Unbundle oemof and solph

Unbundle oemof and solph to clarify the structure.

At the moment we find oemof=solph in the public discussion. In my opinion oemof is solph, tespy, feedinlib, windpowerlib, cydets, oemof_thermal and oemof_tabular.

Therefore, I would suggest to create a solph repository. The oemof repository would be a dummy package that installs all stable (typical?) oemof packages.

Open question:

1. Where do we want to place the network stuff? a. as subpackage in the solph package b. in a new package (oemof_base, …) c. …
2. Will the outputlib part of the solution above (1) or an own package? 3 …

I just want to start an open discussion about that and we may substantiate it on the next dev-meeting.

[joroeder]

One further thought: From a technical perspective, feedinlib and demandlib are very similar. They both prepare/ create time series. Does it make sense to merge them?

How about putting all other precalcaultions of timesieres there as well? like heatpump COPs?

[p-snft]

In my opinion, unbundling makes perfectly sense. I agree, that the oemof repository can be used as a meta-repository that pulls all oemof packages. This way, also backward compatibility can probably be maintained. Further, it would make things more comfortable, as e.g. the examples might be automatically updated. (We could use git submodules for that.

Concerning the questions, I have the feeling that both, network and outputlib, could be seperate repositories:

1. Network can be used without solph. (Potentially. I did not yet do much for scoop, yet.)
2. Outputlib can potentially work without solph. In fact, here at ENaQ we detach optimisation from analysation. The second step relies on a model containing results, which might e.g. have been post-processed.

Then: 3. Would e.g. the TESPy package become oemof.tespy?