np.lib.index_tricks
has a number of functions (and classes) that can streamline indexing. np.s_
is one such function. It is actually an instance of a class that has a __get_item__
method, so it uses the []
notation that you want.
An illustration of its use:
In [249]: np.s_[1:5:2,::3]
Out[249]: (slice(1, 5, 2), slice(None, None, 3))
In [250]: np.arange(2*10*4).reshape(2,10,4)[_]
Out[250]:
array([[[40, 41, 42, 43],
[52, 53, 54, 55],
[64, 65, 66, 67],
[76, 77, 78, 79]]])
In [251]: np.arange(2*10*4).reshape(2,10,4)[1:5:2,::3]
Out[251]:
array([[[40, 41, 42, 43],
[52, 53, 54, 55],
[64, 65, 66, 67],
[76, 77, 78, 79]]])
Notice that it constructs the same tuple of slices that ajcr
did.
_
is the temporary variable that IPython uses for the last result.
To pass such a tuple to a function, try:
def myfunction(some_object_from_which_an_array_will_be_made, my_index=np.s_[:,:]):
my_array = whatever(some_object_from_which_an_array_will_be_made)
return my_array[my_index]
I = np.s_[1:5:2,::3]
myfunction(obj, my_index=I)