mat-io 0.1.1

A package for reading MATLAB .mat files, with support for MATLAB datatypes like table and string

Mat-IO Module

The mat-io module provides tools for reading .mat files, particularly for extracting contents from user-defined objects or MATLAB datatypes such as datetime, table and string. It uses a wrapper built around scipy.io to extract raw subsystem data from MAT-files, which is then parsed and interpreted to extract object data. Currently supported MATLAB objects are:

• string
• datetime
• duration
• table
• timetable
• Enumeration Instance Arrays
• User-defined objects

Note: load_from_mat() uses a modified fork of scipy. The fork currently contains a few minor changes to scipy.io to return variable names and object metadata for all objects in a MAT-file. This change is available on Github and can be installed directly from the branch. You can also view the changes under patches/scipy_changes.patch and apply it manually. Note that you might need to rebuild as parts of the Cython code was modified. Follow the instruction on the official SciPy documentation.

Usage

Install using pip

pip install mat-io

Example

To read subsystem data from a .mat file:

from matio import load_from_mat

file_path = "path/to/your/file.mat"
data = load_from_mat(file_path, raw_data=False)
print(data)

Parameters

• file_path: str Full path to the MAT-file.

• raw_data: bool, optional

  • If False (default), returns object data as raw object data
  • If True, converts data into respective Pythonic datatypes (e.g., string, datetime and table).

• mdict: dict, optional Dictionary into which MATLAB variables will be inserted. If None, a new dictionary is created and returned.

• spmatrix: bool, optional (default = True) Whether to read MATLAB sparse matrices as SciPy sparse matrix coo_matrix.

• **kwargs: Additional keyword arguments passed to scipy.io.loadmat. These include:

  • byte_order
  • mat_dtype
  • chars_as_strings
  • matlab_compatible
  • verify_compressed_data_integrity
  • variable_names
  • uint16_codec

MATLAB objects

MATLAB objects are returned as a dictionary with the following fields:

• _Class: The class name
• _Props: A structured numpy.ndarray containing the property names and their contents. Dimensions are determined by the object dimensions.

If the raw_data parameter is set to False, then load_from_mat converts these objects into a corresponding Pythonic datatype. This conversion is detailed here.

Contribution

There's still lots to do! I could use your help in the following:

• Reverse engineer the MAT-file structure to include support for more objects like categorical, calendarDuration and others
• Write object data into MAT-files
• Write tests
• Algorithmic optimization to integrate within the scipy.io framework
• Documentation of the MAT-file format

Feel free to create a PR if you'd like to add something, or open up an issue if you'd like to discuss! I've also opened an issue with scipy.io detailing some of the workflow, as well as a PR to develop this iteratively. Please feel free to contribute there as well!

Acknowledgement

Big thanks to mahalex for their detailed breakdown of MAT-files. A lot of this wouldn't be possible without it.