[pymvpa] dataset.selectFeatures again

[Susanne Greiner]

Dear all,

just a short syntax question: why is it not possible to select Samples 
and Features the same way when using nifti datasets? In my opinion 
someone new to pyMVPA who has a look at the example files gets across 
the dataset.selectSamples function. This function is used in almost all 
examples with a boolean N.array and looks very intuitive.
Why can't selecting features with boolean arguments be easy in exactly 
the same way? If you try it with a Dataset, it works fine, but when 
using a MappedDataset the analogon does not work any more :-( and the 
''array is not broadcastable to correct shape''
Is that really wanted (or needed for some reason)?
code showing the precise problem below,

lg

Susanne

-------------------------------

from mvpa.suite import *
import numpy as N

# load PyMVPA example dataset or any other NiftiDataset
#
attr = SampleAttributes('data/attributes.txt')
dataset = NiftiDataset(samples='data/bold.nii.gz',
                       labels=attr.labels,
                       chunks=attr.chunks,
                       mask='data/mask.nii.gz')

# only considering 10 samples and 10 features
dataset = dataset.selectSamples(range(10))
dataset = dataset.selectFeatures(range(10))


# this works
dataset = 
dataset.selectSamples(N.array([True,False,True,False,False,True,False,True,False,False],dtype='bool'))

# this does not work but produces: array is not broadcastable to correct 
shape
dataset = 
dataset.selectFeatures(N.array([True,False,True,False,False,True,False,True,False,False],dtype='bool'))

# this works and intuitivly the line above should do the same thing or 
why not
#dataset = dataset.selectFeatures([0,2,5,7])

# this works but the dataset is not mapped any more :-(
#dataset = 
dataset.selectFeatures(N.array([True,False,True,False,False,True,False,True,False,False],dtype='bool'),plain=True)

# this works and everything is fine, but it seems not very straight 
forward to me
#dataset = 
dataset.selectFeatures(where(N.array([True,False,True,False,False,True,False,True,False,False],dtype='bool'))[0])