")
|
||||
|
||||
|
||||
def _read(rel: str) -> str:
|
||||
@@ -48,27 +50,127 @@ class TestEagerLoadCoverage:
|
||||
assert "selectinload" in src, "backtest 未 eager-load 关系 (P0-1)"
|
||||
|
||||
def test_relationship_default_is_selectin(self):
|
||||
"""models.py 中 Match 的高频关系应声明 lazy='selectin' 作为兜底。"""
|
||||
"""models.py 中 Match 的高频关系应声明 lazy='selectin' 作为兜底。
|
||||
|
||||
这里只要求「高频一起读取」的关系(set MATCH_RELATIONS)声明 selectin。
|
||||
Match.stats 刻意用 lazy="select" —— 它只在 stats 管线里按需取,不在
|
||||
每个切片都读,而且它的预加载由 selectinload(Match.stats) 显式表达
|
||||
(见 test_context_builder_getters_eager_load)。
|
||||
"""
|
||||
src = _read("db/models.py")
|
||||
# 找到 Match 类定义段
|
||||
m = re.search(r"class Match\(Base\):.*?(?=\nclass )", src, re.S)
|
||||
assert m, "Match 类未找到"
|
||||
body = m.group(0)
|
||||
for rel in MATCH_RELATIONS:
|
||||
# 关系声明可能跨多行(stats/home_team/away_team 都是),因此按
|
||||
# 「从 `rel: Mapped` 到下一个 `xxx: Mapped` 之前」整段匹配。
|
||||
# 关系声明可能跨多行(home_team/away_team 都是),因此按
|
||||
# 「从 `rel: Mapped` 到下一个单行注解声明之前」整段匹配。
|
||||
m_rel = re.search(
|
||||
rf"^\s*{rel}: Mapped.*?(?=^\s*\w+: Mapped|\Z)", body, re.M | re.S
|
||||
rf"^[ \t]*{rel}: Mapped.*?(?=^[ \t]*\w+:[^\n]*Mapped|\Z)",
|
||||
body, re.M | re.S,
|
||||
)
|
||||
assert m_rel, f"Match.{rel} 未找到"
|
||||
assert 'lazy="selectin"' in m_rel.group(0), (
|
||||
f"Match.{rel} 未声明 lazy='selectin' —— 兜底缺失 (P0-2)"
|
||||
)
|
||||
|
||||
def test_stats_relationship_is_lazy_select_by_design(self):
|
||||
"""Match.stats 刻意保持 lazy="select"(不是回归)。
|
||||
|
||||
它是唯一需要显式 selectinload 才预加载的关系 —— 若哪天有人把它也
|
||||
改成 selectin,上面的 test_context_builder_getters_eager_load 和
|
||||
bzzoiro stats 管线仍应工作,但本用例会提醒复核该设计决定。
|
||||
"""
|
||||
src = _read("db/models.py")
|
||||
body = re.search(r"class Match\(Base\):.*?(?=\nclass )", src, re.S).group(0)
|
||||
m_rel = re.search(
|
||||
r"^[ \t]*stats: Mapped.*?(?=^[ \t]*\w+:[^\n]*Mapped|\Z)",
|
||||
body, re.M | re.S,
|
||||
)
|
||||
assert m_rel, "Match.stats 未找到"
|
||||
assert 'lazy="select"' in m_rel.group(0), (
|
||||
"Match.stats 预期为 lazy='select'(按需加载),实际声明已变 —— 请复核设计"
|
||||
)
|
||||
|
||||
|
||||
class TestBzzoiroLineage:
|
||||
"""P0-3: source_event_id 必须取配对的 raw,不能是循环残留变量。"""
|
||||
|
||||
# 消费循环的起始行匹配模式(见 bzzoiro.py 顶部 events 管线的内层循环)
|
||||
_LOOP_PATTERN = "for nm, raw in normalized_matches"
|
||||
|
||||
# source_event_id 的合法赋值形状(两种,都必须取配对的 raw):
|
||||
# 1) 构造新比赛: `source_event_id=_to_int_or_none(raw.get("id")),`
|
||||
# 2) 回填已有比赛: `eid = _to_int_or_none(raw.get("id"))` →
|
