[Suggestion] Dealing with the data node attribute "name"

[GeigerJ2]

Currently, when going the route, PWD -> WfMS -> PWD, in the second step, the "name"s of input nodes are re-generated (from the function arguments) by the write_workflow_json functions that uses these helpers, e.g., see here:

python-workflow-definition/python_workflow_definition/src/python_workflow_definition/shared.py

Lines 52 to 105 in 00bc3c5

	def update_node_names(workflow_dict: dict) -> dict:
	node_names_final_dict = {}
	input_nodes = [n for n in workflow_dict[NODES_LABEL] if n["type"] == "input"]
	node_names_dict = {
	n["id"]: list(
	set(
	[
	e[TARGET_PORT_LABEL]
	for e in workflow_dict[EDGES_LABEL]
	if e[SOURCE_LABEL] == n["id"]
	]
	)
	)[0]
	for n in input_nodes
	}
	counter_dict = Counter(node_names_dict.values())
	node_names_useage_dict = {k: -1 for k in counter_dict.keys()}
	for k, v in node_names_dict.items():
	node_names_useage_dict[v] += 1
	if counter_dict[v] > 1:
	node_names_final_dict[k] = v + "_" + str(node_names_useage_dict[v])
	else:
	node_names_final_dict[k] = v
	for n in workflow_dict[NODES_LABEL]:
	if n["type"] == "input":
	n["name"] = node_names_final_dict[n["id"]]
	return workflow_dict
	def set_result_node(workflow_dict):
	node_id_lst = [n["id"] for n in workflow_dict[NODES_LABEL]]
	source_lst = list(set([e[SOURCE_LABEL] for e in workflow_dict[EDGES_LABEL]]))
	end_node_lst = []
	for ni in node_id_lst:
	if ni not in source_lst:
	end_node_lst.append(ni)
	node_id = len(workflow_dict[NODES_LABEL])
	workflow_dict[NODES_LABEL].append(
	{"id": node_id, "type": "output", "name": "result"}
	)
	workflow_dict[EDGES_LABEL].append(
	{
	TARGET_LABEL: node_id,
	TARGET_PORT_LABEL: None,
	SOURCE_LABEL: end_node_lst[0],
	SOURCE_PORT_LABEL: None,
	}
	)
	return workflow_dict

One can still, in principle, modify the "name"s in the PWD, however, when one starts with this PWD:

{
  "version": "0.0.1",
  "nodes": [
	...
    {"id": 3, "type": "input", "value": 1, "name": "a"},
    ...
  ],
  "edges": [
    {"target": 0, "targetPort": "x", "source": 3, "sourcePort": null},
	...
  ]
}

loads it into the WfMS and then exports it again to PWD, the result will be this:

{
  "version": "0.0.1",
  "nodes": [
	...
    {"id": 3, "type": "input", "value": 1, "name": "x"},   <-- This got replaced with the value from the targetPort
    ...
  ],
  "edges": [
    {"target": 0, "targetPort": "x", "source": 3, "sourcePort": null},
	...
  ]
}

Hence, in the PWD JSON schema, they somewhat present redundant information from the port information (as said above, they can still be added/modified, it's just not respected by the WfMS, or persistent on interconversions). pyiron and jobflow don't support global workflow inputs and outputs (yet), instead, these are based on the arguments of the first function(s) in the workflow. So a proper implementation is not possible at this point, but one could probably still add some helper functions instead.

As the input/output "name" attributes don't break the code and the workflows can still be executed either way, we still already keep them in the PWD for now and postpone a proper implementation for later.