guzhujushi
6fca82e3fc
口径由用户第二轮改定；Codex 修一处真 bug + 工程侧。 - 真 bug：call_params 写在 try 里、却在 except 被引用，5 种失败输入 5 个全炸 UnboundLocalError （本该返回 success=False 的 Result，变成把异常抛给调用方）。按原意前移到 try 之前，未改口径； 失败时照样带上 n_steps / n_paths / start / random_state - 用户口径：n_steps/n_paths 拆 TypeError/ValueError；加 steps 形状与 NaN 两条校验； 默认步长改 rng.choice([-1.0, 1.0])；params 补 random_state；metrics 改 mean_terminal/std_terminal - 工程：docstring 同步（Returns 的 metrics、Raises 的分工、去掉已删的 start 校验）；+3 条测试 （失败也必须返回带 params 的 Result、NaN 判失败、终点 std ~= sqrt(t)），solve 段 41 条 - 验证：208 收集 / 177 passed / 31 skipped / 0 xfail，ruff 0，pyright 0 - docs：STATUS 决策 54 + 3 条新口径；CHANGELOG；易错点 26
435 lines
16 KiB
Python
435 lines
16 KiB
Python
"""solve 包测试。"""
|
||
|
||
from __future__ import annotations
|
||
|
||
import numpy as np
|
||
import pytest
|
||
|
||
_NOT_IMPLEMENTED = "未实现：见 docs/03-roadmap.md L0/L1"
|
||
|
||
|
||
def _normal_batch(rng, size, **kwargs):
|
||
"""`monte_carlo` 的 sampler 契约：收下实现传来的 `size`，一次抽够。"""
|
||
return rng.normal(0, 1, size)
|
||
|
||
|
||
def test_monte_carlo_integral_converges_to_analytic_value():
|
||
"""∫₀¹ x² dx = 1/3。n=2e5 时误差应 < 0.01。"""
|
||
from mcm_toolkit.solve import monte_carlo_integral
|
||
|
||
r = monte_carlo_integral(lambda X: X[:, 0] ** 2, bounds=[(0, 1)], n=200_000, random_state=0)
|
||
assert r.success
|
||
assert abs(r.metrics["mean"] - 1 / 3) < 0.01 # 体积必须是正的
|
||
assert len(r.data) == 200_000 # data 是 n 个样本各自的贡献
|
||
|
||
|
||
def test_monte_carlo_integral_odd_dimension_sign():
|
||
"""∫∫∫ x dV over [0,1]³ = 1/2 —— 奇数维最容易暴露"体积符号写反"。"""
|
||
from mcm_toolkit.solve import monte_carlo_integral
|
||
|
||
r = monte_carlo_integral(lambda X: X[:, 0], bounds=[(0, 1)] * 3, n=200_000, random_state=0)
|
||
assert abs(r.metrics["mean"] - 0.5) < 0.01
|
||
|
||
|
||
def test_monte_carlo_integral_signed_bounds():
|
||
"""∫₀² x dx = 2 —— 边界带宽度时，体积算错和符号写反都会现形。"""
|
||
from mcm_toolkit.solve import monte_carlo_integral
|
||
|
||
r = monte_carlo_integral(lambda X: X[:, 0], bounds=[(0, 2)], n=200_000, random_state=0)
|
||
assert abs(r.metrics["mean"] - 2.0) < 0.02
|
||
|
||
|
||
def test_monte_carlo_integral_2d_and_params():
|
||
from mcm_toolkit.solve import monte_carlo_integral
|
||
|
||
bounds = [(0, 1), (0, 1)]
|
||
r = monte_carlo_integral(lambda X: X[:, 0] * X[:, 1], bounds=bounds, n=200_000, random_state=0)
|
||
assert abs(r.metrics["mean"] - 0.25) < 0.01
|
||
assert r.name == "monte_carlo_integral"
|
||
assert r.params["dimension"] == 2
|
||
assert r.params["bounds"] == bounds
|
||
|
||
|
||
def test_monte_carlo_integral_rejects_bad_bounds():
|
||
"""四类非法 bounds 都要在入口抛 ValueError：区间反了 / 漏括号 / 空 / 参差不齐。"""
|
||
from mcm_toolkit.solve import monte_carlo_integral
|
||
|
||
