[tool.ruff]
target-version = "py311"
line-length = 100
extend-exclude = [
    ".tmp_*_venv",
    "backend/.venv",
    "backend/data",
    "frontend",
]

[tool.ruff.lint]
# Correctness rules (E4/E7/E9/F) plus the low-noise hygiene set: I (import
# order), UP (py311-modern syntax), B (bugbear footguns), SIM (simplifiable
# constructs). All four auto-fix mechanically; more opinionated sets (N/ANN/RUF)
# can be enabled incrementally later.
# B008 stays allowed: FastAPI's Depends()/Query() in parameter defaults is the
# framework's own idiom, not the footgun bugbear targets.
select = ["E4", "E7", "E9", "F", "I", "UP", "B", "SIM"]
ignore = ["B008"]

[tool.ruff.lint.per-file-ignores]
# 助手工具/接口的"非法输入必须被拒绝"用例刻意断言任意异常：要测的行为就是
# "拒绝路径存在"，具体抛什么类型由 handler 自己的单测钉住，这里收窄反而会把
# 用例绑死在实现细节上。
"backend/tests/test_assistant_claim_tools.py" = ["B017"]
"backend/tests/test_assistant_tools.py" = ["B017"]
"backend/tests/test_drill.py" = ["B017"]
"backend/tests/test_interview.py" = ["B017"]
"backend/tests/test_pdf_exporter_photo.py" = ["B017"]

[tool.coverage.run]
branch = true
source = ["app"]
# Required for `pytest -n auto`: without it every xdist worker writes the same
# `.coverage` SQLite file and the run dies at teardown with
# "attempt to write a readonly database" -- after all tests have already passed,
# which is the worst possible place to fail. With parallel = true each process
# writes `.coverage.<host>.<pid>` and pytest-cov combines them itself.
parallel = true

[tool.coverage.report]
show_missing = true
skip_covered = true
# Current suite is about 87%; keep a conservative floor so normal refactors have room
# while a large untested regression still fails CI.
fail_under = 80

[tool.coverage.xml]
output = "coverage.xml"