Handling of vector-valued coupling data

[RafalKulaga]

For structural solver handling of vector function space is required, however this is currently not supported by the fenicsadapter.py . At first it seemed that it can be fixed easily, as there are Vector equivalents of Scalar functions in PySolverInterface.py (e.g. writeBlockScalarData() and writeBlockVectorData()), but actually there is no difference in the body of these functions, hence it does not help.

Anyway, I am first going to change from 'PySolverInterface.py' to 'precice' and then try to cope with this issue (#25).

[uekerman]

I changed the name of the issue. Did I get it right, is that what you want to do?

What do you mean by?

... but actually there is no difference in the body of these functions, hence it does not help.

[RafalKulaga]

Yes.

I should now refer to precice.pyx functions and not PySolverInterface due to #25. I just mean that simple changing from e.g. write_block_scalar_data() to write_block_vector_data() in fenicsadapter.py does not help. Using vector-valued functions yields the same error in both cases. This is understandable as the implementation of "scalar" and "vector" version of read and write does not differ (it is the same code). Please compare
https://github.com/precice/precice/blob/develop/src/precice/bindings/python/precice.pyx#L259-L281 and https://github.com/precice/precice/blob/develop/src/precice/bindings/python/precice.pyx#L300-L322

I will push the draft of structural solver, so the error (TypeError: only size-1 arrays can be converted to Python scalars) can be reproduced.

[uekerman]

This is understandable as the implementation of "scalar" and "vector" version of read and write does not differ (it is the same code)

Well, it does call two different preCICE methods that do different things. Probably, I understand better what you mean when you push the draft of the structural solver.

Also for the vector-valued case: len(value_indices) != len(values).

Still values needs to be one-dimensional array (of size numberOfVertices times dimensions). Maybe you are looking for the method ravel ?

[RafalKulaga]

@uekerman Thanks, I did not know that it must be one-dim array. Using ravel helped, but just to make sure: from [ [u1, v1], [u2, v2] ...] form we want [ u1, v1, u2, v2, ...] and not [ u1, u2, ... v1, v2, ... ], right?

The issue cannot be closed yet though, there are further modifications in fenicsadapter.py required.

[uekerman]

but just to make sure: from [ [u1, v1], [u2, v2] ...] form we want [ u1, v1, u2, v2, ...] and not [ u1, u2, ... v1, v2, ... ], right?

Exactly! See also the API documentation

[BenjaminRodenberg]

Closed. Vector valued data is supported since #38 has been merged.