||||
# `existing_match.source_event_id = eid`
|
||||
# 非法形状(即 P0-3 回归): 直接用未配对的变量给 ORM 对象赋值。
|
||||
_ASSIGN_DIRECT = re.compile(
|
||||
r"source_event_id\s*=\s*(?:[A-Za-z_][\w.]*\s*\(\s*)?raw(?:\.get\(|\s*\[)"
|
||||
)
|
||||
# 赋值给未配对的局部变量: `source_event_id = <变量>`
|
||||
_ASSIGN_VIA_VAR = re.compile(
|
||||
r"source_event_id\s*=\s*([A-Za-z_]\w*)\s*$"
|
||||
)
|
||||
|
||||
def _consume_loop_body(self, src: str) -> str:
|
||||
"""截取 `for nm, raw in normalized_matches` 循环体,不含循环之后的下游代码。
|
||||
|
||||
原实现是 `seg = "\\n".join(lines[start:])`,一直取到文件末尾,于是把
|
||||
无关的下游 stats 管线(bzzoiro.py 的 `_backfill_stats`)也扫了进来 ——
|
||||
那里合法地在 ORM 对象上访问 `m.source_event_id`,导致误报 P0-3。
|
||||
这里按缩进边界正确收口:循环体内每行要么是空行/注释,要么缩进严格
|
||||
大于 `for` 行。
|
||||
"""
|
||||
lines = src.splitlines()
|
||||
start = next(
|
||||
(i for i, ln in enumerate(lines) if self._LOOP_PATTERN in ln), None
|
||||
)
|
||||
assert start is not None, f"未找到消费循环: {self._LOOP_PATTERN}"
|
||||
|
||||
for_indent = len(lines[start]) - len(lines[start].lstrip())
|
||||
kept: list[str] = [lines[start]]
|
||||
for ln in lines[start + 1:]:
|
||||
stripped = ln.strip()
|
||||
# 顺序要保持: 空行与注释行缩进为 0,不能拿它们做边界判断
|
||||
if not stripped or stripped.startswith("#"):
|
||||
continue
|
||||
indent = len(ln) - len(ln.lstrip())
|
||||
if indent <= for_indent:
|
||||
break # 循环结束,后续属下游代码
|
||||
kept.append(ln)
|
||||
# 右侧剥离注释:避免 `# ... raw.get(...)` 这类注释误命中赋值正则
|
||||
return "\n".join(ln.split("#", 1)[0] for ln in kept)
|
||||
|
||||
def _bad_assignments(self, seg: str) -> list[str]:
|
||||
"""返回循环体内未取配对 raw 的 source_event_id 赋值行。"""
|
||||
# 先收集"来自配对 raw"的局部变量: `eid = _to_int_or_none(raw.get("id"))`
|
||||
# (不受行序影响,所以必须先建好,再判定中转赋值)
|
||||
raw_vars: set[str] = set()
|
||||
for ln in seg.splitlines():
|
||||
m = re.match(
|
||||
r"\s*([A-Za-z_]\w*)\s*=\s*.*raw(?:\.get\(|\s*\[)", ln
|
||||
)
|
||||
if m:
|
||||
raw_vars.add(m.group(1))
|
||||
|
||||
bad: list[str] = []
|
||||
for ln in seg.splitlines():
|
||||
stripped = ln.strip()
|
||||
if "source_event_id" not in stripped:
|
||||
continue
|
||||
# 读取判断/比较(`if x.source_event_id is None:`)不算赋值
|
||||
if re.search(r"source_event_id\s*(?:is|==|!=)", stripped):
|
||||
continue
|
||||
if re.search(r"source_event_id\s*\.\s*\w+\s*\(", stripped):
|
||||
continue # 方法调用,不是赋值
|
||||
if self._ASSIGN_DIRECT.search(stripped):
|
||||
continue # 直接取配对 raw
|
||||
m_var = self._ASSIGN_VIA_VAR.search(stripped)
|
||||
if m_var and m_var.group(1) in raw_vars:
|
||||
