[Suggestion] Compromise between flowrep and PWD #19
Description
I guess I found an acceptable compromise between @JNmpi and @liamhuber to fill the gap between flowrep and @jan-janssen's PWD.
I would suggest to delete inputs and outputs from flowrep, and remove invisible function arguments, in order not to have to inspect the actual functions. Let's take the following workflow as an example:
def operation(x: float, y: float) -> tuple[float, float]:
return x + y, x - y
def add(x: float = 2.0, y: float = 1) -> float:
return x + y
def multiply(x: float, y: float = 5) -> float:
return x * y
def example_workflow(a=10, b=20):
c, d = operation(a, b)
e = add(c, y=d)
f = multiply(e)
return f{
"nodes": {
"operation_0": {
"function": {
"module": operation.__module__,
"qualname": operation.__qualname__,
"version": "not_defined",
},
"type": "Function",
},
"add_0": {
"function": {
"module": add.__module__,
"qualname": add.__qualname__,
"version": "not_defined",
},
"type": "Function",
},
"multiply_0": {
"function": {
"module": multiply.__module__,
"qualname": multiply.__qualname__,
"version": "not_defined",
},
"type": "Function",
},
},
"edges": [
("inputs.a", "operation_0.inputs.0"),
("inputs.b", "operation_0.inputs.1"),
("operation_0.outputs.output[0]", "add_0.inputs.0"),
("operation_0.outputs.output[1]", "add_0.inputs.y"),
("add_0.outputs.output", "multiply_0.inputs.0"),
("multiply_0.outputs.output", "outputs.f"),
],
"label": "example_macro",
"type": "Workflow",
}I put module, qualname etc., but I guess I would even want to have the actual python functions instead. Anyway, this dictionary can be passed to other tools such as semantikon to retrieve the full IO information. This representation has the following advantages
- It does not contain IO (as @JNmpi wished)
- No inspection required to create this representation (as @liamhuber wished)
- 0, 1 etc. for the input arguments are actually even required for some functions, such as
range - This creates a clear separation between
semantikonandflowrep, asflowrepdoes not need to know anything about the underlying functions
Let me know what you think