python - 修改函数内的屏蔽数组的一部分

python - modify part of a masked array inside a function

我需要修改函数内的屏蔽数组的一部分,例如:

import numpy.ma as ma

arr_2d = ma.masked_all((5,5))
arr_3d = ma.masked_all((5,5,5))
arr_3d[0,1] = 5

def func1(arr, val):
    arr[:] = val

看起来很简单,但是......

>>> func1(arr_3d[0], 1)
>>> arr_3d[0]
masked_array(data =
 [[-- -- -- -- --]
 [1.0 1.0 1.0 1.0 1.0]
 [-- -- -- -- --]
 [-- -- -- -- --]
 [-- -- -- -- --]],
             mask =
 [[ True  True  True  True  True]
 [False False False False False]
 [ True  True  True  True  True]
 [ True  True  True  True  True]
 [ True  True  True  True  True]],
       fill_value = 1e+20)

这似乎与始终在数组的一部分上设置共享掩码有关,以便将掩码作为副本传递给函数

我希望除了显式传递掩码、返回数据副本或传递 带有索引的更大数组。

最近的 numpy 中的警告是:

In [738]: func1(A[1],1)
/usr/local/bin/ipython3:2: MaskedArrayFutureWarning: setting an item on 
a masked array which has a shared mask will not copy the mask and also 
change the original mask array in the future.
Check the NumPy 1.11 release notes for more information.

http://docs.scipy.org/doc/numpy/release.html#assigning-to-slices-views-of-maskedarray

Currently a slice of a masked array contains a view of the original data and a copy-on-write view of the mask. Consequently, any changes to the slice’s mask will result in a copy of the original mask being made and that new mask being changed rather than the original.

此操作后,A 的第 1 行仍被屏蔽,但 A[,:].data` 已更改。

In [757]: B=np.ma.masked_all((5))
...
In [759]: B[0]=5     # direct __setitem__ change to B
In [760]: B
Out[760]: 
masked_array(data = [5.0 -- -- -- --],
             mask = [False  True  True  True  True],
       fill_value = 1e+20)
In [761]: func1(B[3:],1)
/usr/local/bin/ipython3:2: MaskedArrayFutureWarning: ....

In [762]: B      # no change to mask
Out[762]: 
masked_array(data = [5.0 -- -- -- --],
             mask = [False  True  True  True  True],
       fill_value = 1e+20)
In [763]: B.data      # but data is changed
Out[763]: array([ 5.,  0.,  0.,  1.,  1.])

A[1,:]=1 是对掩码 __setitem__ 的直接使用,它可以完全负责设置 datamask。在您的函数中 A 是原始视图,通过 A.__getitem__ 调用获得。显然,开发人员担心对此视图掩码的更改是否会影响原始掩码。

我们可能要看看开发者的讨论;警告表明最近发生了一些变化。

============

问题不在于函数中的使用,而在于视图

In [764]: B1=B[3:]
In [765]: B1[:]=2
/usr/local/bin/ipython3:1: MaskedArrayFutureWarning:...
In [766]: B
Out[766]: 
masked_array(data = [5.0 -- -- -- --],
             mask = [False  True  True  True  True],
       fill_value = 1e+20)
In [767]: B.data
Out[767]: array([ 5.,  0.,  0.,  2.,  2.])

警告描述了现在正在发生的事情,可能还会持续一段时间。据说这种做法会改变。

遵循更改说明建议:

In [785]: B1=B[3:]
In [787]: B1._sharedmask
Out[787]: True
In [790]: B1._sharedmask=False
In [791]: B1[:]=4
In [792]: B1
Out[792]: 
masked_array(data = [4.0 4.0],
             mask = [False False],
       fill_value = 1e+20)
In [793]: B     # mask has been changed along with data
Out[793]: 
masked_array(data = [5.0 -- -- 4.0 4.0],
             mask = [False  True  True False False],
       fill_value = 1e+20)

如此变化

 def func1(arr,val):
     arr._sharedmask=False
     arr[:]=val

将停止警告,并修改原数组的掩码。