A better loadmat for SciPy

In the scipy.io package there is a function loadmat for loading Matlab formated data files.

The loadmat function is not useful for reading Matlab struct arrays, as it loads them into structured NumPy arrays, making it difficult to access the elements by struct field names.

A better solution is to load the struct contents into a Python dict, which is what the following function does:

import numpy as np
import scipy.io


def loadmat(filename):
    """Improved loadmat (replacement for scipy.io.loadmat)
    Ensures correct loading of python dictionaries from mat files.

    Inspired by: https://stackoverflow.com/a/29126361/572908
    """

    def _has_struct(elem):
        """Determine if elem is an array
        and if first array item is a struct
        """
        return isinstance(elem, np.ndarray) and (
            elem.size > 0) and isinstance(
            elem[0], scipy.io.matlab.mio5_params.mat_struct)

    def _check_keys(d):
        """checks if entries in dictionary are mat-objects. If yes
        todict is called to change them to nested dictionaries
        """
        for key in d:
            elem = d[key]
            if isinstance(elem,
                          scipy.io.matlab.mio5_params.mat_struct):
                d[key] = _todict(elem)
            elif _has_struct(elem):
                d[key] = _tolist(elem)
        return d

    def _todict(matobj):
        """A recursive function which constructs from
        matobjects nested dictionaries
        """
        d = {}
        for strg in matobj._fieldnames:
            elem = matobj.__dict__[strg]
            if isinstance(elem,
                          scipy.io.matlab.mio5_params.mat_struct):
                d[strg] = _todict(elem)
            elif _has_struct(elem):
                d[strg] = _tolist(elem)
            else:
                d[strg] = elem
        return d

    def _tolist(ndarray):
        """A recursive function which constructs lists from cellarrays
        (which are loaded as numpy ndarrays), recursing into the
        elements if they contain matobjects.
        """
        elem_list = []
        for sub_elem in ndarray:
            if isinstance(sub_elem,
                          scipy.io.matlab.mio5_params.mat_struct):
                elem_list.append(_todict(sub_elem))
            elif _has_struct(sub_elem):
                elem_list.append(_tolist(sub_elem))
            else:
                elem_list.append(sub_elem)
        return elem_list

    data = scipy.io.loadmat(
        filename, struct_as_record=False, squeeze_me=True)
    return _check_keys(data)

---

Update October 30, 2019
Dieter Werthmüller kindly suggested to include the above functionality in SciPy. Here is his comment on the Github issue on loadmat improvements.

As stated in Dieter's comment, I make the code available under a liberal FOSS license (say, the SciPy BSD 3-Clause license), and hope that someone will rework it into a pull request for inclusion in the SciPy project.

Update May 21, 2020
The code is now in scipy! Huge thanks to Clemens Brunner who authored the pull request that was approved today.