[Scipy-tickets] [SciPy] #1230: scipy.io.matlab.loadmat does not squeeze

SciPy Trac scipy-tickets@scipy....
Mon Jul 12 13:30:33 CDT 2010


#1230: scipy.io.matlab.loadmat does not squeeze
----------------------+-----------------------------------------------------
 Reporter:  tpikonen  |       Owner:  somebody
     Type:  defect    |      Status:  new     
 Priority:  normal    |   Milestone:  0.8.0   
Component:  scipy.io  |     Version:  devel   
 Keywords:  mat-file  |  
----------------------+-----------------------------------------------------
 loadmat, even when called with squeeze_me=1 returns scalars as 0-d arrays,
 which is very annoying. Also chars_as_strings does not do what is
 expected. Example with the git head of scipy:

 octave:1> a = struct("f1", 0.5, "f2", "A String")
 a =
 {
   f1 =  0.50000
   f2 = A String
 }

 octave:2> save -mat testcase.mat a

 Python 2.6.5+ (release26-maint, Jul  6 2010, 12:58:20)
 [GCC 4.4.4] on linux2
 Type "help", "copyright", "credits" or "license" for more information.
 >>> import scipy.io.matlab as mio
 >>> adict = mio.loadmat('testcase.mat', struct_as_record=0, squeeze_me=1,
 chars_as_strings=1)
 >>> asq = adict['a']
 >>> asq.f1
 array(0.5)
 >>> asq.f2
 array(u'A String',
       dtype='<U8')
 >>> scipy.__version__
 '0.9.0.dev'
 >>> numpy.__version__
 '1.4.1'

 What should be (IMHO) returned with these input arguments to loadmat are
 asq.f1 == 0.5 and asq.f2 == 'A String'.

 Investigation into io.matlab code shows that the dtype conversion, which
 uses the type table obtained from miobase.convert_dtypes always results in
 values with dtype.isbuiltin == 0, even for basic types such as floats.
 This in turn breaks mio_utils.squeeze_element.

 Patching convert_dtypes to not change the byteorder with newbyteorder
 makes scalar values read out ok, but probably breaks something else.

-- 
Ticket URL: <http://projects.scipy.org/scipy/ticket/1230>
SciPy <http://www.scipy.org>
SciPy is open-source software for mathematics, science, and engineering.


More information about the Scipy-tickets mailing list