[Question] Creating a list from node outputs

[ligerzero-ai]

wf.ISIF2_structures = collect_structure_from_nodejob(wf.ISIF2_job.outputs.vasp_output, energy_diff_threshold=0.2, job_name="ISIF2")
wf.ISIF5_structures = collect_structure_from_nodejob(wf.ISIF5_job.outputs.vasp_output, energy_diff_threshold=0.2, job_name="ISIF5")
wf.ISIF7_structures = collect_structure_from_nodejob(wf.ISIF7_job.outputs.vasp_output, energy_diff_threshold=0.2, job_name="ISIF7")
wf.ASSYST_base_structures = std.AppendToList(new_element=wf.ISIF2_structures.outputs.filtered_structures)
wf.ASSYST_base_structures = std.AppendToList(wf.ASSYST_base_structures, new_element=wf.ISIF5_structures.outputs.filtered_structures)
wf.ASSYST_base_structures = std.AppendToList(wf.ASSYST_base_structures, new_element=wf.ISIF7_structures.outputs.filtered_structures)

it complains:

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[5], line 53
     51 wf.ISIF7_structures = collect_structure_from_nodejob(wf.ISIF7_job.outputs.vasp_output, energy_diff_threshold=0.2, job_name="ISIF7")
     52 wf.ASSYST_base_structures = std.AppendToList(new_element=wf.ISIF2_structures.outputs.filtered_structures)
---> 53 wf.ASSYST_base_structures = std.AppendToList(wf.ASSYST_base_structures, new_element=wf.ISIF5_structures.outputs.filtered_structures)
     54 wf.ASSYST_base_structures = std.AppendToList(wf.ASSYST_base_structures, new_element=wf.ISIF7_structures.outputs.filtered_structures)
     55 wf.pull()

File /cmmc/ptmp/hmai/mambaforge/envs/pyiron_workflow/lib/python3.12/site-packages/pyiron_workflow/nodes/composite.py:404, in Composite.__setattr__(self, key, node)
    402     super().__setattr__(key, node)
    403 elif isinstance(node, Node):
--> 404     self.add_child(node, label=key)
    405 elif (
    406     isinstance(node, type)
    407     and issubclass(node, Node)
    408     and key in self.children.keys()
    409 ):
    410     # When a class is assigned to an existing node, try a replacement
    411     self.replace_child(key, node)

File /cmmc/ptmp/hmai/mambaforge/envs/pyiron_workflow/lib/python3.12/site-packages/pyiron_workflow/nodes/composite.py:286, in Composite.add_child(self, child, label, strict_naming)
    281     raise TypeError(
    282         f"Only new {Node.__name__} instances may be added, but got "
    283         f"{type(child)}."
    284     )
    285 self._cached_inputs = None  # Reset cache after graph change
--> 286 return super().add_child(child, label=label, strict_naming=strict_naming)

File /cmmc/ptmp/hmai/mambaforge/envs/pyiron_workflow/lib/python3.12/site-packages/pyiron_workflow/mixin/semantics.py:270, in SemanticParent.add_child(self, child, label, strict_naming)
    268     pass
    269 else:
--> 270     label = self._get_unique_label(label, strict_naming)
    272     if self._this_child_is_already_at_a_different_label(child, label):
    273         self.children.inv.pop(child)

File /cmmc/ptmp/hmai/mambaforge/envs/pyiron_workflow/lib/python3.12/site-packages/pyiron_workflow/mixin/semantics.py:315, in SemanticParent._get_unique_label(self, label, strict_naming)
    313 if label in self.child_labels:
    314     if strict_naming:
--> 315         raise AttributeError(
    316             f"{label} is already the label for a child. Please remove it "
    317             f"before assigning another child to this label."
    318         )
    319     else:
    320         label = self._add_suffix_to_label(label)

AttributeError: ASSYST_base_structures is already the label for a child. Please remove it before assigning another child to this label.

How am I supposed do this? I want to collate the structures in a single list.

[ligerzero-ai]

FWIW, I have already tried:

wf.ASSYST_base_structures = collect_structures(node_list = [wf.ISIF2_job,
                                                             wf.ISIF5_job,
                                                             wf.ISIF7_job],
                                                energy_diff_threshold = 0.1,
                                                job_names=[f"{os.getcwd()}/{structure_folder}/ISIF2",
                                                           f"{os.getcwd()}/{structure_folder}/ISIF5",
                                                           f"{os.getcwd()}/{structure_folder}/ISIF7"])

but then it complains about trying to do something with NotData, which I assume means that the nodes have not run.

