[Matrix-SIG] I have working code that implements active-casting concept.

[Scott M. Ransom]

Rick White wrote:
> 
> Travis's changes (including moving array_set into the Numeric module,
> etc.) are big steps toward removing the major shortcomings of Numeric
> for our work (astronomical data analysis and image processing).

Hear! Hear!  I completely agree.  

And I would like to thank Travis for stepping up to the plate and making
the changes.

With these changes, I will almost certainly become a more active member
of the NumPy community -- and will attempt to spread the NumPy gospel
among my astronomer colleagues.  (Down with IDL!)

As for a request for particular "single precision problems", I think
Travis' current changes have addressed them completely (at least in my
case).

The main problem came from passing single precision arrays to/from NumPy
from/to C (or Fortran) library functions.  While this was certainly
possible in the past, it became a royal pain if you wanted to manipulate
these arrays in Python before or after calling the library functions --
the arrays were up-cast to double (I got around this before by writing a
simple C function to downcast double precision NumPy arrays to single
precision).

Hopefully this is now a problem of the past.

Thanks again Travis,

Scott


-- 
Scott M. Ransom                   Address:  Harvard-Smithsonian CfA
Phone:  (617) 495-4142                      60 Garden St.  MS 10 
email:  ransom@cfa.harvard.edu              Cambridge, MA  02138
PGP Fingerprint: D2 0E D0 10 CD 95 06 DA  EF 78 FE 2B CB 3A D3 53