Error in converting ndarray to double

When I write in MATLAB,

y = py.numpy.array(1:10);

I receive ndarray but I am unable to convert it into double for further use like following line

x=2*double(y)

Error using double

Conversion to double from py.numpy.ndarray is not possible.

[Ameer Hamza]

Which MATLAB release are you using?

[Sajid Sarwar]

MATLAB 2017a

[Suyash Sule]

I am experiencing the same error on Matlab 2020a as well. A few online answers suggest that double() should work but it doesnt. double(np.array.array()) also does not work. Please help.

Answer 1

[Michael Loibl]

I actually have the same problem. Now, I found a nasty (but in my case helpful) solution. I am using the fact that I am able to convert Python lists to Matlab. I did the following type conversions:

y_py=py.numpy.array(1:10);
y_py_list=y.tolist();
y_mat_cell=cell(y_py_list);
y_mat=cell2mat(y_mat_cell);

Nevertheless, it is also unclear to me, why the conversion seems to work that easily for other people. Therefore, I would be happy to see further more sofisticated and more general answers.

[Florian Werner]

To convert two-dimensional numpy-arrays with the same nasty solution, you can use the following code:

npArray = py.numpy.random.rand(3,3)
pyList = npArray.tolist();
% List to cell
matCell = cell(pyList)';
for i = 1:length(matCell)
    if isa(matCell{i}, 'py.list')
        matCell{i} = cell2mat(cell(matCell{i}));
    end
end
mat = cell2mat(matCell);

[Lucademicus]

Suddenly my code stopped working, while it worked yesterday. I get the same error

> Error using double

> Conversion to double from py.numpy.ndarray is not possible.

I'm in R2023b so this mystery is still there.

[Lucademicus]

A restart of MATLAB made it work again. Not sure what happened, but perhaps this helps anybody

[Petri M]

Matlab version 2025b and Python 3.13, having the same problem: double(...) does not work.

But this tolist conversion works, so still thanks to Michael !

Answer 2

[Chris H]

I experienced exactly this issue when running MATLAB R2023a on Linux (RHEL 8.8), when using a python3.9 install via miniconda.

I do not know why using the python 3.9 install from miniconda created this issue, but when I deactivated the conda environment (which I had set to my default environment via my bashrc) and installed python39 and python39-numpy via yum/dnf, suddenly everything started working. You can keep your default conda environment outside of MATLAB and then within MATLAB select the python3.9 installed at the OS level via the command:

pyenv(Version='/usr/bin/python3.9')

(so long as python is installed at that location)