for bad in ([(1, 0)], [0, 1], [], [(0, 1), (0, 2, 3)]):
|
||
with pytest.raises(ValueError):
|
||
monte_carlo_integral(lambda X: X[:, 0], bounds=bad, n=100, random_state=0)
|
||
|
||
|
||
def test_monte_carlo_integral_failure_message_stays_failure():
|
||
"""失败时 message 不能被包装层改写成"成功"（易错点 22）。"""
|
||
from mcm_toolkit.solve import monte_carlo_integral
|
||
|
||
r = monte_carlo_integral(lambda X: 1.0, bounds=[(0, 1)], n=100, random_state=0)
|
||
assert r.success is False
|
||
assert "成功" not in r.message
|
||
assert "失败" in r.message
|
||
assert r.params["dimension"] == 1 # 失败也要留下追溯信息
|
||
|
||
|
||
def test_monte_carlo_integral_accepts_numpy_integer_n():
|
||
"""§4.7：`np.int64` 不是 `int` 的子类，注解与运行期都要收（易错点 1）。"""
|
||
from mcm_toolkit.solve import monte_carlo_integral
|
||
|
||
r = monte_carlo_integral(lambda X: X[:, 0], bounds=[(0, 1)], n=np.int64(1000), random_state=0)
|
||
assert r.success
|
||
assert len(r.data) == 1000
|
||
|
||
|
||
def test_monte_carlo_probability_of_standard_normal():
|
||
"""P(Z > 0) = 0.5，n=5000 时误差应 < 0.05。"""
|
||
from mcm_toolkit.solve import monte_carlo_probability
|
||
|
||
r = monte_carlo_probability(lambda z: z > 0, _normal_batch, n=5000, random_state=0)
|
||
assert r.success
|
||
assert abs(r.metrics["probability"] - 0.5) < 0.05
|
||
|
||
|
||
def test_monte_carlo_is_reproducible():
|
||
from mcm_toolkit.solve import monte_carlo_probability
|
||
|
||
a = monte_carlo_probability(lambda z: z > 0, _normal_batch, n=1000, random_state=9)
|
||
b = monte_carlo_probability(lambda z: z > 0, _normal_batch, n=1000, random_state=9)
|
||
assert a.metrics["probability"] == b.metrics["probability"]
|
||
|
||
|
||
def test_monte_carlo_returns_n_results_and_documented_metrics():
|
||
from mcm_toolkit.solve import monte_carlo
|
||
|
||
r = monte_carlo(_normal_batch, lambda x: x**2, n=2000, random_state=0)
|
||
assert r.success
|
||
assert r.data.shape == (2000,)
|
||
assert set(r.metrics) == {"mean", "std", "median", "p05", "p95", "ci_low", "ci_high"}
|
||
|
||
|
||
def test_monte_carlo_calls_simulator_only_once():
|
||
"""`simulator` 只被调用一次 —— 这是"必须向量化"的可执行版本。"""
|
||
from mcm_toolkit.solve import monte_carlo
|
||
|
||
counter = {"calls": 0}
|
||
|
||
def simulator(params):
|
||
counter["calls"] += 1
|
||
return params
|
||
|
||
r = monte_carlo(_normal_batch, simulator, n=1000, random_state=0)
|
||
assert r.success
|
||
assert counter["calls"] == 1
|
||
|
||
|
||
def test_monte_carlo_data_is_reproducible_with_same_seed():
|
||
from mcm_toolkit.solve import monte_carlo
|
||
|
||
a = monte_carlo(_normal_batch, lambda x: x, n=2000, random_state=11)
|
||
b = monte_carlo(_normal_batch, lambda x: x, n=2000, random_state=11)
|
||
assert np.array_equal(a.data, b.data)
|
||
|
||
|
