Fix docs: non-string annotations are converted, not rejected

[Builder106]

Problem

Several docstrings and docs claim that set_annotations=True requires string annotations and raises NonStringAnnotationError when they are not strings. For example, concatenate_functions:

All annotations must be strings; otherwise, a NonStringAnnotationError is raised.

This has been inaccurate since #67 ("do not require string annotations from users"). The current behavior is to convert non-string annotations to their string representation (via ensure_annotations_are_strings → _get_str_repr). In fact, NonStringAnnotationError is never raised anywhere in the source, is not exported from the package, and has no tests:

# functions WITHOUT `from __future__ import annotations`, so annotations are
# real type objects, not strings
def a() -> int: return 1
def b(a: int) -> float: return a + 1.5

f = concatenate_functions({"a": a, "b": b}, targets="b", set_annotations=True)
# No error. inspect.signature(f).return_annotation == "float"  (the string)

Changes

• Updated the stale docstrings in dag.py (FunctionExecutionInfo, concatenate_functions, _create_combined_function_from_dag, _create_concatenated_function, create_execution_info) to describe the actual convert behavior and removed the false NonStringAnnotationError Raises: entries.
• Updated the exception table in docs/api.md and the set_annotations note in docs/usage_patterns.ipynb.
• Added a regression test in tests/test_without_string_annotations.py asserting that concatenate_functions(..., set_annotations=True) converts non-string annotations rather than raising.

Verification

• pytest: 195 passed
• ruff check / ruff format --check: clean
• ty check: clean

One question for maintainers

NonStringAnnotationError is now effectively orphaned — defined in dags/exceptions.py but never raised, not exported, and untested. I left the class in place so this PR stays documentation-only and doesn't change the (informal) exception surface. Happy to either remove it in this PR or open a follow-up if you'd prefer to deprecate/drop it.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB