Re: [Cdat-discussion] Pb with CDAT beta 5 install

[Andrew C. <ac...@gm...>]

I think Julien raises an important point. Personally, I like that I
can install CDAT as a single unit, and have a complete set of packages
available, similar to the Enthought python distributions, and other
bundles. Especially with a beta, I'd rather build the whole thing,
python and all, and keep it seperate from my other python
installation. I've also had more sucess building it as a unit, at
least on OSX.

However for production use I'd like the CDAT packages to work
comfortably with my main python install - it's nice being able to swap
between using cdms and Jeff W.'s netcdf interface, or use matplotlib
for line plots but vcs for spatial maps and so on.

I suppose a key question for the CDAT developers is which  entails
more work: designing CDAT primarily as a set of python packages to be
installed into an existing installation, or designing CDAT as a
stand-alone bundle where the versions of various packages can be more
tightly controlled?

-Andrew


On Fri, Oct 10, 2008 at 2:41 AM, Julien Vienne <jv...@cl...> wrote:

> Well, I'm not sure this is a good idea since CDAT is not really based on.
> Personaly, I already have matplolib installed and I would prefer CDAT
> not installing another one.
> In general, I think CDAT should check only installed softwares and be
> built in consequence, using only existing dependencies (if possible)
> ++
> Jul.


-------------------------
Andrew Charles