||
def test_monte_carlo_follows_set_seed_when_random_state_omitted():
|
||
from mcm_toolkit.core import set_seed
|
||
from mcm_toolkit.solve import monte_carlo
|
||
|
||
set_seed(5)
|
||
a = monte_carlo(_normal_batch, lambda x: x, n=500)
|
||
set_seed(5)
|
||
b = monte_carlo(_normal_batch, lambda x: x, n=500)
|
||
assert np.array_equal(a.data, b.data)
|
||
|
||
|
||
def test_monte_carlo_estimates_known_expectation():
|
||
"""E[X²] = 1（X ~ N(0, 1)）；n = 2e5 时误差应 < 0.02。"""
|
||
from mcm_toolkit.solve import monte_carlo
|
||
|
||
r = monte_carlo(_normal_batch, lambda x: x**2, n=200_000, random_state=0)
|
||
assert abs(r.metrics["mean"] - 1.0) < 0.02
|
||
|
||
|
||
def test_monte_carlo_docstring_example_runs():
|
||
"""docstring 的 Examples 必须真的能跑通（此前那一版是坏的）。"""
|
||
from mcm_toolkit.solve import monte_carlo
|
||
|
||
draw = lambda rng, size, **k: rng.normal(0, 1, size) # noqa: E731
|
||
r = monte_carlo(draw, lambda x: x**2, n=10000, random_state=0)
|
||
assert r.success
|
||
assert abs(r.metrics["mean"] - 1.0) < 0.1
|
||
|
||
|
||
def test_monte_carlo_accepts_numpy_integer_n():
|
||
"""§4.7：`np.int64` 不是 `int` 的子类，不能漏。"""
|
||
from mcm_toolkit.solve import monte_carlo
|
||
|
||
r = monte_carlo(_normal_batch, lambda x: x, n=np.int64(100), random_state=0)
|
||
assert r.success
|
||
assert r.data.shape == (100,)
|
||
|
||
|
||
def test_monte_carlo_counts_boolean_outcomes():
|
||
"""布尔结果按 0/1 统计 —— 事件概率是 MC 最常见的用法。"""
|
||
from mcm_toolkit.solve import monte_carlo
|
||
|
||
r = monte_carlo(
|
||
lambda rng, size, **k: rng.normal(0, 1, size) > 0,
|
||
lambda x: x,
|
||
n=5000,
|
||
random_state=0,
|
||
)
|
||
assert r.success
|
||
assert r.data.dtype == float
|
||
assert abs(r.metrics["mean"] - 0.5) < 0.05
|
||
|
||
|
||
def test_monte_carlo_rejects_nan_results():
|
||
"""结果里出现 NaN 就判失败，不静默给出一列 nan 指标。"""
|
||
from mcm_toolkit.solve import monte_carlo
|
||
|
||
r = monte_carlo(
|
||
lambda rng, size, **k: np.r_[np.zeros(size - 1), np.nan],
|
||
lambda x: x,
|
||
n=10,
|
||
random_state=0,
|
||
)
|
||
assert not r.success
|
||
assert "NaN" in r.message
|
||
|
||
|
||
def test_monte_carlo_rejects_bad_arguments():
|
||
from mcm_toolkit.solve import monte_carlo
|
||
|
||
with pytest.raises(TypeError):
|
||
monte_carlo(123, lambda x: x) # type: ignore[arg-type] # 就是要传错
|
||
with pytest.raises(TypeError):
|
||
monte_carlo(_normal_batch, 123) # type: ignore[arg-type] # 就是要传错
|
||
with pytest.raises(TypeError):
|
||
monte_carlo(_normal_batch, lambda x: x, n=True)
|
||
with pytest.raises(TypeError):
|
||
monte_carlo(_normal_batch, lambda x: x, n=10.0) # type: ignore[arg-type] # 就是要传错
|
||
with pytest.raises(ValueError):
|
||
monte_carlo(_normal_batch, lambda x: x, n=0)
|
||
|
||
|
||
