[Numpy-discussion] Re: [SciPy-user] Fwd: Scipy_core and numarray

[Travis Oliphant]

Andrew Jaffe wrote:

>Hi All,
>
>I've tried to post via the newsgroup interface, but my messages have
>never appeared, so here's another try!...
>
>I'm currently a numarray user from the astrophysics community and I
>suppose I want to understand a bit more about the relationship between
>the two projects. Do the two projects exist side-by-side? Will numarray
>be superseded by scipy_core? I like the speed advantage of scipy_core,
>but I there may be features of numarray I don't want to give up (or
>legacy code I need to interfact with).
>  
>
I'm interested in the "features" you don't want to give up.  SciPy core 
should have all features of numarray. 

Also, scipy core, numarray, and Numeric can all live together.  If you 
are using sufficient versions of each, you can exchange data via the 
array interface.

The ONLY reason I started scipy core was to unify Numeric and numarray 
users.   My selling documentation, should not be construed as anything 
else but trying to help people who can't find time to contribute in 
other ways to contribute to that goal.  

If I was "just trying to make money,"  I would be doing something far 
different than this...

>
>Along the latter lines, I have a more practical question: one
>numarray capability that I used a lot was the ability to create
>uninitialized arrays, such as
>
>     arr = numarray.array(shape=(3,5), type=Float64)
>  
>
It's actually there in Numeric (since 23.7 I think).

Numeric.empty((3,5),'d')

and also in (new) scipy

scipy.empty((3,5),scipy.float64)

In (new) scipy you can also create new arrays from buffers very easily.

-Travis