🐛 fix ε accuracy

mbcp/mp_math/const.py

@@ -15,5 +15,5 @@ PI = math.pi
 E = math.e
 GOLDEN_RATIO = (1 + math.sqrt(5)) / 2
 GAMMA = 0.57721566490153286060651209008240243104215933593992
-EPSILON = 1e-8
+EPSILON = 0.0000000000001
 

mbcp/mp_math/equation.py

@@ -63,10 +63,11 @@ def get_partial_derivative_func(func: MultiVarsFunc, var: int | tuple[int, ...],
             return (func(*args_list_plus) - func(*args_list_minus)) / (2 * epsilon)
         return partial_derivative_func
     elif isinstance(var, tuple):
-        for i in var:
-            print("测试第i个变量：", i)
-            func = get_partial_derivative_func(func, i, epsilon)
-            print("测试第i个变量的偏导：", func(1, 2))
-        return func
+        def high_order_partial_derivative_func(*args: Var) -> Var:
+            result_func = func
+            for v in var:
+                result_func = get_partial_derivative_func(result_func, v, epsilon)
+            return result_func(*args)
+        return high_order_partial_derivative_func
     else:
         raise ValueError("Invalid var type")

tests/test_partial_derivative.py

@@ -6,6 +6,7 @@
 import logging
 
 from mbcp.mp_math.mp_math_typing import RealNumber
+from mbcp.mp_math.utils import Approx
 
 
 def three_var_func(x: RealNumber, y: RealNumber) -> RealNumber:
@@ -26,6 +27,7 @@ class TestPartialDerivative:
         def df_dx(x, y):
             """原函数关于x的偏导"""
             return 3 * (x ** 2) * (y ** 2) - 3 * (y ** 3) - y
+
         logging.info(f"Expected: {df_dx(1, 2)}, Actual: {partial_derivative_func(1, 2)}")
         assert Approx(partial_derivative_func(1, 2)) == df_dx(1, 2)
 
@@ -40,6 +42,7 @@ class TestPartialDerivative:
         def df_dy(x, y):
             """原函数关于y的偏导"""
             return 2 * (x ** 3) * y - 9 * x * (y ** 2) - x
+
         logging.info(f"Expected: {df_dy(1, 2)}, Actual: {partial_derivative_func(1, 2)}")
         assert Approx(partial_derivative_func(1, 2)) == df_dy(1, 2)
 
@@ -54,6 +57,7 @@ class TestPartialDerivative:
         def df_dxdy(x, y):
             """原函数关于y和x的偏导"""
             return 6 * x ** 2 * y - 9 * y ** 2 - 1
+
         logging.info(f"Expected: {df_dxdy(1, 2)}, Actual: {partial_derivative_func(1, 2)}")
         assert Approx(partial_derivative_func(1, 2)) == df_dxdy(1, 2)
 
@@ -68,6 +72,7 @@ class TestPartialDerivative:
         def df_dydx(x, y):
             """原函数关于x和y的偏导"""
             return 6 * x * y ** 2
+
         logging.info(f"Expected: {df_dydx(1, 2)}, Actual: {partial_derivative_func(1, 2)}")
         assert Approx(partial_derivative_func(1, 2)) == df_dydx(1, 2)
 
@@ -82,5 +87,15 @@ class TestPartialDerivative:
         def d3f_dx3(x, y):
             """原函数关于x的三阶偏导"""
             return 6 * (y ** 2)
+
         logging.info(f"Expected: {d3f_dx3(1, 2)}, Actual: {partial_derivative_func(1, 2)}")
         assert Approx(partial_derivative_func(1, 2)) == d3f_dx3(1, 2)
+
+    def test_possible_error(self):
+        from mbcp.mp_math.equation import get_partial_derivative_func
+        def two_vars_func(x: RealNumber, y: RealNumber) -> RealNumber:
+            return x ** 2 * y ** 2
+
+        partial_func = get_partial_derivative_func(two_vars_func, 0)
+        partial_func_2 = get_partial_derivative_func(two_vars_func, (0, 0))
+        assert Approx(partial_func_2(1, 2)) == 8