Notes | numpy-10 Iterative array

#%% md
NumPy  Iterator object  numpy.nditer  Provides a flexible way to access one or more array elements .
 The basic task of iterator is to access array elements .
 So let's use  arange()  Function to create a  2X3  Array , And use  nditer  Iterate over it b
#%%
import numpy as np

a = np.arange(6).reshape(2, 3) # 2r3c
a
#%%
#  Output through iteration 
#  The above example is not a standard  C  perhaps  Fortran  The order , The order of selection is consistent with the array memory layout , This is done to improve the efficiency of access , The default is row order first （row-major order, Or rather,  C-order）
for _ in np.nditer(a): #  First floor for You can output the entire array  |  Output line by line ( level )
    print(_, end=', ')
#%%
a.T #  reverse 
#%%
#  It reflects that by default, you only need to access each element , Regardless of their specific order . We can see this by iterating the transpose of the above array , And with  C  Sequential access array transposed  copy  Compare the ways 
# a  and  a.T  The traversal order is the same , That is, they are stored in the same order in memory 
for _ in np.nditer(a.T): #  First floor for You can output the entire array  |  have no feelings   The order of output is the same  |  Using or  a  This iterates 
    print(_, end=', ')
#%%
#  however  a.T.copy(order = 'C')  The traversal results are different , That's because it's different from the first two storage methods , The default is to access by line 
for _ in np.nditer(a.T.copy(order='C')): #  First floor for You can output the entire array  |  The order of output is different  |  This is output line by line ( level )  This is using replication ( Copy sort changes )
    print(_, end=', ')
#%%
#  Control traversal order 
# for x in np.nditer(a, order='F'):Fortran order, That is, sequence first 
# for x in np.nditer(a.T, order='C'):C order, That is, line order takes precedence 
#%%
import numpy as np

a = np.arange(0, 60, 5) # 0~60( Not included 60)  step  5
a
#%%
a = a.reshape(3, 4) # 3r4c
a
#%%
b = a.T
b # 4r3c
#%%
for _ in np.nditer(b):
    print(_, end=', ') #  vertical ( Column )
#%%
c = b.copy(order='C')
c
#%%
for _ in np.nditer(c):
    print(_, end=', ') #  level ( That's ok )
#%%
c = b.copy(order='F')
c
#%%
for _ in np.nditer(c):
    print(_, end=', ') #  vertical ( Column )
#%%
#  You can explicitly set , Mandatory  nditer  Objects use some sort of order 
for _ in np.nditer(a, order='C'):
    print(_, end=', ') #  level ( That's ok )
#%%
for _ in np.nditer(a, order='F'):
    print(_, end=', ') #  vertical ( Column )
#%% md
##  Change the value of the element in the array 
nditer  Object has another optional parameter  op_flags.  By default ,nditer  The array to be iterated is considered as a read-only object （read-only）, In order to traverse the array at the same time , The implementation of array elements is worth modifying , Must specify  read-write  perhaps  write-only  The pattern of 
#%%
import numpy as np

for _ in np.nditer(a, op_flags=['readwrite']):
    _[...] = 2 * _ #  All elements   multiply 2 operation 

a
#%% md
##  Use an external loop 
#%%
# nditer  Class constructor has  flags  Parameters , It can accept the following values 
"""  Parameters   describe  c_index  Can track  C  Index of order  |  That's ok  f_index  Can track  Fortran  Index of order  |  Column  multi_index  Each iteration can track one index type  external_loop  The given value is a one-dimensional array with multiple values , Instead of a zero dimensional array  """
#%%
for _ in np.nditer(a, flags=['external_loop'], order='F'):
    print(_, end=', ')
#%% md
##  Broadcast iteration 
#%%
#  If two arrays are broadcast ,nditer  Composite objects can iterate over them at the same time .  Hypothetical array  a  The dimensions are  3X4, Array  b  The dimensions are  1X4 , Then use the following iterators （ Array  b  Broadcast to  a  Size ）
import numpy as np

a = np.arange(0, 60, 5).reshape(3, 4)
a
#%%
b = np.array([1, 2, 3, 4], dtype=int)
b
#%%
#  To iterate   This makes it easy to check the calculation method 
for x,y in np.nditer([a, b]):
    print('x = %d, y = %d' % (x, y))
#%%