[SciPy-dev] integration of Unum in scipy

DE MENTEN Sebastien sebastien.dementen at alumni.insead.edu
Mon Feb 20 22:30:05 CST 2006

> > One weakness at that time was its integration with Numeric. Using
> > functions such as sum, cumsum & co on Unum objects was not giving
> > ideal result. Today, with the added capabilities of ufuncs (namely
> > array_wrap) this problem vanishes.
> Well, I wouldn't say that the problem "vanishes" so much as "maybe
> tractable." Only experience will tell how tractable.

At least, the point (my personal Unumeric showstopper) which bothered me
really vanished :)

> > Hence, it could make sense to integrate it nicely with scipy.
> It would certainly be nice for Unum to be able to handle numpy arrays
> well.
> However, we cannot accept Unum into scipy since it is GPLed.

I think the author (Pierre Denis) could change his mind and the license
as he does not feel too strong about it (it was more a default choice if
I remember correctly). Anyway, in case of positive answer I will ask him
permission for both the integration and the license change.

The question is more about a formal integration in scipy:
 a) The code is mature, clearly written and not huge (engine code is 600
loc and unit definitions code is 1 loc per unit). Its integration should
not be a big deal.
 b) The package nicely fills a gap in the scientific toolkit.

Hence, is the license the only problem?


More information about the Scipy-dev mailing list