Minor improvement of snapshots handling

[lkstrp]

Closes #1082

Changes proposed in this Pull Request

Current state

• Snapshots can be single or multi-indexed (have investment periods). This is currently represented by ambiguous index names:
  • Without investment periods:
    • n.snapshots.index.name - 'snapshot'
    • n.snapshots.index.names: ['snapshot']
  • With investment periods:
    • n.snapshots.index.name - 'snapshot'.
    • n.snapshots.index.names - ['period', 'timestep'].
• The same timestep/snapshot data is referred to by both timestep and snapshot. This can lead to all kinds of bugs, especially with the upcoming scenario dimension, and should be harmonised.

Two ways to fix this

• Rename timestep to snapshot
  • Pro: Only renames the index for networks with multiple investment periods
  • Con: Still no unambiguous naming. Since n.snapshots is both snapshot and period data, it is an arbitray access to the timestep data. But renaming n.snapshots is very intrusive.
• Rename snapshot to timestep
  • Pro: Cleaner. Ambiguous n.snapshots would be resolved.
  • Con: Would change the index name for all networks without multiple investment periods. I do not know how much this would break.

This PR implements the first option. But I would rather do the second and get rid of snapshot as an index name altogether.

Multiindex names
In any case, I would remove the single index name for multi-indexed snapshots. Since idx.names==[idx.name] for a single dimension, we can drop it for multi-indexed snapshots. This would result in clear and unique dimension names for each dimension and fix the current behaviour.

Some tests fail because linopy can't handle a tuple of coordinates, both single indexed and multi indexed at the moment. This could be fixed here or in linopy. But let me know what you think first.

Checklist

• Unit tests for new features were added (if applicable).
• A note for the release notes doc/release_notes.rst of the upcoming release is included.
• I consent to the release of this PR's code under the MIT license.

[fneum]

This PR implements the first option. But I would rather do the second and get rid of snapshot as an index name altogether.

I agree with your decision to implement the first option for now (less intrusive). Perhaps with #1037 we can then rename the index level to timestep. The main reason why we used the term "snapshot" in the first place rather than "timestep" is because it could also be used for stochastic problem definitions. With #1037 the intended use is different.

In any case, I would remove the single index name for multi-indexed snapshots.

Yes!

Some tests fail because linopy can't handle a tuple of coordinates, both single indexed and multi indexed at the moment. This could be fixed here or in linopy. But let me know what you think first.

Sounds more like it should be addressed in linopy.

[fneum]

index_name should be removed from docstring

[fneum]

not intended?

[fneum]

Mostly restructuring?

[lkstrp]

I agree with your decision to implement the first option for now (less intrusive). Perhaps with #1037 we can then rename the index level to timestep.

@fneum, hmm, not sure I agree. If people are using index names in n.snapshots, this is a breaking change and can't be made backwards compatible. Probably not a problem for many users, but still. Merging this pr with option 1, and fixing it later with 1037, would just mean breaking things twice for those users.
So I would rather go with option 2 here and put it already in v0.32.

[fneum]

Ok, convinced. It's just an index name after all :)

[lkstrp]

For the history books:
This PR confuses dimension with dimension levels.
Periods and timesteps are just levels of the snapshots dimension, and this cannot be changed:

• timesteps are not unique for each investment_period, but contain year data. The year cannot be removed either, since timestamps cannot exist without year information.
• It is also possible to define different numbers of timesteps per investment period.
  For them to be independent dimensions, both would be necessary.

A follow up PR will clarify that.