Return dataframe from xmsmesher

[kcpevey]

The helper function:

def xmsmesh_to_dataframe(pts, cells):
    """
    Convert mesh pts and cells to dataframe
    
    Args:
      pts (MultiPolyMesherIo.points): Points from a MultiPolyMesherIo
      cells (MultiPolyMesherIo.cells: Cells from a MultiPolyMesherIo
      
    Returns:
      pd.DataFrame: MultiPolyMesherIo points in a dataframe
      pd.DataFrame: MultiPolyMesherIo cells in a dataframe
    """
    r_pts = pd.DataFrame(pts, columns=['x', 'y', 'z'])
    r_cells = pd.DataFrame([(cells[x+2], cells[x+3], cells[x+4]) for x in range(0, len(cells), 5)], columns=['v0', 'v1', 'v2'])
    return r_pts, r_cells

would logically reside in the package itself and allow for users to reformat the output into dataframes either with a keyword flag, or as a method on xmsmesh.meshing.MultiPolyMesherIo class. Personally, I'd prefer something like this:

mesh_io = xmsmesh.meshing.MultiPolyMesherIo(poly_inputs=input_polygon, refine_points=refine_points)
where
mesh_io.dframe() would return the equivalent of xmsmesh_to_dataframe above.

Other ideas would be:

• have dataframes be default output
• add a mesh_io.array() method that returns a np.array.

[sdc50]

To be consistent with xarray and pandas I would suggest method names like to_dataframe or to_array.

I really think that the generate_mesh function should just take all of the arguments that the MultiPolyMesherIo object does and handle the creation of that object within the function. Then that function can also have an argument like format where the return type is specified. For example:

generate_mesh(poly_inputs=input_polygon, refine_points=refine_points, format='dataframe')

Valid options for the format argument could be 'dataframe', 'array', 'xarray', 'list', 'dict'

To clarify, currently generate_mesh currently returns a tuple of a bool and list of strings (succeeded, errors) This should be changed so that if it does not succeed it raises an error with the error messages. The return value should be points and cells in the specified format (i.e. as a tuple of dataframes or arrays or lists, or as an xarray or dict).

If there were a widely accepted Python data structure for an unstructured grid (could xmsgrid become this?) then I think generate_mesh should, by default, return that kind of object. If such an object existed I would assume it would have standard conversion methods like to_dataframe, to_array etc.