fix(registry): pass element id to delete() on re-register, not weakref wrapper id

[xodn348]

Summary

Fixes a bug in UIElementRegistry.register() where delete() was called with the Python id of the weakref.ref wrapper rather than the id of the wrapped UIElement. Because delete() derives registered_python_id by dereferencing the weakref and calling id() on the result, the two ids never match, so delete() always returns early — skipping the ctx.function_registry.delete(namespace=object_id) call that is supposed to clean up function-registry entries for the old element on cell re-run.

The fix is a single-character change: id(self._objects[object_id]) → id(self._objects[object_id]()).

Closes #9551

📝 Summary

Fixes a bug in UIElementRegistry.register() where delete() was called with the Python id of the weakref.ref wrapper rather than the id of the wrapped UIElement. Because delete() derives registered_python_id by dereferencing the weakref and calling id() on the result, the two ids never match, so delete() always returns early — skipping the ctx.function_registry.delete(namespace=object_id) call that is supposed to clean up function-registry entries for the old element when a cell is re-run with the same UI element id.

The fix is a one-character change: dereference the weakref before calling id():

# Before (buggy):
self.delete(object_id, id(self._objects[object_id]))
# id() of a weakref.ref object ≠ id() of the element it wraps.
# delete()'s guard always fails → function_registry.delete() is never called.

# After (fixed):
self.delete(object_id, id(self._objects[object_id]()))
# id() of the dereferenced element. Guard succeeds → cleanup runs correctly.
# If the referent was GC'd, id(None) is passed; delete()'s None-path still
# removes the entry as intended.

📋 Pre-Review Checklist

• This change was discussed through an issue (Json cache containing error prevents correct notebook execution #9551, reported 2026-05-14).
• Any AI generated code has been reviewed line-by-line by the human PR author, who stands by it.

✅ Merge Checklist

• I have read the contributor guidelines.
• Tests have been added for the changes made.

Local verification

$ cd /tmp/marimo

# pre-commit run on touched files
$ pre-commit run --files marimo/_plugins/ui/_core/registry.py tests/_plugins/ui/_core/test_registry.py
typos....................................................................Passed
ruff check...............................................................Passed
ruff format..............................................................Passed

# Registry-specific tests (including new regression test)
$ uv run --group test pytest tests/_plugins/ui/_core/test_registry.py -v --tb=short
platform linux -- Python 3.13.12, pytest-9.0.3
collected 9 items

tests/_plugins/ui/_core/test_registry.py::test_cached_element_still_registered PASSED
tests/_plugins/ui/_core/test_registry.py::test_resolve_simple_noop PASSED
tests/_plugins/ui/_core/test_registry.py::test_resolve_array_noop PASSED
tests/_plugins/ui/_core/test_registry.py::test_resolve_lens_array PASSED
tests/_plugins/ui/_core/test_registry.py::test_resolve_lens_nested PASSED
tests/_plugins/ui/_core/test_registry.py::test_lens_not_bound PASSED
tests/_plugins/ui/_core/test_registry.py::test_parent_bound_to_view PASSED
tests/_plugins/ui/_core/test_registry.py::test_register_passes_element_id_not_weakref_id PASSED
tests/_plugins/ui/_core/test_registry.py::test_dont_delete_element_with_wrong_python_id PASSED
9 passed in 0.42s

# Broader CI-mirror scope (pytest tests/ -k "not test_cli")
$ uv run --group test pytest tests/_plugins/ui/ tests/_runtime/ -q -k "not test_cli" --tb=short
1831 passed, 765 skipped, 7 xfailed, 8 xpassed in 86s

=== LOCAL_TEST_PASSED ===

Risk

The changed line is inside the if object_id in self._objects: branch of register(), which only fires on cell re-run where the same UI element id is re-registered. The existing test_dont_delete_element_with_wrong_python_id test (which deliberately passes python_id=-1) is unaffected. No public API is changed.

[xodn348]

I have read the CLA Document and I hereby sign the CLA

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
marimo-docs	Ready	Preview, Comment	May 21, 2026 9:26am

[github-actions]

Thank you for your submission, we really appreciate it. Like many open-source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution. You can sign the CLA by just posting a Pull Request Comment same as the below format.

---

I have read the CLA Document and I hereby sign the CLA

---

_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}

[akshayka]

@xodn348 did you verify that this actually fixes the bug in #9551?

[mscolnick]

i do think #9551 is a different issue completely