def test_monte_carlo_probability_returns_documented_metrics():
|
||
from mcm_toolkit.solve import monte_carlo_probability
|
||
|
||
r = monte_carlo_probability(lambda z: z > 0, _normal_batch, n=20000, random_state=0)
|
||
assert r.success
|
||
assert set(r.metrics) == {"probability", "ci_low", "ci_high", "std_error"}
|
||
assert r.data.shape == (20000,)
|
||
assert 0.0 <= r.metrics["ci_low"] <= r.metrics["probability"]
|
||
assert r.metrics["probability"] <= r.metrics["ci_high"] <= 1.0
|
||
|
||
|
||
def test_monte_carlo_probability_std_error_matches_binomial_formula():
|
||
"""`std_error` 就是比例的标准误 √(p(1-p)/n)。"""
|
||
from mcm_toolkit.solve import monte_carlo_probability
|
||
|
||
n = 20000
|
||
r = monte_carlo_probability(lambda z: z > 0, _normal_batch, n=n, random_state=0)
|
||
p = r.metrics["probability"]
|
||
assert np.isclose(r.metrics["std_error"], np.sqrt(p * (1 - p) / n))
|
||
|
||
|
||
def test_monte_carlo_probability_ci_narrows_with_n():
|
||
"""区间宽度随 n 增大而收窄（O(1/√n)）。"""
|
||
from mcm_toolkit.solve import monte_carlo_probability
|
||
|
||
def width(n):
|
||
r = monte_carlo_probability(lambda z: z > 0, _normal_batch, n=n, random_state=0)
|
||
return r.metrics["ci_high"] - r.metrics["ci_low"]
|
||
|
||
assert width(100_000) < width(1_000)
|
||
|
||
|
||
def test_monte_carlo_probability_higher_confidence_is_wider():
|
||
from mcm_toolkit.solve import monte_carlo_probability
|
||
|
||
def width(confidence):
|
||
r = monte_carlo_probability(
|
||
lambda z: z > 0, _normal_batch, n=5000, random_state=0, confidence=confidence
|
||
)
|
||
return r.metrics["ci_high"] - r.metrics["ci_low"]
|
||
|
||
assert width(0.99) > width(0.90)
|
||
|
||
|
||
def test_monte_carlo_probability_ci_matches_statsmodels_wilson():
|
||
"""区间必须对得上独立实现（statsmodels 的 wilson）—— `^` 写成异或那次就靠它拦住。"""
|
||
from statsmodels.stats.proportion import proportion_confint
|
||
|
||
from mcm_toolkit.solve import monte_carlo_probability
|
||
|
||
n = 5000
|
||
r = monte_carlo_probability(lambda z: z > 0, _normal_batch, n=n, random_state=0)
|
||
count = int(r.data.sum())
|
||
lo, hi = proportion_confint(count, n, alpha=0.05, method="wilson")
|
||
assert np.isclose(r.metrics["ci_low"], lo)
|
||
assert np.isclose(r.metrics["ci_high"], hi)
|
||
|
||
|
||
def test_monte_carlo_probability_rejects_bad_confidence():
|
||
from mcm_toolkit.solve import monte_carlo_probability
|
||
|
||
for bad in (0.0, 1.0, -0.5, 1.5):
|
||
with pytest.raises(ValueError):
|
||
monte_carlo_probability(lambda z: z > 0, _normal_batch, n=100, confidence=bad)
|
||
|
||
|
||
def test_random_walk_shape_and_start():
|
||
"""`data` 形状 (n_paths, n_steps + 1)，第 0 列是起点。"""
|
||
from mcm_toolkit.solve import random_walk
|
||
|
||
