[Numpy-discussion] the path forward

Alan G Isaac aisaac@american....
Tue Apr 29 15:55:53 CDT 2008


On Tue, 29 Apr 2008, Bruce Southey apparently wrote:
> There is no additional benefit from having row or column 
> shapes or metadata because the row/column nature is 
> usually predetermined and would be represented by the 
> shape of the corresponding matrix. 

The problem is that ``x[0]`` being 2d has produced a variety 
of anomalies, and the natural fix is for ``x[0]`` to be 1d.

Gael has argued strongly that she should be able to use the 
following notation: ``x[0,:]*A*x[:,0]``.  But this will work
only if ``x[0,:]`` is 2d or if it is 1d but has an "orientation".

So *if* you think ``x[0] == x[0,:]`` is desirable, *and* you 
want to satisfy Gael, *then* it seems you must introduce 1d 
"oriented" vectors.

I believe Travis is also suggesting that we travel that 
road, taking a first step as follows:
for now let ``x[0]`` be a 1d array to quickly fix the 
anomalies, but let ``x[0,:]`` continue to be a matrix
until the vector code is written, at which point ``x[0]`` 
and ``x[0,:]`` we be the same "row vector".

Or so I have understood things.

Cheers,
Alan Isaac





More information about the Numpy-discussion mailing list