Access array contents from a .mat file loaded using Scipy.io.loadmat - python

Question

UPDATE: This is a long question that boils down to, can someone explain the numpy array class to me? I answered my own question below.

I am working on a project to import data from matlab into a mysql database whose contents will be made available through a django website. I want to use Scipy.io.loadmat to get the information from matlab into a form I can use in python so that I can enter the data into the database with the django api.

My problem is that I cannot work with the data imported by scipy.io.loadmat. It is loaded in the form of several nested arrays and some of the variable names seem to be lost.

Here is the matlab code for a test structure that I have created for a trial:

sensors.time = [0:1:10].';
sensors.sensor1 = {};
sensors.sensor1.source_type = 'flight';                          
sensors.sensor1.source_name = 'flight-2';                       
sensors.sensor1.channels = {};
sensors.sensor1.channels.channel1.name = '1';                    
sensors.sensor1.channels.channel1.local_ori = 'lateral';         
sensors.sensor1.channels.channel1.vehicle_ori = 'axial';         
sensors.sensor1.channels.channel1.signals = {};
sensors.sensor1.channels.channel1.signals.signal1.filtered = 'N';
sensors.sensor1.channels.channel1.signals.signal1.filtered_description = 'none'; 
sensors.sensor1.channels.channel1.signals.signal1.data = sin(sensors.time)+0.1*rand(11,1); 

>> sensors
      time: [11x1 double]
      sensor1: [1x1 struct]
>> sensors.sensor1
      source_type: 'flight'
      source_name: 'flight-2'
      channels: [1x1 struct]
>> sensors.sensor1.channels
      channel1: [1x1 struct]
>> sensors.sensor1.channels.channel1
      name: '1'
      local_ori: 'lateral'
      vehicle_ori: 'axial'
      signals: [1x1 struct]
>> sensors.sensor1.channels.channel1.signals
      signal1: [1x1 struct]
>> sensors.sensor1.channels.channel1.signals.signal1
      filtered: 'N'
      filtered_description: 'none'
      data: [11x1 double]

I can easily visualize this structure as a python dictionary, so it does not seem like this should be such a complicated exercise.

Here is the python code I used to read the file in (eventually I want to read in multiple files):

from scipy
import os, glob

path = 'C:\Users\c\Desktop\import'
for f in glob.glob( os.path.join(path, '*.mat')):
    matfile = scipy.io.loadmat(f, struct_as_record=True)

This is the resulting dictionary from loadmat:

>>> matfile
{'sensors': array([[ ([[0], [1], [2], [3], [4], [5], [6], [7], [8], [9], [10]],[[(array([u'flight'], 
      dtype='<U6'), array([u'flight-2'], 
      dtype='<U8'), array([[ ([[(array([u'1'], 
      dtype='<U1'), array([u'lateral'], 
      dtype='<U7'), array([u'axial'], 
      dtype='<U5'), array([[ ([[(array([u'N'], 
      dtype='<U1'), array([u'none'], 
      dtype='<U4'), array([[ 0.06273465],[ 0.84363597],[ 1.00035443],[ 0.22117587],[-0.68221775],[-0.87761299],[-0.24108487],[ 0.71871452],[ 1.04690773],[ 0.46512366],[-0.51651414]]))]],)]],
      dtype=[('signal1', '|O4')]))]],)]], 
      dtype=[('channel1', '|O4')]))]])]], 
      dtype=[('time', '|O4'), ('sensor1', '|O4')]), '__version__': '1.0', '__header__': 'MATLAB 5.0 MAT-file, Platform: PCWIN64, Created on: Tue Jun 07 18:38:32 2011', '__globals__': []}

The data is all there, but I don't know how to access these class objects. I would like to be able to loop over contents so that I can process, multiple sensors, then multiple channels for each sensor, etc.

Any explanations to help me simplify this data structure or suggested changes to make this easier would be greatly appreciated.

---

Update, based on Nick's suggestion here is the repr(matfile) and the dir(matfile)

>>> repr(matfile)
"{'sensors': array([[ ([[0], [1], [2], [3], [4], [5], [6], [7], [8], [9], [10]], [[(array([u'flight'], \n      dtype='<U6'), array([u'flight-2'], \n      dtype='<U8'), array([[ ([[(array([u'1'], \n      dtype='<U1'), array([u'lateral'], \n      dtype='<U7'), array([u'axial'], \n      dtype='<U5'), array([[ ([[(array([u'N'], \n      dtype='<U1'), array([u'none'], \n      dtype='<U4'), array([[ 0.0248629 ],\n       [ 0.88663486],\n       [ 0.93206871],\n       [ 0.22156497],\n       [-0.65819207],\n       [-0.95592508],\n       [-0.22584908],\n       [ 0.66569432],\n       [ 1.06956739],\n       [ 0.51103298],\n       [-0.53732649]]))]], [[(array([u'Y'], \n      dtype='<U1'), array([u'1. 5 Hz High Pass, 2. remove offset'], \n      dtype='<U35'), array([[ 0.        ],\n       [ 0.84147098],\n       [ 0.90929743],\n       [ 0.14112001],\n       [-0.7568025 ],\n       [-0.95892427],\n       [-0.2794155 ],\n       [ 0.6569866 ],\n       [ 0.98935825],\n       [ 0.41211849],\n       [-0.54402111]]))]])]], \n      dtype=[('signal1', '|O4'), ('signal2', '|O4')]))]],)]], \n      dtype=[('channel1', '|O4')]))]])]], \n      dtype=[('time', '|O4'), ('sensor1', '|O4')]), '__version__': '1.0', '__header__': 'MATLAB 5.0 MAT-file, Platform: PCWIN64, Created on: Wed Jun 08 10:58:19 2011', '__globals__': []}"

