用scipy.io通過mat文件在Python和Matlab/Octave之間進(jìn)行數(shù)據(jù)交換

點(diǎn)擊打開鏈接

http://docs.scipy.org/doc/scipy/reference/tutorial/io.html

如果更喜歡用python或Octave/Matlab,但又想兼而有之, 可以考慮

File IO (scipy.io)

See also

numpy-reference.routines.io(in numpy)

MATLAB files

loadmat(file_name[,?mdict,?appendmat])Load MATLAB file

savemat(file_name,?mdict[,?appendmat,?...])Save a dictionary of names and arrays into a MATLAB-style .mat file.

whosmat(file_name[,?appendmat])List variables inside a MATLAB file

The basic functions

We’ll start by importingscipy.ioand calling itsiofor convenience:

>>>

>>>importscipy.ioassio

If you are using IPython, try tab completing onsio. Among the many options, you will find:

sio.loadmatsio.savematsio.whosmat

These are the high-level functions you will most likely use when working with MATLAB files. You’ll also find:

sio.matlab

This is the package from whichloadmat,savematandwhosmatare imported. Withinsio.matlab, you will find themiomodule This module contains the machinery thatloadmatandsavematuse. From time to time you may find yourself re-using this machinery.

How do I start?

You may have a.matfile that you want to read into Scipy. Or, you want to pass some variables from Scipy / Numpy into MATLAB.

To save us using a MATLAB license, let’s start inOctave. Octave has MATLAB-compatible save and load functions. Start Octave (octaveat the command line for me):

octave:1>a=1:12a=123456789101112octave:2>a=reshape(a,[134])a=ans(:,:,1)=123ans(:,:,2)=456ans(:,:,3)=789ans(:,:,4)=101112octave:3>save-6octave_a.mata% MATLAB 6 compatibleoctave:4>lsoctave_a.matoctave_a.mat

Now, to Python:

>>>

>>>mat_contents=sio.loadmat('octave_a.mat')>>>mat_contents{'a': array([[[? 1.,? 4.,? 7.,? 10.],[? 2.,? 5.,? 8.,? 11.],[? 3.,? 6.,? 9.,? 12.]]]),'__version__': '1.0','__header__': 'MATLAB 5.0 MAT-file, written byOctave 3.6.3, 2013-02-17 21:02:11 UTC','__globals__': []}>>>oct_a=mat_contents['a']>>>oct_aarray([[[? 1.,? 4.,? 7.,? 10.],[? 2.,? 5.,? 8.,? 11.],[? 3.,? 6.,? 9.,? 12.]]])>>>oct_a.shape(1, 3, 4)

Now let’s try the other way round:

>>>

>>>importnumpyasnp>>>vect=np.arange(10)>>>vect.shape(10,)>>>sio.savemat('np_vector.mat',{'vect':vect})

Then back to Octave:

octave:8>loadnp_vector.matoctave:9>vectvect=0123456789octave:10>size(vect)ans=110

If you want to inspect the contents of a MATLAB file without reading the data into memory, use thewhosmatcommand:

>>>

>>>sio.whosmat('octave_a.mat')[('a', (1, 3, 4), 'double')]

whosmatreturns a list of tuples, one for each array (or other object) in the file. Each tuple contains the name, shape and data type of the array.

MATLAB structs

MATLAB structs are a little bit like Python dicts, except the field names must be strings. Any MATLAB object can be a value of a field. As for all objects in MATLAB, structs are in fact arrays of structs, where a single struct is an array of shape (1, 1).

octave:11>my_struct=struct('field1',1,'field2',2)my_struct={field1=1field2=2}octave:12>save-6octave_struct.matmy_struct

We can load this in Python:

>>>

>>>mat_contents=sio.loadmat('octave_struct.mat')>>>mat_contents{'my_struct': array([[([[1.0]], [[2.0]])]],dtype=[('field1', 'O'), ('field2', 'O')]), '__version__': '1.0', '__header__': 'MATLAB 5.0 MAT-file, written by Octave 3.6.3, 2013-02-17 21:23:14 UTC', '__globals__': []}>>>oct_struct=mat_contents['my_struct']>>>oct_struct.shape(1, 1)>>>val=oct_struct[0,0]>>>val([[1.0]], [[2.0]])>>>val['field1']array([[ 1.]])>>>val['field2']array([[ 2.]])>>>val.dtypedtype([('field1', 'O'), ('field2', 'O')])

In versions of Scipy from 0.12.0, MATLAB structs come back as numpy structured arrays, with fields named for the struct fields. You can see the field names in thedtypeoutput above. Note also:

>>>

>>>val=oct_struct[0,0]

and:

octave:13>size(my_struct)ans=11

So, in MATLAB, the struct array must be at least 2D, and we replicate that when we read into Scipy. If you want all length 1 dimensions squeezed out, try this:

>>>

>>>mat_contents=sio.loadmat('octave_struct.mat',squeeze_me=True)>>>oct_struct=mat_contents['my_struct']>>>oct_struct.shape()

Sometimes, it’s more convenient to load the MATLAB structs as python objects rather than numpy structured arrays - it can make the access syntax in python a bit more similar to that in MATLAB. In order to do this, use thestruct_as_record=Falseparameter setting toloadmat.

