9

I have imported a Matlab *.mat file using scipy.io and trying to extract the 2D data from it. There are several arrays inside, and when I am trying to get them I got stuck at the last operation.

The data looks like the image below. When I try to index it: IndexError: too many indices for array

I have googled to the point that it looks like a single valued tuple, where the only element is my array. This in principle must be indexable, but it doesn't work. The type(data) returns <class 'numpy.ndarray'>

So the question is: how do I get my 2D array out of this data structure?

    data[0] # Doesn't work.

array in question

Improve this question
4
  • 1
    array[0]? Without telling us what you have tried, it's hard to say where your problem might lie. Commented Jul 3, 2018 at 8:27
  • Could you add an example to generate a similar matrix or at least how you are trying to retrieve the data? What does it outputs if you execute my_array.shape? Commented Jul 3, 2018 at 8:27
  • 1
    If I knew how to generate a similar matrix, I would've known the answer. I have found a way around though, some voodoo coding, I will post the reply asap. Commented Jul 3, 2018 at 8:30
  • Short answer is data.item(). Commented Jun 10, 2022 at 14:23

2 Answers 2

Reset to default
8

A search on loadmat should yield many SO questions that will help you pick apart this result. loadmat has to translate MATLAB objects into Python/numpy approximations.

data = io.loadmat(filename)

should produce a dictionary with some cover keys and various data keys. list(data.keys()) to identify those.

x = data['x']

should match the x variable in the MATLAB workspace. It could be a 2d, order F array, corresponding to a MATLAB matrix.

It could be (n,m) object dtype array, corresponding to a MATLAB cell.

It could be a structured array, where the field names correspond to a MATLAB struct attributes.

In your case it looks like you have a 0d object dtype array. The shape is (), an empty tuple (1d has (n,) shape, 2d has (n,m) shape, etc). You can pull the element out of a () array with:

 y[()]
 y.item()

The [()] looks odd, but it's logical. For a 1d array y[1] can be written as y[(1,)]. For 2d, y[1,2] and y[(1,2)] are the same. The indexing tuple should match the number of dimensions. Hence a () can index a () shape array.

Improve this answer
Sign up to request clarification or add additional context in comments.

1 Comment

This behavior of 0-d arrays numpy is extremely unintuitive (and hard to discover). This could be fixed by changing the IndexError message for 0-D arrays to IndexError: …; Use x[()] or x.item() to extract the value from a 0-D array.
2

After some voodoo coding I have found a funny way to solve this:

The initial data is the zero-dimensional where the only element is the 2D array. The way to get this element out apparently is:

    z = data.item()[()][0]
    print(z)

The final result is below I got my 2D array:

Workable 2D data

Improve this answer

1 Comment

Are you telling yourself that your own answer doesn't work?

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.