Can now initialize a universe with a Topology object

[dotsdl]

As an alternative to giving a topology file, now an existing Topology object can be given to a Universe on init. This allows many Universe objects to share the same Topology object, so changes to the topology features of one changes them all.

[dotsdl]

Not sure if this was the best place to put this functionality. I don't think it really hurts, though. But that's what this PR is for. :D

[hainm]

kwargs.pop('topology_format') is enough. :D

updated: oops, mistaken with .get method.

kwargs.get('format', None) is None

should be

kwargs.get('format') is None

[dotsdl]

This was already here; the reason it counts as a change is because of the indentation into another if-else level. I see the point and will probably remove it, but as it is it is more explicit without looking up the default rules for dict.get().

[hainm]

but as it is it is more explicit without looking up the default rules for dict.get().

agree.

[kain88-de]

The argument should go into the definition of the init. That is much better. Kwargs should only be used to pass arguments to called functions.

[dotsdl]

@kain88-de I agree, and if it were completely up to me I would make it so that the Universe init took topfile followed by trajfile, with the possibility for trajfile to be a list of filenames. Because instead we use *args style arguments for this, it's not easy to do unless we're willing to break a very visible API piece.

[richardjgowers]

I think the other thing using *args lets us do easily is having no traj/all in one files (gro/pdb). Otherwise you'd need Universe(top, None) or something ugly

[dotsdl]

None could always be the default for traj. I don't think it would be a problem, but it would be a clear change.

[richardjgowers]

I dunno, I dislike how long the current Universe init is, so I'd prefer a class method like Universe.from_Topology. Is there a reason this won't work here?

[dotsdl]

@richardjgowers I think the init should be broken up into methods, but that it should still be able to take Topology objects as a topologies. I think this is a good place to discuss this, as I do think the Universe init could use some rethought.

[richardjgowers]

Maybe we should remember the filename in the Topology object, so here self.filename = topology.filename. It will be annoying if somewhere down the line we've lost memory of what file this came from

[richardjgowers]

Actually, with #643, it does just make sense to keep the (topology, traj) signature, even if it's a not a topology file, so I like this.