Extract array from object in Python

Question

I have a variable named mesh of type object extracted from a MATLAB .mat file. (EDIT: Reproducible example below)

In [1]: mesh
Out[1]: 
array([[ array([[ (array([[ 89, 108]], dtype=uint8), 
         array([[-131.659809],
          [-131.659809],
          [-131.659809],
          ..., 
          [  52.022239],
          [  52.022239],
          [  52.022239]]), 
         array([[ 189.358345],
          [ 187.271049],
          [ 185.183753],
          ..., 
          [ -29.807736],
          [ -31.895032],
          [ -33.982328]]))]],
         dtype=[('dim', 'O'), ('x', 'O'), ('y', 'O')])]], dtype=object)

How can I access the individual arrays dim, x and y?

Reproducible example:

I am unable to assign dtype=uint8 to an array as in the imported object.

Also, the imported object has object size (1,1) while this example has object size (1,1,1,1).

import numpy as np

mesh = np.array([[ np.array([[ (np.array([[ 6, 6]], dtype=float), 
     np.array([[-131.659809],
      [-131.659809],
      [-131.659809], 
      [  52.022239],
      [  52.022239],
      [  52.022239]]), 
     np.array([[ 189.358345],
      [ 187.271049],
      [ 185.183753],
      [ -29.807736],
      [ -31.895032],
      [ -33.982328]]))]],
     dtype=[('dim', 'O'), ('x', 'O'), ('y', 'O')])]], dtype=object)

Please provide a reproducible example, check out minimal reproducible example — juanpa.arrivillaga
– juanpa.arrivillaga, Commented Jan 22, 2018 at 8:43
You should be able to access it like mesh['dim'] and mesh['x']... there seems to be a needless wrapping of an object type array, i.e., and array of other arrays. so to really dig down, you need to index into that, so for example mesh['dim'][0] should return your actual array of dimensions, e.g. array([[7, 6]]) — juanpa.arrivillaga
– juanpa.arrivillaga, Commented Jan 22, 2018 at 9:08
@juanpa.arrivillaga Yes, I agree and I can see that there is unnecessary wrapping. But that is the way I am getting the imported data from the .mat file by using scipy.io.loadmat. — nish-ant
– nish-ant, Commented Jan 22, 2018 at 9:20
@juanpa.arrivillaga Sorry, I changed the example slightly. Actually, I am importing a structure called dataSRC having following content: dataSRC.mesh 1x1 struct; dataSRC.field 1x1 struct; dataSRC.pod 1x1 struct. ` The structure mesh has following content: dataSRC.mesh.dim [89,108]; dataSRC.mesh.x 9612x1 double; dataSRC.mesh.y 9612x1 double — nish-ant
– nish-ant, Commented Jan 22, 2018 at 9:24

JRasmusBm · Accepted Answer · 2018-01-22 09:54:56Z

To access the separate arrays above, one could do the following:

>>> mesh[0][0][0][0][0] # dim
array([[ 6.,  6.]])
>>> mesh[0][0][0][0][1] # x
array([[-131.659809],
   [-131.659809],
   [-131.659809],
   [  52.022239],
   [  52.022239],
   [  52.022239]])
>>> mesh[0][0][0][0][2] # y
array([[ 189.358345],
   [ 187.271049],
   [ 185.183753],
   [ -29.807736],
   [ -31.895032],
   [ -33.982328]])

And to set a value, one could for example do:

>>> mesh[0][0][0][0][0][0][0]=1    
>>> mesh
array([[[[ (array([[ 1.,  6.]]), array([[-131.659809],
   [-131.659809],
   [-131.659809],
   [  52.022239],
   [  52.022239],
   [  52.022239]]), array([[ 189.358345],
   [ 187.271049],
   [ 185.183753],
   [ -29.807736],
   [ -31.895032],
   [ -33.982328]]))]]]], dtype=object)

Disclaimer

This answer is a hack, not a solution. It is not meant to be interpreted as a general answer, but rather is specific to the exact example above. It does not deal with numpy, just with being able to access the specific arrays.

Collectives™ on Stack Overflow

Extract array from object in Python

1 Answer 1

Disclaimer

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Disclaimer

Comments

Your Answer

Sign up or log in

Post as a guest

Related