These are chat archives for SHTOOLS/SHTOOLS

Aug 2016
Mark Wieczorek
Aug 26 2016 12:44

@all This pull request is about ready to merge: SHTOOLS/SHTOOLS#51

I suggest that you look through it (as well as the two revised Intro notebooks), and if there are no problems, we can merge this monday. The PyPI PR is probably ready to merge at the same time. I would like to inspect the random phases PR, but don't expect there to be any problems. As for the power spectrum PR, I have not had time to look at this, but from a cursory glance, I think it is a good idea. I suggest that we hold off on this one until the next major release where we fix the Fortran STOP issue. I suspect that we will need to rewrite some other routines that call this.

As I note in the PR, there are still a few things I would like to do. However, it might be best to just release this as is as v3.3.1 next week so that we can upload it to PyPI. My schedule will be too busy in september to do anything meaningful. Nevertheless, in a couple months (before AGU), I think that we can release 3.4, which fixes all major issues. After 3.4, we can start thinking about how to proceed with more fundamental changes, like perhaps converting some of the fortran code to python.

Ilya Oshchepkov
Aug 26 2016 13:02
In all operators maybe it is better to rename arguments from cnm and grid to other, because it is more obvious that this is another class instances
Ilya Oshchepkov
Aug 26 2016 16:20
Also, in all operators in SHCoeffs it is better to explicitly compare shape of the self.coeffs or self.lmax... Of course, there will be numpy error if they don't have same shape, but anyway
Or it is posible to fill with zeros the array with the smaller lmax and equalize the shape
Ilya Oshchepkov
Aug 26 2016 16:25
Yes, only for __sub__ and __add__, but not for __mult__...