[SciPy-user] HDF5 vs FITS (was: Fast saving/loading of huge matrices)

Francesc Altet faltet at carabos.com
Sun Apr 22 05:57:04 EDT 2007


El dv 20 de 04 del 2007 a les 16:26 -0400, en/na James Turner va
escriure:
>  > As a side note, we do use this richness of hdf5 in our experiment, to
>  > store say the time of an experimental run, the temperature of the room...
> 
> It sounds like HDF5 provides much the same capabilities as FITS, the
> main file standard used for some decades in astronomy. It also sounds
> like there may be a lot of overlap between Pytables and STScI's binary
> tables, as implemented in PyFITS. I imagine that's why Pytables was
> based on numarray, come to think of it...

Well, I must recognize that I didn't know about STScI's TABLES packages
until a couple of years ago (when PyTables was already more than three
years old), so no idea on the amount of overlap between both projects.
However, I'd say that both projects are rather different in aims or, at
very least, on the formats that they are based on.  So, probably, most
of the overlap would come basically from the similarity of the
names ;-).  TABLES being born before 1998 has a clear precedence for
reclaiming the name over PyTables which first public version was
released in 2002, but as its application fields are quite different (in
principle), I don't think there is not much point in thinking about
changing the name of the latter, IMO.

And no, I didn't choose numarray for the very first version of PyTables
because it was used already used for TABLES (in fact, I think that
TABLES appeared much before than numarray; correct me if I'm wrong), but
because numarray was the only Python package with a powerful
implementation of recarrays, objects that were critical for the PyTables
aims.  Let me stress out that the PyTables project is in debt with the
excellent (although, with the advent of NumPy, the qualifier 'venerable'
is beginning to be applicable also) numarray package and its developers,
because without them PyTables wouldn't exist (or at least, wouldn't have
many of its current features).

Cheers,

-- 
Francesc Altet    |  Be careful about using the following code --
Carabos Coop. V.  |  I've only proven that it works, 
www.carabos.com   |  I haven't tested it. -- Donald Knuth




More information about the SciPy-User mailing list