Workflow constructor raises TypeError when input_spec is an instance of SpecInfo

[NicolasGensollen]

The documentation of the Workflow constructor says that input_spec can be a list of strings, a SpecInfo, or a BaseSpec instance:

pydra/pydra/engine/core.py

Line 807 in 219c721

input_spec: ty.Optional[ty.Union[ty.List[ty.Text], SpecInfo, BaseSpec]] = None,

However, when providing a SpecInfo I get a TypeError: unhashable type: 'SpecInfo'.

Looking at the code, it seems like this case isn't handled properly such that we end up iterating over the SpecInfo instance (line 860), triggering the error described above:

pydra/pydra/engine/core.py

Lines 843 to 863 in 219c721

	if input_spec:
	if isinstance(input_spec, BaseSpec):
	self.input_spec = input_spec
	else:
	self.input_spec = SpecInfo(
	name="Inputs",
	fields=[("_graph_checksums", ty.Any)]
	+ [
	(
	nm,
	attr.ib(
	type=ty.Any,
	metadata={
	"help_string": f"{nm} input from {name} workflow"
	},
	),
	)
	for nm in input_spec
	],
	bases=(BaseSpec,),
	)

It could be that SpecInfo was supposed to inherit from BaseSpec, but it doesn't seem to be the case atm:

pydra/pydra/engine/specs.py

Line 65 in 219c721

class SpecInfo:

[satra]

@NicolasGensollen - this seems like an oversight and is a bug.

would you like to send a PR to resolve this ? i think the only specific check that would be needed is that the SpecInfo object has BaseSpec as it's bases, similar to the constructor here:

pydra/pydra/engine/core.py

Line 862 in 219c721

bases=(BaseSpec,),

[NicolasGensollen]

@satra Sure I'll send a PR! Thanks for the quick reply ! 👍