[SciPy-user] io.loadmat

[Ryan Krauss]

Any news on this?  It is not a show stopper for me but has forced me
to first process my experimental data on my wife's computer with old
scipy.  I don't want to lose this functionality in new scipy.

Ryan

On 2/1/06, Ryan Krauss <ryanlists at gmail.com> wrote:
> Thanks Travis.  I was able to get the latest SVN and that one now
> loads.  I have one more (I think this will be the last).  Same place:
> http://www.imdl.gatech.edu/ryan/python/
>
> I replaced the other link with a file called matfile2.tar.gz
> It should unpack a mat file called vertical1_11_24_04_lowfreq.mat
>
> This is the format that my experimental setup on campus saves.  It is
> supposed to be mat compliant.  I was able to load this exact same data
> file using old scipy in windows on my wife's computer last night.
>
> With the latest SVN I get the following:
> In [3]: temp=io.loadmat('vertical2_11_24_04_lowfreq.mat')
> ---------------------------------------------------------------------------
> exceptions.ValueError                                Traceback (most
> recent call last)
>
> /home/ryan/thesis/actuator_modeling/torque_testing/data_Nov2004/vertical_data/<ipython
> console>
>
> /usr/lib/python2.4/site-packages/scipy/io/mio.py in loadmat(name,
> dict, appendmat, basename)
>     790         if (M > 1):
>     791             fid.close()
> --> 792             raise ValueError, "Unsupported binary format."
>     793         if (O != 0):
>     794             fid.close()
>
> ValueError: Unsupported binary format.
>
> Just to try and get some more information, I hacked up mio.py starting
> around line 790:
>         M,rest = divmod(header[0],1000)
>         O,rest = divmod(rest,100)
>         P,rest = divmod(rest,10)
>         T = rest
>
>         print('M='+str(M))
>         print('O='+str(O))
>         print('P='+str(P))
>         print('T='+str(T))
>         if (M > 1):
>             print('Bad M')
> #            fid.close()
> #            raise ValueError, "Unsupported binary format."
>         if (O != 0):
>             print('Bad O')
> #            fid.close()
> #            raise ValuError, "Hundreds digit of first integer should be zero."
>
>         if (T not in [0,1]):
>             print('Bad T')
> #            fid.close()
> #            raise ValueError, "Cannot handle sparse matrices, yet."
>
>
> And then I get this:
> In [1]: temp=io.loadmat('vertical2_11_24_04_lowfreq.mat')
> M=0
> O=0
> P=1
> T=0
> M=1818583
> O=9
> P=0
> T=7
> Bad M
> Bad O
> Bad T
> ---------------------------------------------------------------------------
> exceptions.MemoryError                               Traceback (most
> recent call last)
>
> /home/ryan/thesis/actuator_modeling/torque_testing/data_Nov2004/vertical_data/<ipython
> console>
>
> /usr/lib/python2.4/site-packages/scipy/io/mio.py in loadmat(name,
> dict, appendmat, basename)
>     808         storage = {0:'d',1:'f',2:'i',3:'h',4:'H',5:'B'}[P]
>     809
> --> 810         varname = fid.fread(header[-1],'char')[:-1]
>     811         varname = varname.tostring()
>     812         defnames.append(varname)
>
> /usr/lib/python2.4/site-packages/scipy/io/mio.py in read(self, count,
> stype, rtype, bs, c_is_b)
>     283         if count == 0:
>     284             return zeros(0,rtype)
> --> 285         retval = numpyio.fread(self, count, stype, rtype, bs)
>     286         if len(retval) == 1:
>     287             retval = retval[0]
>
> MemoryError:
>
> So that it looks like it is passing through with valid values for
> M,T,O,and P once and then getting bad values on a second pass.  I
> don't know what any of this means.
>
> I believe this is a matlab v4 file.  Old scipy and octave can both open it.
>
> Thanks,
>
> Ryan
>
> On 2/1/06, Travis Oliphant <oliphant.travis at ieee.org> wrote:
> > Ryan Krauss wrote:
> >
> > >I am having trouble loading Matlab .mat files that loaded just fine
> > >under the old scipy.  Here is the error message:
> > >
> > >
> > Hi Ryan.  With latest SVN of scipy, I'm able to load your matfile just
> > fine.
> >
> > I'm not sure what the issue was, but I did change the getsize_type
> > function in mio.py to make sure that the second argument returned is
> > always a character.   There was a chance it could be returned as a
> > type-object.
> >
> > If  you can't get the latest SVN version, let me know what version of
> > numpy you are using and I can give you a couple-line edit to the mio.py
> > file that should work.
> >
> > -Travis
> >
> >
> >
> > _______________________________________________
> > SciPy-user mailing list
> > SciPy-user at scipy.net
> > http://www.scipy.net/mailman/listinfo/scipy-user
> >
>