Get a `TensorField` from `TensorFieldData`

[igirault]

Currently the class TensorFieldData enables to retrieve a list of fields via the static method TensorFieldData::GetFields. More exactly, this static method return a list of Grids, stored in the method as a public member named grids.

This would be great if TensorFieldData could also have a member of type TensorField, so that it could be directly retrieved via a similar static method TensorFieldData::GetTensorField. This would make the implementation of boundary conditions easier, as we could manipulate directly a TensorField instead of a list of Fields when handling vectors for instance.

[PhilipDeegan]

you mean like this ?

[igirault]

I mean having the member that is an instance of this type and wraps the grids member

[PhilipDeegan]

once you have using data_type = core::TensorField<field_type, PhysicalQuantity, rank>; it's quite simpler to do what you want which I guess would look like

  static auto getTensorField(SAMRAI::hier::Patch const& patch, int id)
  {
      auto const& patchData = std::dynamic_pointer_cast<data_t>(patch.getPatchData(id));
      if (!patchData)      
          throw std::runtime_error("cannot cast to TensorFieldData");
      data_t tf{patchData->name_, patchData->quantity_}; // need to add name_ to tf_data class       
      tf.setBuffer(&patchData->grids);
      return tf;
  }

if you need it you're probably the best person to add it

[PhilipDeegan]

A tensorfield is basically a list of fields, so unless you need the name of the tensorfield directly rather than the name of each component, there isn't really a whole lot of difference IMO