Re: [Pytables-users] h5file size difference between dataset created with earray and carray

[Francesc A. <fa...@py...>]

A Friday 26 June 2009 17:44:48 Didier Roissé escrigué:
> Hello,
> I created a HDF5 file containing an array with the following dimension
> (5000,1). I notice a big difference file size between the array creation
> with createCArray and createEArray.
> For creation of the table with createCArray, i use this script :
>
> h5ft=openFile('test_carray.h5','w')
> ca=h5ft.createCArray(h5ft.root,"foo",Float32Col(),(5000,1),"carray test")
> for ii in range(5000):
>   ca[ii][0]=1.0

Here is culprit.  When assigning to a PyTables object, always specify the 
indexes in one shot.  For example, replace the above line by:

  ca[ii,0]=1.0

and you are done.  Or you can avoid the last index completely too:

  ca[ii]=1.0

as the input is broadcasted to the affected index.

What was happening here is that ca[ii] returns the element ii of the CArray as 
a (temporary) NumPy array, and when you do ca[ii][0]=1.0, you are modifying 
that NumPy array, not the value on disk.

Hope that helps,

-- 
Francesc Alted