python的各种max方法：max,np.nanmax,np.amax

求最大值是算法中比较常见的问题，python中有多种求最大值的方法，而且对于不同的对象，适用方法不一样。

1 三种不同的对象：pd.Series，np.array，list

不为空的情况

s1 = pd.Series([1, 2, 3, np.nan])
a1 = pd.Series([1, 2, 3, np.nan]).values
l1 = [1, 2, 3, np.nan]
print(type(s1))
print(type(a1))
print(type(l1))
# ===================================================
[Out]:
<class 'pandas.core.series.Series'>
<class 'numpy.ndarray'>
<class 'list'>

为空的情况

s2 = pd.Series([])
a2 = pd.Series([]).values
l2 = []

2 xxx.max()方法

对于非空对象，Series返回最大的数值，array会返回nan，list则会报错

s1_max_1 = s1.max()
print(s1_max_1)
a1_max_1 = a1.max()
print(a1_max_1)
l1_max_1 = l1.max()
print(l1_max_1)
# ===================================================
[Out]:
3.0
nan
AttributeError: 'list' object has no attribute 'max'

对于空对象，Series返回nan，array和list则会报错

s2_max_1 = s2.max()
print(s2_max_1)
a2_max_1 = a2.max()
print(a2_max_1)
l2_max_1 = l2.max()
print(l2_max_1)
# ===================================================
[Out]:
nan
ValueError: zero-size array to reduction operation maximum which has no identity
AttributeError: 'list' object has no attribute 'max'

2 max(xxx)方法

对于非空对象，各类对象都是返回最大的数值

s1_max_2 = max(s1)
print(s1_max_2)
a1_max_2 = max(a1)
print(a1_max_2)
l1_max_2 = max(l1)
print(l1_max_2)
# ===================================================
[Out]:
3.0
3.0
3

对于空对象，各类对象都会报错

s2_max_2 = max(s2)
print(s2_max_2)
a2_max_2 = max(a2)
print(a2_max_2)
l2_max_2 = max(l2)
print(l2_max_2)
# ===================================================
[Out]:
ValueError: max() arg is an empty sequence
ValueError: max() arg is an empty sequence
ValueError: max() arg is an empty sequence

3 np.nanmax(xxx)方法

对于非空对象，各类对象都是返回最大的数值

s1_max_3 = np.nanmax(s1)
print(s1_max_3)
a1_max_3 = np.nanmax(a1)
print(a1_max_3)
l1_max_3 = np.nanmax(l1)
print(l1_max_3)
# ===================================================
[Out]:
3.0
3.0
3.0

对于空对象，各类对象都会报错

s2_max_3 = np.nanmax(s2)
print(s2_max_3)
a2_max_3 = np.nanmax(a2)
print(a2_max_3)
l2_max_3 = np.nanmax(l2)
print(l2_max_3)
# ===================================================
[Out]:
ValueError: zero-size array to reduction operation maximum which has no identity
ValueError: zero-size array to reduction operation fmax which has no identity
ValueError: zero-size array to reduction operation maximum which has no identity

4 np.amax(xxx)方法

对于非空对象，Series返回最大的数值，array和list会返回nan

s1_max_4 = np.amax(s1)
print(s1_max_4)
a1_max_4 = np.amax(a1)
print(a1_max_4)
l1_max_4 = np.amax(l1)
print(l1_max_4)
# ===================================================
[Out]:
3.0
nan
nan

对于空对象，Series返回nan，array和list会报错

s2_max_4 = np.amax(s2)
print(s2_max_4)
a2_max_4 = np.amax(a2)
print(a2_max_4)
l2_max_4 = np.amax(l2)
print(l2_max_4)

# ===================================================
[Out]:
nan
ValueError: zero-size array to reduction operation maximum which has no identity
ValueError: zero-size array to reduction operation maximum which has no identity

5 总结

1. 对于空对象的array和list，各种方法都会报错，所以需要加if条件判断，避免报错；
2. 对于空对象的Series，只能采用.max()和np.amax()，会返回nan；
3. 对于非空对象，建议都转成Series再求最大值，因为各个方法都是准确的；
4. 对于非空的各类对象，max()和np.nanmax()计算的结果都是准确的；
5. 对于非空对象，如果有nan值，.max和np.amax()会返回nan，这是个坑，需要避免；