[FEATURE] Merge grids

[Thijss]

It would be nice if we could merge multiple grids via e.g. the following syntax:

grid_1_and_2 = grid_1.merge(grid_2)

The challenge here is to avoid id conflicts in the resulting grid.
The proposal is to regenerate all ids to avoid conflicts.

Note 1: regenerating all ids also means that all references to ids (i.e. node ids) need to be updated too.

Note 2: Teams that require a UUID on assets across all grids, need to implement this themselves by extending the Grid

[vincentkoppen]

We have aimplementation of this in our private project, which is really basic.

def _increment_grid_ids_by_offset(grid: GridExtension, offset: int) -> None:
    for array in grid.all_arrays():
        if isinstance(array, IdArray):
            array.id += offset

        if isinstance(array, NodeArray | SymPowerSensorArray | SymVoltageSensorArray | AsymVoltageSensorArray):
            continue

        elif isinstance(array, TransformerTapRegulatorArray):
            array.regulated_object += offset

        elif isinstance(array, BranchArray):
            array.from_node += offset
            array.to_node += offset

        elif isinstance(array, Branch3Array):
            array.node_1 += offset
            array.node_2 += offset
            array.node_3 += offset

        elif isinstance(array, SymGenArray | SymLoadArray | SourceArray):
            array.node += offset

        else:
            raise NotImplementedError(f"The array of type {type(array)} is not implemented for appending")

[vincentkoppen]

A problem with the above implementation is that all the references need to be hardcoded, since we cannot detect which columns are id columns.
More specifically if a user extends the grid with an additional id column/reference, we will not update those. If we want a more robust solution, I think we might need to extend our typing.

So that we have:

type ID_COLUMN = numpy.int32
class IdArray(..):
  id: NDArray[ID_COLUMN]

class BranchArray(..): 
  from_node: NDArray[ID_COLUMN]
  to_node: NDArray[ID_COLUMN]
...

Now you could loop over all the arrays and columns, check if it is of ID_COLUMN type and if so update by the offset. This simplifies the function and maintainance later on. But we need to introduce and agree on the new typing.

We might even discuss this with the general power-grid-model team and see if they can introduce this typing in general