Problems inheriting from gridData's Grid class

[AstroMike]

I was using Grid as a base class for a class (in cosmology). For now let's call this CosmoGrid. But I ran into some issues.

First, the arithmetic operations (add etc) return Grid. I am not a OOP ninja, but I think if you have them return self.__class__(...) instead, then they will return an instance of the same subclass (CosmoGrid).

Second, one might want to wish to pass into Grid, not an ndarray but a subclass of an ndarray: for example an array of "Quantities" (that have astropy.units units attached, or a masked array). Will not the line self.grid = numpy.asarray(grid) convert this into a normal ndarray? (the docs say: Returns out: ... If a is a subclass of ndarray, a base class ndarray is returned.) The alternative might be numpy.asanyarray, for which the docs say "Convert the input to an ndarray, but pass ndarray subclasses through."

[richardjgowers]

Hello @AstroMike !

re 1) yes OOP says that you should use self.__class__ to avoid mangling subclassers, so this is a bug in the code

re 2) I think using np.asanyarray might fix this, sounds like another good fix

[orbeckst]

@AstroMike a PR would be most welcome – developer bandwidth is a bit limited at the moment. We're currently mostly occupied with getting a 1.0 of MDAnalysis out. This will likely be of 0% importance to you ;-). But we're very happy to make GridDataFormats better/more correct/more useful to anyone who finds a use for it.

[orbeckst]

@richardjgowers do you want to knock this one out? I want to do a 0.4.1 today or tomorrow – there were a whole bunch of bug fixes that should have been pushed out long ago