Indexing rows, columns, elements of numpy.matrix

Question

I don't understand the rules for accessing elements, rows and columns in a numpy.matrix . Here is an example:

import numpy as np
m = np.matrix( [ [ 0, 1, 2 ] ] )
print( "m:", m )
print( "m[0]:", m[ 0 ] )
print( "m[0,]", m[ 0, ] )
print( "m[0,:]", m[ 0, : ] )
print( "m[0,i]", m[ 0, 0 ], m[ 0 , 1 ], m[ 0, 2 ] )

and here's what it prints:

m: [[0 1 2]]
m[0]: [[0 1 2]]
m[0,] [[0 1 2]]
m[0,:] [[0 1 2]]
m[0,i] 0 1 2

It seems that numpy.matrix works differently than numpy.array or Python 2D lists. Can someone point to a succinct explanation of how indexing works for numpy.matrix ? And, is there any more succinct way than the last line of code to extract all the elements of a (1,n) numpy.matrix as if it were a 1D container?

Answer 1

A matrix is always 2d. Row indexing (which most of your examples are doing) still returns a 2d array.

In [1]: m = np.matrix([[ 0,1,2]])                                               
In [2]: m                                                                       
Out[2]: matrix([[0, 1, 2]])
In [3]: m[0]                                                                    
Out[3]: matrix([[0, 1, 2]])      # note the 'matrix' display

indexing an element:

In [4]: m[0,1]                                                                  
Out[4]: 1

matrix as regular array:

In [6]: m.A                                                                     
Out[6]: array([[0, 1, 2]])      # still 2d

as list:

In [7]: m.tolist()                                                              
Out[7]: [[0, 1, 2]]

m as 1d array:

In [8]: m.A1                                                                    
Out[8]: array([0, 1, 2])           # in effect m.A.ravel()

You don't have to use np.matrix , especially if it's confusing. It's in the process of being deprecated.