Fix node registration

[liamhuber]

Adds the node registration method back, although at the moment it's just called in Creator.__init__ so you don't see it.

TODO:

• Tests
• Think about what API you want to support -- should we also handle lists of nodes, or modules containing a nodes attribute (as long as it has the right type), or URLs? (EDIT: Right now I'm still only accepting string import paths to modules that have a nodes: list[Node] property, but the Creator.register method has good encapsulation of the logic for checking whether (a) newly proposed packages are just a re-registration of an existing package, and (b) whether the proposed package will actually be usable for generating nodes, so extending this in the future to other node package identifiers should be easy. For now I want to prioritize getting any fix in so we at least have some way to register things!)
• Stop registering atomistics by default and update the demo and docs as needed
• More gracefully handle re-registering things -- if it's the same nodes being set to the same domain name, just let it pass silently! This may be helpful when...
• We encourage somewhere in the docs that macros register the necessary packages inside the macro definition whenever the macro uses .create... (i.e. unless the macro is part of the same .py file that all its children are coming from, in which case they should be available in-scope without .create

Outside of scope:

• Actually supporting all manner of registration types (like a URL) -- I just want infrastructure that will be flexible enough to handle that in the future, not to get it all solved now!

Closes #51

[github-actions]

👈 Launch a binder notebook on branch pyiron/pyiron_workflow/fix_node_registration

[github-actions]

Pull Request Test Coverage Report for Build 6700105758

• 66 of 67 (98.51%) changed or added relevant lines in 4 files are covered.
• 53 unchanged lines in 7 files lost coverage.
• Overall coverage decreased (-0.05%) to 87.774%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
pyiron_workflow/interfaces.py	41	42	97.62%

Files with Coverage Reduction	New Missed Lines	%
node_library/standard.py	1	95.65%
pyiron_workflow/node_library/standard.py	2	92.31%
channels.py	7	91.35%
function.py	8	91.57%
interfaces.py	8	87.18%
composite.py	9	91.36%
node_library/atomistics.py	18	0.0%

Totals
Change from base Build 6620662485:	-0.05%
Covered Lines:	3209
Relevant Lines:	3656

---

💛 - Coveralls

[codacy-production]

Coverage summary from Codacy

See diff coverage on Codacy

Coverage variation	Diff coverage
❌ -1.15% (target: -1.00%)	81.43%

Coverage variation details

	Coverable lines	Covered lines	Coverage
Common ancestor commit (091abf1)	1802	1501	83.30%
Head commit (6189e02)	1843 (+41)	1514 (+13)	82.15% (-1.15%)

Coverage variation is the difference between the coverage for the head and common ancestor commits of the pull request branch: <coverage of head commit> - <coverage of common ancestor commit>

Diff coverage details

	Coverable lines	Covered lines	Diff coverage
Pull request (#53)	70	57	81.43%

Diff coverage is the percentage of lines that are covered by tests out of the coverable lines that the pull request added or modified: <covered lines added or modified>/<coverable lines added or modified> * 100%

See your quality gate settings    Change summary preferences

[liamhuber]

The 3.9 tests break:

...
TypeError: unsupported operand type(s) for |: 'type' and 'type'

The above exception was the direct cause of the following exception:
...
ValueError: The package identifier is pyiron_workflow.node_library.standard is not valid. Please ensure it is an importable module with a list of Node objects stored in the variable `nodes`.

At first I was confused why importing from pyiron_workflow.node_library.standard would not work, but scrolling up the trace it's actually breaking when trying to process the type hint. This makes sense since 3.9 doesn't handle | but requires typing.Union, hence why we brute-force skip almost all the tests until the version is >= 3.10.

This failure happens in:

ERROR: executors.test_cloudprocesspool (unittest.loader._FailedTest)
ERROR: test_macro (unittest.loader._FailedTest)
ERROR: test_util (unittest.loader._FailedTest)
ERROR: test_workflow (unittest.loader._FailedTest)

In each case starting with

ImportError: Failed to import test module: [test module name]

So I guess I've somehow added a "bad" type hint somewhere before we get to the skipif version guard.

[liamhuber]

The problem is when Creator tries to register the standard node package, which makes enough sense since we parse the type hints there and definitely use the new style.

What doesn't yet make sense is why there's a creator existing in, e.g., the tests/unit/executors/test_cloudprocesspool? And since the standard registration happens in __init__ is there really one of these being initialized, or is this some strange side-effect of the Creator being a Singleton?

[liamhuber]

Ah, ok, I think maybe because right in pyiron_workflow.__init__ we import Workflow, which inherits from Composite, which instantiates a Creator as a class attribute.

We no longer get away with this because now we typing.get_type_hints as part of the decorator creating new node classes.

Alright, so all perfectly sensible now how to fix it.

[liamhuber]

Notebook failure is expected since I stopped registering atomistics by default and broke the scatter plot node over to its own package.

[codacy-production]

Coverage summary from Codacy

See diff coverage on Codacy

Coverage variation	Diff coverage
❌ -1.15% (target: -1.00%)	81.43%

Coverage variation details

	Coverable lines	Covered lines	Coverage
Common ancestor commit (091abf1)	1802	1501	83.30%
Head commit (23b8827)	1843 (+41)	1514 (+13)	82.15% (-1.15%)

Coverage variation is the difference between the coverage for the head and common ancestor commits of the pull request branch: <coverage of head commit> - <coverage of common ancestor commit>

Diff coverage details

	Coverable lines	Covered lines	Diff coverage
Pull request (#53)	70	57	81.43%

Diff coverage is the percentage of lines that are covered by tests out of the coverable lines that the pull request added or modified: <covered lines added or modified>/<coverable lines added or modified> * 100%

See your quality gate settings    Change summary preferences

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[liamhuber]

Codacy Coverage Report — -1.15% coverage variation (-1% target)

Coverage is just angry because pyiron_workflow.node_library.atomistics is no longer getting caught in the unit tests, but that's kind of the point; the unit tests are meant to test the infrastructure not the nodes we build on top of it. When the nodes get spun off to their own repos they can have their own tests.