Hi Daniel,

On 05/27/2014 10:01 PM, Daniel Wheeler wrote:

Hi,

I noticed that the name "Domain" has changed to "FEDomain".

https://github.com/sfepy/sfepy/commit/3af511599110b22e327944b009b37acd3c35c3...

While this is probably a change for the better, I would recommend being very cautious about such changes. For example, it breaks code that I'm hoping to package and release at some stage. This seems to be an API change and Sfepy probably needs a policy for this. Maybe some discussion on the mailing list for example. In traditional release models with X.Y.Z release numbers, this would probably require changing the X value. I realize that Sfepy has rolling releases which is fine, but similar caution should probably be employed.

Yes, you are right, such a change would better be announced on the list. (In past, I usually got no replies to similar enhancement proposals, so I got sloppy.)

Note that this state will not probably stay long: both FEDomain and IGDomain (for isogeometric analysis) are still subclasses of Domain. This interface change was a quick (& a bit dirty) way to proceed with the release. I think it is good to keep the releases regular (4 each year), even if some new code is not completely polished (worse is better etc.), to get a feedback - IMO most people take the releases and not the git master.

Anyway, with Field it was possible to keep the original interface, so Domain might follow that pattern. It's just there is still a lot to be done in terms of linearization of IGA solutions (for postprocessing) and similar issues - FEDomain and IGDomain differ more than FEField and IGField.

(If you do not want to do Domain -> FEDomain everywhere, import FEDomain as Domain for the moment.)

Cheers, r.