[AstroPy] AstroPy Digest, Vol 58, Issue 16

Perry Greenfield perry at stsci.edu
Tue Jun 14 15:57:31 EDT 2011

On Jun 14, 2011, at 2:27 PM, Tommy Grav wrote:
> I also think there needs to be a certain control by certain key  
> contributing
> people. One problem seems to be that when people just develop a  
> package
> and contribute it to something like Astrolib there is no continuity  
> between
> the package and the use of them. I think the strength of numpy is that
> is sprung from the work of a few people and any new developer has to
> adopt a certain style of coding that maintains the package overall  
> consistency.
> Scipy seems to be a little looser, but still has some central  
> control by a
> small group of developers. I think astropy/astrolib needs something  
> similar,
> making it a package in itself, not just a collection of packages  
> gathered from
> a wide variety of sources. There has to be a certain focus to make  
> it successful
> I think.
> Cheers
> Tommy

You must remember that it was easier with these two projects,  
particularly numpy. numpy is essentially a pretty clear integrated  
whole with a series of previous owner from which it was derived. At  
one point it was developed primarily by Jim Hugunin, then we did  
numarray, then Travis essentially merged the two.

Not quite so simple for scipy, but scipy started out mostly as gluing  
together existing libraries. As such, there weren't a lot of  
architectural decisions to make about interfaces (and I think some  
complain about the lack of consistency still as a result). These  
involve many, mostly mature, libraries that had their own primary  
developers (and still do for the most part).

In this case, we are talking about remaking much of the astronomical  
software from scratch with many different kinds of potential users. So  
I think in many respects, this is much more difficult than either  
numpy or scipy had to deal with in organizational terms.


More information about the AstroPy mailing list