For-node pain?

[liamhuber]

I already wince every time I have to write a custom function to access the output of the for nodes

Wait, this should not hurt! Can you give me an example? If there's something that's is regularly annoying, we should 100% fix it. Does this have to do with type-hinting like #703?

Don't forget, you can get the outputs as individual channels instead of a dataframe!

import pyiron_workflow as pwf

wf = pwf.Workflow("no_custom_reading")
wf.produce_list = pwf.for_node(
    pwf.std.Add,
    iter_on="obj",
    obj=[1, 2, 3],
    other=10,
    output_as_dataframe=False,
)
wf.consume_list = pwf.for_node(
    pwf.std.Add,
    iter_on="other",
    obj=-10,
    other=wf.produce_list.outputs.add,
    output_as_dataframe=False,
)
wf()

Gives the output, but also the input being iterated over (#674):

{'produce_list__obj': [1, 2, 3],
 'consume_list__other': [11, 12, 13],
 'consume_list__add': [1, 2, 3]}

TBH, I'd actually prefer output_as_dataframe=False as the default, or to get rid of the pandas dependence entirely...

[ligerzero-ai]

Ill get back to you on this soon, but not this week soon.

[ligerzero-ai]

Consider:

wf.gb_seg_structure_generator = for_node(
        create_seg_structure_and_output_dir,
        structure=structure,
        iter_on=("defect_site"),
        defect_site=defect_sites,
        structure_basename=structure_basename,
        element=element,
        parent_dir=parent_dir,
)
wf.gb_seg_structure_list = get_df_col_as_list(
    wf.gb_seg_structure_generator.outputs.df, "structure"
)
wf.gb_seg_structure_dirs = get_df_col_as_list(
    wf.gb_seg_structure_generator.outputs.df, "output_dir"
)

I have to access the outputs of the node object (in this case df) like so:

@pwf.as_function_node
def get_df_col_as_list(df, col):
    output_list = df[col].to_list()
    return output_list

[liamhuber]

I have to access the outputs of the node object (in this case df) like so:

I don't have quick access to that node, but try this:

wf.gb_seg_structure_generator = for_node(
        create_seg_structure_and_output_dir,
        structure=structure,
        iter_on=("defect_site"),
        defect_site=defect_sites,
        structure_basename=structure_basename,
        element=element,
        parent_dir=parent_dir,
        output_as_dataframe=False,   # Add this!!!
)
print(wf.gb_seg_structure_generator.outputs.structure.value)
print(wf.gb_seg_structure_generator.outputs.output_dir.value)

By switching off dataframe output, you should get direct access to what you're looking for in the outputs without any get_df_col_as_list middleman

[liamhuber]

TBH, I would not only like to make the list output the default, I want to remove the dataframe output option altogether. Here and the InputsToDataframe transformer are the only places we use pandas and I would be happy to drop a dependency...