mat 7.3

Load MATLAB 7.3 .mat files into Python.

Starting with MATLAB 7.3, .mat files have been changed to store as custom hdf5 files. This means they cannot be loaded by scipy.io.loadmat any longer and raise.

NotImplementedError: Please use HDF reader for matlab v7.3 files, e.g. h5py

Quickstart

This library loads MATLAB 7.3 HDF5 files into a Python dictionary.

import mat73
data_dict = mat73.loadmat('data.mat')

As easy as that!

By enabling use_attrdict=True you can even access sub-entries of structs as attributes, just like in MATLAB:

data_dict = mat73.loadmat('data.mat', use_attrdict=True) 
struct = data_dict['structure'] # assuming a structure was saved in the .mat
struct[0].var1 == struct[0]['var1'] # it's the same!

You can also specifiy to only load a specific variable or variable tree, useful to reduce loading times

data_dict = mat73.loadmat('data.mat', only_include='structure') 
struct = data_dict['structure'] # now only structure is loaded and nothing else

data_dict = mat73.loadmat('data.mat', only_include=['var/subvar/subsubvar', 'tree1/']) 
tree1 = data_dict['tree1'] # the entire tree has been loaded, so tree1 is a dict with all subvars of tree1
subsubvar = data_dict['var']['subvar']['subsubvar'] # this subvar has been loaded

Installation

To install, run:

pip install mat73

Alternatively for most recent version:

pip install git+https://github.com/skjerns/mat7.3

Datatypes

The following MATLAB datatypes can be loaded

MATLAB	Python
logical	np.bool_
single	np.float32
double	np.float64
int8/16/32/64	np.int8/16/32/64
uint8/16/32/64	np.uint8/16/32/64
complex	np.complex128
char	str
struct	list of dicts
cell	list of lists
canonical empty	[]
missing	None
sparse	scipy.sparse.csc
Other (ie Datetime, ...)	Not supported

Short-comings

• This library will only load mat 7.3 files. For older versions use scipy.io.loadmat
• Proprietary MATLAB types (e.g datetime, duriation, etc) are not supported. If someone tells me how to convert them, I'll implement that
• For now, you can't save anything back to the .mat. It's a bit more difficult than expected, so it's not on the roadmap for now
• See also hdf5storage, which can indeed be used for saving .mat, but has less features for loading
• See also pymatreader which has a (maybe even better) implementation of loading MAT files, even for older ones