>>>

>>>mat_contents=sio.loadmat('octave_struct.mat',struct_as_record=False)>>>oct_struct=mat_contents['my_struct']>>>oct_struct[0,0].field1array([[ 1.]])

struct_as_record=Falseworks nicely withsqueeze_me:

>>>

>>>mat_contents=sio.loadmat('octave_struct.mat',struct_as_record=False,squeeze_me=True)>>>oct_struct=mat_contents['my_struct']>>>oct_struct.shape# but no - it's a scalarTraceback (most recent call last):File"", line1, inAttributeError:'mat_struct' object has no attribute 'shape'>>>type(oct_struct)>>>oct_struct.field11.0

Saving struct arrays can be done in various ways. One simple method is to use dicts:

>>>

>>>a_dict={'field1':0.5,'field2':'a string'}>>>sio.savemat('saved_struct.mat',{'a_dict':a_dict})

loaded as:

octave:21>loadsaved_structoctave:22>a_dicta_dict=scalarstructurecontainingthefields:field2=astringfield1=0.50000

You can also save structs back again to MATLAB (or Octave in our case) like this:

>>>

>>>dt=[('f1','f8'),('f2','S10')]>>>arr=np.zeros((2,),dtype=dt)>>>arrarray([(0.0, ''), (0.0, '')],dtype=[('f1', '>>arr[0]['f1']=0.5>>>arr[0]['f2']='python'>>>arr[1]['f1']=99>>>arr[1]['f2']='not perl'>>>sio.savemat('np_struct_arr.mat',{'arr':arr})

MATLAB cell arrays

Cell arrays in MATLAB are rather like python lists, in the sense that the elements in the arrays can contain any type of MATLAB object. In fact they are most similar to numpy object arrays, and that is how we load them into numpy.

octave:14>my_cells={1,[2,3]}my_cells={[1,1]=1[1,2]=23}octave:15>save-6octave_cells.matmy_cells

Back to Python:

>>>

>>>mat_contents=sio.loadmat('octave_cells.mat')>>>oct_cells=mat_contents['my_cells']>>>print(oct_cells.dtype)object>>>val=oct_cells[0,0]>>>valarray([[ 1.]])>>>print(val.dtype)float64

Saving to a MATLAB cell array just involves making a numpy object array:

>>>

>>>obj_arr=np.zeros((2,),dtype=np.object)>>>obj_arr[0]=1>>>obj_arr[1]='a string'>>>obj_arrarray([1, 'a string'], dtype=object)>>>sio.savemat('np_cells.mat',{'obj_arr':obj_arr})

octave:16>loadnp_cells.matoctave:17>obj_arrobj_arr={[1,1]=1[2,1]=astring}

IDL files

readsav(file_name[,?idict,?python_dict,?...])Read an IDL .sav file

Matrix Market files

mminfo(source)Queries the contents of the Matrix Market file ‘filename’ to extract size and storage information.

mmread(source)Reads the contents of a Matrix Market file ‘filename’ into a matrix.

mmwrite(target,?a[,?comment,?field,?precision])Writes the sparse or dense arrayato a Matrix Market formatted file.

Wav sound files (scipy.io.wavfile)

read(filename[,?mmap])Return the sample rate (in samples/sec) and data from a WAV file

write(filename,?rate,?data)Write a numpy array as a WAV file

Arff files (scipy.io.arff)

Module to read ARFF files, which are the standard data format for WEKA.

ARFF is a text file format which support numerical, string and data values. The format can also represent missing data and sparse data.

See theWEKA websitefor more details about arff format and available datasets.

Examples

>>>

>>>fromscipy.ioimportarff>>>fromcStringIOimportStringIO>>>content="""...@relation foo...@attribute width? numeric...@attribute height numeric...@attribute color? {red,green,blue,yellow,black}...@data...5.0,3.25,blue...4.5,3.75,green...3.0,4.00,red...""">>>f=StringIO(content)>>>data,meta=arff.loadarff(f)>>>dataarray([(5.0, 3.25, 'blue'), (4.5, 3.75, 'green'), (3.0, 4.0, 'red')],dtype=[('width', '>>metaDataset: foowidth's type is numericheight's type is numericcolor's type is nominal, range is ('red', 'green', 'blue', 'yellow', 'black')

loadarff(f)Read an arff file.

Netcdf (scipy.io.netcdf)

netcdf_file(filename[,?mode,?mmap,?version])A file object for NetCDF data.

Allows reading of NetCDF files (version ofpupynerepackage)