r = random_walk(n_steps=100, n_paths=5, start=2.5, random_state=0)
|
||
assert r.success
|
||
assert r.data.shape == (5, 101)
|
||
assert np.all(r.data[:, 0] == 2.5)
|
||
|
||
|
||
def test_random_walk_default_steps_are_plus_minus_one():
|
||
"""默认是对称随机游走：每一步都恰好是 ±1。"""
|
||
from mcm_toolkit.solve import random_walk
|
||
|
||
r = random_walk(n_steps=200, n_paths=3, random_state=0)
|
||
assert set(np.unique(np.diff(r.data, axis=1))) == {-1.0, 1.0}
|
||
|
||
|
||
def test_random_walk_is_reproducible():
|
||
from mcm_toolkit.solve import random_walk
|
||
|
||
a = random_walk(n_steps=50, n_paths=4, random_state=7)
|
||
b = random_walk(n_steps=50, n_paths=4, random_state=7)
|
||
assert np.array_equal(a.data, b.data)
|
||
|
||
|
||
def test_random_walk_accepts_custom_step_sampler():
|
||
"""换成高斯增量就是带漂移的高斯游走 —— 建模差异全在 step_sampler 里。"""
|
||
from mcm_toolkit.solve import random_walk
|
||
|
||
r = random_walk(
|
||
n_steps=100,
|
||
n_paths=3,
|
||
step_sampler=lambda rng, size: rng.normal(0.5, 1.0, size),
|
||
random_state=0,
|
||
)
|
||
assert r.success
|
||
assert r.data.shape == (3, 101)
|
||
assert abs(r.data[:, -1].mean() - 50.0) < 20.0 # 100 步 × 漂移 0.5
|
||
|
||
|
||
def test_random_walk_rejects_bad_arguments():
|
||
"""§4.7 的分工：类型不对抛 TypeError，取值不对（<= 0）抛 ValueError。"""
|
||
from mcm_toolkit.solve import random_walk
|
||
|
||
with pytest.raises(TypeError):
|
||
random_walk(n_steps=2.5) # type: ignore[arg-type] # 就是要传错
|
||
with pytest.raises(TypeError):
|
||
random_walk(n_steps=True)
|
||
with pytest.raises(TypeError):
|
||
random_walk(n_steps=10, n_paths=1.5) # type: ignore[arg-type] # 就是要传错
|
||
with pytest.raises(ValueError):
|
||
random_walk(n_steps=0)
|
||
with pytest.raises(ValueError):
|
||
random_walk(n_steps=10, n_paths=0)
|
||
r = random_walk(n_steps=10, start="a") # type: ignore[arg-type] # 就是要传错
|
||
assert r.success is False # 口径待定：入口抛 TypeError 也可以，见 STATUS"下一步"
|
||
|
||
|
||
def test_random_walk_failure_returns_result_not_exception():
|
||
"""失败也要返回带 params 的 Result，不能把异常抛给调用方 —— 包装层引用未赋值变量那次就靠它拦住。"""
|
||
from mcm_toolkit.solve import random_walk
|
||
|
||
def transposed(rng, size):
|
||
return rng.normal(0, 1, (size[1], size[0]))
|
||
|
||
r = random_walk(n_steps=4, n_paths=6, step_sampler=transposed, random_state=0)
|
||
assert r.success is False
|
||
assert "形状" in r.message
|
||
assert r.params["n_paths"] == 6 # 失败也要留下追溯信息
|
||
assert r.data is None
|
||
|
||
|
||
def test_random_walk_rejects_nan_steps():
|
||
"""增量含 NaN 判失败 —— 与 monte_carlo 的 NaN 口径（决策 48）一致。"""
|
||
from mcm_toolkit.solve import random_walk
|
||
|
||
r = random_walk(
|
||
n_steps=4,
|
||
n_paths=2,
|
||
step_sampler=lambda rng, size: np.full(size, np.nan),
|
||
random_state=0,
|
||
)
|
||
