[pymvpa] Surface voxel selection saving error

[Swaroop Guntupalli]

Hi Nick,

I had a voxel selection pre-computed from a while ago (Oct. 2013).
When I try to save it now using h5py, it gives the following error (pasted
below).

Is there a way to resolve this without re-computing the voxel selection?
I can point you to the voxel selection on hydra if that helps.

Thanks,
Swaroop


/PyMVPA/mvpa2/base/hdf5.py in obj2hdf(hdf, obj, name, memo, noid, **kwargs)
    758     # try disassembling the object
    759     try:
--> 760         pieces = obj.__reduce__()
    761         if __debug__:
    762             debug('HDF5', "Reduced '%s' (ref: %i) in [%s]"

/PyMVPA/mvpa2/misc/surfing/volume_mask_dict.pyc in __reduce__(self)
    568         return (self.__class__,
    569                 (self._volgeom, self._source),
--> 570                 self.__getstate__())
    571
    572     @deprecated("should be used for testing compatibility only - "

/PyMVPA/mvpa2/misc/surfing/volume_mask_dict.pyc in __getstate__(self)
    599         # here a new tuple is created
    600         s3 = _dict_with_arrays2array_tuple(s[3])
--> 601         s4 = _dict_with_arrays2array_tuple(s[4])
    602         ss = s[:3] + (s3, s4)
    603

/PyMVPA/mvpa2/misc/surfing/volume_mask_dict.pyc in
_dict_with_arrays2array_tuple(d)
    938         # probably src2aux, so run recursively
    939         return dict((k, _dict_with_arrays2array_tuple(v))
--> 940                         for k, v in d.iteritems())
    941
    942     keys = np.asarray(d.keys())

/PyMVPA/mvpa2/misc/surfing/volume_mask_dict.pyc in <genexpr>((k, v))
    938         # probably src2aux, so run recursively
    939         return dict((k, _dict_with_arrays2array_tuple(v))
--> 940                         for k, v in d.iteritems())
    941
    942     keys = np.asarray(d.keys())

/PyMVPA/mvpa2/misc/surfing/volume_mask_dict.pyc in
_dict_with_arrays2array_tuple(d)
    965             # ensure all values in the dict have the same datatype
    966             raise ValueError('Type mismatch for keys %s and %s: %s
!= %s' %
--> 967                                 (keys[0], key, data.dtype, v.dtype))
    968
    969         idxs = np.arange(length) + pos

ValueError: Type mismatch for keys 0 and 41: float64 != float32
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-exppsy-pymvpa/attachments/20151012/bd6038b5/attachment.html>