Retrospective while

[liamhuber]

Introducing a new While node in analogy to the For node. As a second-order, stateful, dynamic node, retrospective provenance gets automatically handled by the way we have individual body nodes. Cyclic loops are specified by reference to IO channel name pairs (strings, ugh...) between the body and the test, and the body and itself:

import pyiron_workflow.api as pwf
from pyiron_workflow.nodes.while_loop import while_node

wf = pwf.Workflow("my_while_loop")
wf.x0 = pwf.std.UserInput(0)
wf.limit = pwf.std.UserInput(5)
wf.step = pwf.std.UserInput(2)
wf.add_while = while_node(
    pwf.std.LessThan,  # Test
    pwf.std.Add,  # Body
    [("add", "obj")],  # body-to-test
    [("add", "obj")],  # body-to-body
    strict_condition_hint=False,  # LessThan doesn't hint it's boolean return...
    test_obj=wf.x0,
    test_other=wf.limit,
    body_obj=wf.x0,
    body_other=wf.step
)
wf.xf = wf.add_while.outputs.add  # While output maps to body output
wf()
>>> {'add_while__add': 6}

TODO:

• Debug the caching tests
• Add tests for the cyclic edge maps (disallow non-existent edges)
• Test and allow continuing from a failed state
  • On failure, we get the provenance graph leading up to the failure; that will suffice for now
• Add and test executors for the body and condition
• Docstrings
• Include in the deepdive
• (Maybe?) Replace other while examples with this (they still keep working since they're just a hard-code on signal control + some provenance logging, but since they're not encouraged I'm inclined to eliminate them...) Other than the deep dive, I only found the integration test. I actually like the integration test as an example of what we could accomplish by exploiting the separation of data and signal edges, so I'll leave it. Of course, the new While class has significantly better provenance behaviour, so it's the recommended attack for these loops.

Closes #501

@samwaseda, I guess you may be interested to keep a high-level eye here to see if you like my concrete implementation of what we're talking about over in pyiron/flowrep#24. I will naturally try to keep everything here fully compliant with what we decide there.

[github-actions]

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

[codecov]

Codecov Report

Attention: Patch coverage is 98.55072% with 2 lines in your changes missing coverage. Please review.

Project coverage is 91.87%. Comparing base (66ab17b) to head (5d9b3b4).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
pyiron_workflow/create.py	50.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #650      +/-   ##
==========================================
+ Coverage   91.61%   91.87%   +0.25%     
==========================================
  Files          33       34       +1     
  Lines        3458     3593     +135     
==========================================
+ Hits         3168     3301     +133     
- Misses        290      292       +2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[liamhuber]

@Tara-Lakshmipathy @ligerzero-ai, no action necessary on your part, just an FYI that there is a new interface for while-loops available. It's directly analogous to the existing for-loop interface. It's (hopefully) much easier to use than the old by-hand way of forcing a cyclic macro, and has the advantage of producing a full retrospective provenance graph automatically (unlike the by-hand approach which re-ran the same node repeatedly).