Python中的 numpy.require()

当返回正确的标志满足传递给编译代码的要求（可能通过 ctypes）时， numpy.require()函数对数组很有用。

Syntax: numpy.require(a, dtype=None, requirements=None)

Parameters:

a :  array_like

dtype : data-type

requirements : str or list of str 

The requirements list can be any of the following.

• ‘F’ : ‘F_CONTIGUOUS’  – ensure a Fortran-contiguous array.
• ‘C’  : ‘C_CONTIGUOUS’ – ensure a C-contiguous array.
• ‘A’ : ‘ALIGNED’  – ensure a data-type aligned array.
• ‘W’ : ‘WRITEABLE’  – ensure a writable array.
• ‘O’ : ‘OWNDATA’ – ensure an array that owns its own data.
• ‘E’ : ‘ENSUREARRAY’ – ensure a base array, instead of a subclass.

Returns :  ndarray

Exception : ValueError – Raises ValueError

代码#1：

Python3

# Python program explaining
# numpy.require() function
  
# importing numpy
import numpy as np
  
# creating 4 x 4 array
data = np.arange(16).reshape(4, 4)
  
data.flags

输出：

C_CONTIGUOUS : True
  F_CONTIGUOUS : False
  OWNDATA : False
  WRITEABLE : True
  ALIGNED : True
  WRITEBACKIFCOPY : False
  UPDATEIFCOPY : False

代码#2：

蟒蛇3

import numpy as np
  
# Python program explaining
# numpy.require()
b = np.require(data, dtype=np.float32,
               requirements=['A', 'W', 'O', 'C'])
b.flags

输出：

C_CONTIGUOUS : True
  F_CONTIGUOUS : False
  OWNDATA : True
  WRITEABLE : True
  ALIGNED : True
  WRITEBACKIFCOPY : False
  UPDATEIFCOPY : False