Write Datapackage.json

[FelixMau]

To be able to create an energy system from a data package, you must provide a datapackage.json

It is necessary to write down foreign_keys whilst build_datapackage to keep track of timeseries column names and resource names. (Closes #23 ) Note: There is already a method within DataPackage class prepared for this purpose

The data package must provide the following information:

• [x] The path where the resource (and foreign resources) is located.

• [x] Information on fields, including their types and metadata (defaults could be used)

• [x] Foreign keys that provide additional information, such as bus data.

• [x] "Foreign keys" pointing to time series data, (different from ordinary foreign keys.)

• [x] Additionally, a dataset needs to be built to contain all the buses (whilst build_datapackage).

To monitor and check whether code is working properly:

• [ ] write Test
• [ ] define expected and Input data

[FelixMau]

When initializing classes with attributes that can be either a sequence or a scalar sequence() is being called from __init__. This Process turns eventually given default values like the fixed_costs within facades.Storage from integer into _Sequence that is emulating a list with "unknown length" as explained in this example:

>>> s = _Sequence(default=42)
    >>> len(s)
    0
    >>> s[1]
    42

The Sequence class cannot be saved within a regular Datapackage.

Suggestion 1:

When unpacking the AdapterClass to a dict with as_dict (to save it as an Pandas.DataFrame) I suggest that we test if we have recieved a sequence and if yes writing the default value of the _Sequence to the dict.

Suggestion 2:

As it is now a empty list is being returned and when deseralizing the Datapackage it is interpreted as full Sequence leading to errors since the sequence is actually empty. I think we could also improve on the tabular deserialization part and improve its stability by catching this error and take the default value if a faulty iterable is passed. This might be helpful as well for upcomming implementations of periodically changing scalars. Edit: Problem here might be that we do not know the full timeseries length (or amount of steps) yet when initializing the facades, correct?

I would like your feedback regarding the plans on scalars within multi period feature.

[nailend]

@FelixMau

Suggestion 1:

I agree with you, thank you for tackling this. We will probably have to rework this after the deserialization is solved, but we will do this in a separate issue.

Suggestion 2:

Currently, the empty list is detected as empty list which is fine, as we actually save it in the wrong way. So we have to think of a different way of passing a _Sequence via the datapackage. Concerning the periodic-scalar, I would wait until this is/or if it is fixed on the oemof.solph side, but I am keen enhancing _sequence for this. It makes sense to think this together. But we probably have to take care of #116 before

[nailend]

I gave the plumbing a try, see here, do you think thats possible to implement from the adapter side? to somehow catch it if the parameter change per period and are then written as a dict?

But lets tackle this is in a new Issue

[FelixMau]

Test improvements moved to #43 Developing Period merging and reading moved to #40