>>> dir(matfile)
['__class__', '__cmp__', '__contains__', '__delattr__', '__delitem__', '__doc__', '__eq__', '__format__', '__ge__', '__getattribute__', '__getitem__', '__gt__', '__hash__', '__init__', '__iter__', '__le__', '__len__', '__lt__', '__ne__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__setitem__', '__sizeof__', '__str__', '__subclasshook__', 'clear', 'copy', 'fromkeys', 'get', 'has_key', 'items', 'iteritems', 'iterkeys', 'itervalues', 'keys', 'pop', 'popitem', 'setdefault', 'update', 'values', 'viewitems', 'viewkeys', 'viewvalues']

>>> dir(matfile['sensors'])
['T', '__abs__', '__add__', '__and__', '__array__', '__array_finalize__', '__array_interface__', '__array_prepare__', '__array_priority__', '__array_struct__', '__array_wrap__', '__class__', '__contains__', '__copy__', '__deepcopy__', '__delattr__', '__delitem__', '__delslice__', '__div__', '__divmod__', '__doc__', '__eq__', '__float__', '__floordiv__', '__format__', '__ge__', '__getattribute__', '__getitem__', '__getslice__', '__gt__', '__hash__', '__hex__', '__iadd__', '__iand__', '__idiv__', '__ifloordiv__', '__ilshift__', '__imod__', '__imul__', '__index__', '__init__', '__int__', '__invert__', '__ior__', '__ipow__', '__irshift__', '__isub__', '__iter__', '__itruediv__', '__ixor__', '__le__', '__len__', '__long__', '__lshift__', '__lt__', '__mod__', '__mul__', '__ne__', '__neg__', '__new__', '__nonzero__', '__oct__', '__or__', '__pos__', '__pow__', '__radd__', '__rand__', '__rdiv__', '__rdivmod__', '__reduce__', '__reduce_ex__', '__repr__', '__rfloordiv__', '__rlshift__', '__rmod__', '__rmul__', '__ror__', '__rpow__', '__rrshift__', '__rshift__', '__rsub__', '__rtruediv__', '__rxor__', '__setattr__', '__setitem__', '__setslice__', '__setstate__', '__sizeof__', '__str__', '__sub__', '__subclasshook__', '__truediv__', '__xor__', 'all', 'any', 'argmax', 'argmin', 'argsort', 'astype', 'base', 'byteswap', 'choose', 'clip', 'compress', 'conj', 'conjugate', 'copy', 'ctypes', 'cumprod', 'cumsum', 'data', 'diagonal', 'dot', 'dtype', 'dump', 'dumps', 'fill', 'flags', 'flat', 'flatten', 'getfield', 'imag', 'item', 'itemset', 'itemsize', 'max', 'mean', 'min', 'nbytes', 'ndim', 'newbyteorder', 'nonzero', 'prod', 'ptp', 'put', 'ravel', 'real', 'repeat', 'reshape', 'resize', 'round', 'searchsorted', 'setfield', 'setflags', 'shape', 'size', 'sort', 'squeeze', 'std', 'strides', 'sum', 'swapaxes', 'take', 'tofile', 'tolist', 'tostring', 'trace', 'transpose', 'var', 'view']

Obviously I need to learn a bit about objects and classes. How can I pull bits of the array and put them into variables. For example:

time = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
source_type = 'flight'
etc.   

Answer 1

I've run into a similar issue with a fairly complex mat file at our company. I'm still getting my head wrapped around the scipy IO module, but here is what we found.

When you access matfile['sensors'] it returns a scipy.io.matlab.mio5_params.mat_struct object, which we can use to access the contents below. When you print it, it looks like a flat array, but you can still access the dict to get at the individual components. So you could run something like this to start accessing the components:

from scipy.io import loadmat
matfile = loadmat('myfile.mat', squeeze_me=True, struct_as_record=False)
matfile['sensors'].sensor1.channels.channel1.name

In your case you want to be able to iterate over the elements in the structure, which you can do if you access the _fieldnames property of the mat_struct object. From there you can just loop over the field names and access them with getattr:

for field in matfile['sensors']._fieldnames:
    # getattr will return the value for the given key
    print getattr(matfile['sensors'], field)

This is at least allowing us to access the deeply nested elements without having to alter our mat files.

Answer 2

The solution I resorted to was to simplify the MATLAB structure. I eliminated nested structures. Each data set resides in a single file and I used python to loop over all the files of a particular type in the specified directory. (http://bogdan.org.ua/2007/08/12/python-iterate-and-read-all-files-in-a-directory-folder.html, if you would like to see an example of that.)

Importing the flat matlab structure results in a dictionary where the matlab variable names are the keys. Strings come in as arrays of shape (1,) --> [ string ], and numbers come in as arrays of shape (N, M) --> [[ numbers ]].

I still have to learn a bit more about the numpy arrays.