[ligerzero-ai]

Using:

i2l = pwf.inputs_to_list(3)
wf.ISIF_vaspoutputs = i2l([wf.ISIF2_job.outputs.vasp_output, wf.ISIF5_job.outputs.vasp_output, wf.ISIF7_job.outputs.vasp_output])

yields

ReadinessError: InputsToList3 received a run command but is not ready. The node should be neither running nor failed, and all input values should conform to type hints.
InputsToList3 readiness: False
STATE:
running: False
failed: False
INPUTS:
item_0 ready: True
item_1 ready: False
item_2 ready: False

[liamhuber]

So the exception is perfectly expected -- you can't assign new nodes to the same label. MWE:

import pyiron_workflow as pwf
wf = pwf.Workflow("label_is_taken")
wf.some_label = pwf.standard_nodes.UserInput(42)

try:
    wf.some_label = pwf.standard_nodes.UserInput(43)
except AttributeError as e:
    print("Everything is proceeding as I have foreseen --", e)
>>> Selection deleted
import pyiron_workflow as pwf
wf = pwf.Workflow("label_is_taken")
wf.some_label = pwf.standard_nodes.UserInput(42)

try:
    wf.some_label = pwf.standard_nodes.UserInput(43)
except AttributeError as e:
    print("Everything is proceeding as I have foreseen --", e)
>>> Everything is proceeding as I have foreseen -- some_label is already the label for a child. Please remove it before assigning another child to this label.

I want to collate the structures in a single list.

Good question/goal.

AppendToList can indeed be used to achieve this, but you only instantiate the node once, loop its own output back on itself, and then make multiple connections to the new_element input. This is super dependent on cyclic flows and execution signal management. Since we plan to change things so that cyclicity is only possible inside a While node and inputs only accept one connection, this is not a future-safe way of proceeding and I don't recommend it.

Do you know at the time of writing the graph how many of these structures you will have -- i.e. is it explicitly ISIF(2/5/7), or is it ISIF{N} where N is run-time input from the user? If it's fixed at the graph level, you can accomplish this super easily with an inputs-to-list node:

import pyiron_workflow as pwf

wf = pwf.Workflow("collect_fixed_set")
wf.d1 = pwf.standard_nodes.UserInput(10)
wf.d2 = pwf.standard_nodes.UserInput(20)
wf.d3 = pwf.standard_nodes.UserInput(30)
wf.as_list = pwf.inputs_to_list(  # n: int, /, *node_args, ...
    4,
    wf.d3,
    25,  # Can mix and match data and connections, as usual
    wf.d2,
    wf.d1
)
wf()
>>> {'as_list__list': [30, 25, 20, 10]}

I.e. we specify that we're collecting 4 things (under the hood this is dynamically subclassing a generic list node parent class to a subclass with fixed IO -- 4 inputs), then give input connections/data for where to get those things.

If you need to do it where the length of the list is variable at run-time, then I think we need to figure out how to express your action as a "body node" and stick it in a for loop node. If the length isn't variable the inputs-to-list node is cleaner, IMO.

I looked on pyiron_nodes for the assyst nodes but didn't see anything -- is this available somewhere I can read it? It's not necessary, but if we need to go to making a body node macro I won't be able to give any help on formulating it without access to the node package details.

[liamhuber]

Using:

i2l = pwf.inputs_to_list(3)
wf.ISIF_vaspoutputs = i2l([wf.ISIF2_job.outputs.vasp_output, wf.ISIF5_job.outputs.vasp_output, wf.ISIF7_job.outputs.vasp_output])

yields

ReadinessError: InputsToList3 received a run command but is not ready. The node should be neither running nor failed, and all input values should conform to type hints.
InputsToList3 readiness: False
STATE:
running: False
failed: False
INPUTS:
item_0 ready: True
item_1 ready: False
item_2 ready: False

Yeah, the problem here is [wf.ISIF2_job.outputs.vasp_output, wf.ISIF5_job.outputs.vasp_output, wf.ISIF7_job.outputs.vasp_output]. Python doesn't know how to make a list of the output channels and so the node tries to treat the whole list as a single piece of input data instead of a bunch on input connections. You were absolutely on the right track here though, and I think my co-temporaneously written example should clear it up.