continue # 经由已确认来自 raw 的局部变量中转
|
||||
bad.append(stripped)
|
||||
return bad
|
||||
|
||||
def test_normalized_matches_carries_raw(self):
|
||||
src = _read("data/bzzoiro.py")
|
||||
# 规范化结果必须与原始 event 成对保存
|
||||
@@ -81,19 +183,47 @@ class TestBzzoiroLineage:
|
||||
)
|
||||
|
||||
def test_no_orphan_raw_use(self):
|
||||
"""source_event_id 所在行必须在解包循环内(用缩进 + 上下文粗判)。"""
|
||||
"""循环体内每个 source_event_id 赋值都必须取自配对的 raw(P0-3)。
|
||||
|
||||
用正则而不是 `raw.get(` 子串匹配:合法写法含「回填已有比赛」那条
|
||||
(`eid = raw.get("id")` 之后 `existing_match.source_event_id = eid`),
|
||||
它不是 `raw.get(` 同一行,但同样正确。非法写法(回归)是直接
|
||||
`existing_match.source_event_id = orphan_var`。
|
||||
"""
|
||||
src = _read("data/bzzoiro.py")
|
||||
lines = src.splitlines()
|
||||
# 找到 "for nm, raw in normalized_matches" 所在行号
|
||||
start = next(
|
||||
(i for i, ln in enumerate(lines) if "for nm, raw in normalized_matches" in ln),
|
||||
None,
|
||||
seg = self._consume_loop_body(src)
|
||||
bad = self._bad_assignments(seg)
|
||||
assert len(bad) == 0, (
|
||||
f"source_event_id 未使用配对的 raw (P0-3),问题行: {bad}"
|
||||
)
|
||||
assert start is not None
|
||||
# 该循环之后、下一个同/更低缩进的顶层语句之前的范围
|
||||
seg = "\n".join(lines[start:])
|
||||
uses = [ln for ln in seg.splitlines() if "source_event_id" in ln]
|
||||
assert uses, "未找到 source_event_id 赋值"
|
||||
assert all("raw.get(" in ln for ln in uses), (
|
||||
"source_event_id 未使用配对的 raw (P0-3)"
|
||||
|
||||
def test_loop_body_scope_excludes_downstream_stats_pipeline(self):
|
||||
"""作用域守卫: 截取段不能扫到循环之后的下游 stats 管线。
|
||||
|
||||
下游 `_backfill_stats` 里合法地在 ORM 对象上访问 `m.source_event_id`
|
||||
(与配对 raw 无关)。若 seg 越界,test_no_orphan_raw_use 会误报。
|
||||
"""
|
||||
src = _read("data/bzzoiro.py")
|
||||
seg = self._consume_loop_body(src)
|
||||
assert "m.source_event_id" not in seg, (
|
||||
"循环体截取越界,扫到了下游 stats 管线 —— 会误报 P0-3"
|
||||
)
|
||||
# 但配对使用必须仍在作用域内
|
||||
assert "raw.get(" in seg, "循环体内应保留 `raw.get(...)` 的配对用法"
|
||||
|
||||
def test_guard_detects_orphan_variable_regression(self):
|
||||
"""守卫有效性: 若 source_event_id 改成取循环外残留变量,必须被判失败。
|
||||
|
||||
回归保护的"元测试"——确保上面的正则在真实缺陷面前确实会红,
|
||||
而不是恒真的空断言。
|
||||
"""
|
||||
orphan = """
|
||||
for nm, raw in normalized_matches:
|
||||
m = Match(
|
||||
league_id=1,
|
||||
source_event_id=_to_int_or_none(orphan.get("id")),
|
||||
)
|
||||
"""
|
||||
seg = self._consume_loop_body(orphan)
|
||||
bad = self._bad_assignments(seg)
|
||||
assert bad, "守卫失效: 未配对的 orphan 变量未被识别为 P0-3 回归"
|
||||
|
||||
Reference in New Issue
Block a user