assert r.success is False
|
||
assert "NaN" in r.message
|
||
|
||
|
||
def test_random_walk_metrics_are_terminal_positions():
|
||
"""终点位置的标准差按 √t 扩散（t = n_steps）—— 随机游走的核心性质。"""
|
||
from mcm_toolkit.solve import random_walk
|
||
|
||
n_steps = 100
|
||
r = random_walk(n_steps=n_steps, n_paths=2000, start=5.0, random_state=0)
|
||
assert abs(r.metrics["mean_terminal"] - 5.0) < 5.0 # 对称游走，期望还在起点附近
|
||
assert abs(r.metrics["std_terminal"] - np.sqrt(n_steps)) < 5.0 # √t 扩散
|
||
|
||
|
||
def test_random_walk_docstring_example_runs():
|
||
from mcm_toolkit.solve import random_walk
|
||
|
||
r = random_walk(n_steps=100, n_paths=5, random_state=0)
|
||
assert r.data.shape == (5, 101)
|
||
assert np.all(np.abs(np.diff(r.data, axis=1)) == 1)
|
||
|
||
|
||
@pytest.mark.skip(reason=_NOT_IMPLEMENTED)
|
||
def test_markov_steady_state_two_state_chain():
|
||
"""P=[[0.9,0.1],[0.2,0.8]] 的平稳分布是 [2/3, 1/3]。"""
|
||
from mcm_toolkit.solve import markov_steady_state
|
||
|
||
r = markov_steady_state(np.array([[0.9, 0.1], [0.2, 0.8]]))
|
||
assert np.allclose(r.data, [2 / 3, 1 / 3], atol=1e-6)
|
||
|
||
|
||
@pytest.mark.skip(reason=_NOT_IMPLEMENTED)
|
||
def test_knapsack_dp_known_optimum():
|
||
from mcm_toolkit.solve import knapsack_dp
|
||
|
||
r = knapsack_dp(weights=[1, 3, 4, 5], values=[1, 4, 5, 7], capacity=7)
|
||
assert np.isclose(r.metrics["total_value"], 9.0)
|
||
assert np.isclose(r.metrics["total_weight"], 7.0)
|
||
|
||
|
||
@pytest.mark.skip(reason=_NOT_IMPLEMENTED)
|
||
def test_linear_program_basic():
|
||
from mcm_toolkit.solve import linear_program
|
||
|
||
r = linear_program(c=[1, 2], A_ub=[[1, 1]], b_ub=[4], bounds=[(0, None)] * 2)
|
||
assert r.success
|
||
assert np.isclose(r.metrics["objective"], 4.0)
|
||
|
||
|
||
@pytest.mark.skip(reason=_NOT_IMPLEMENTED)
|
||
def test_mm1_metrics_closed_form():
|
||
from mcm_toolkit.solve import mm1_metrics
|
||
|
||
r = mm1_metrics(arrival_rate=2, service_rate=3)
|
||
assert np.isclose(r.metrics["rho"], 2 / 3, rtol=1e-6)
|
||
assert np.isclose(r.metrics["L"], 2.0, rtol=1e-6)
|
||
|
||
|
||
@pytest.mark.skip(reason=_NOT_IMPLEMENTED)
|
||
def test_mm1_rejects_unstable_system():
|
||
"""λ >= μ 时系统不稳定，应 success=False 而不是返回 inf。"""
|
||
from mcm_toolkit.solve import mm1_metrics
|
||
|
||
assert mm1_metrics(arrival_rate=3, service_rate=3).success is False
|
||
|
||
|
||
@pytest.mark.skip(reason=_NOT_IMPLEMENTED)
|
||
def test_simulated_annealing_finds_quadratic_minimum():
|
||
from mcm_toolkit.solve import simulated_annealing
|
||
|
||
f = lambda v: (v[0] - 3.0) ** 2 + (v[1] + 1.0) ** 2 # noqa: E731
|
||
r = simulated_annealing(f, x0=[0.0, 0.0], n_iter=5000, random_state=0)
|
||
assert np.allclose(r.params["x"], [3.0, -1.0], atol=0.1)
|