reregister memo in app harness

[benedikt-bartscher]

No description provided.

[greptile-apps]

Greptile Summary

This PR fixes two related bugs that caused flaky compile errors (Invalid path, unresolved $/app_components/... imports) when using AppHarness with a real pre-imported app: (1) reloading the top module on test start silently dropped every submodule page, rx.State subclass, and @rx.memo defined in submodules, and (2) nested var-memo dependencies registered lazily inside another memo's body were missed when the compiler snapshotted MEMOS before evaluating those bodies.

• AppHarness._initialize_app is split into two clear paths: real apps (app_source is None) are imported once with reload=False and their MEMOS registrations are preserved; generated apps continue to clear MEMOS and reload as before. The _registry_token field defaults to None and cleanup already guards with if self._registry_token is not None, so the early-return path is safe.
• _reregister_used_memo is called at every @rx.memo wrapper use-site (__call__, partial, __rx_var__, component_var), repopulating a cleared MEMOS from usage even in the belt-and-suspenders case.
• materialize_registered_memo_bodies runs a fixed-point loop in compile_app before the MEMOS snapshot, evaluating lazy bodies until no new memos are registered, so deeply nested var-memo dependencies are always included in compilation.

Confidence Score: 5/5

The change is safe to merge. Real-app harness tests no longer reload the app module, preserving all submodule registrations; generated-app tests are unaffected.

Both code paths in _initialize_app are clearly scoped and well-tested. The new materialize_registered_memo_bodies pre-pass is idempotent (lazy bodies cache their result after the first evaluation), and the fixed-point loop is guaranteed to terminate. The _registry_token safety gap for real-app teardown is already handled by the existing None guard. No regressions are introduced on the generated-app path.

No files require special attention.

Important Files Changed

Filename	Overview
packages/reflex-base/src/reflex_base/components/memo.py	Adds _reregister_used_memo (called at every use-site on wrappers) and materialize_registered_memo_bodies (fixed-point pre-pass that evaluates lazy bodies until MEMOS stabilizes), both correctly scoped to non-passthrough memos; logic is sound.
reflex/testing.py	Splits _initialize_app into two clearly separated paths: real apps get reload=False with MEMOS preserved (early return before RegistrationContext setup); generated apps clear MEMOS and reload as before. _registry_token defaults to None and cleanup checks for None, so the early return is safe.
reflex/compiler/compiler.py	Inserts materialize_registered_memo_bodies() immediately before the MEMOS snapshot to settle nested lazy-body registrations before compilation; minimal, targeted change.
tests/units/components/test_memo.py	Adds three targeted regression tests: re-registration on use after a clear, nested var-memo detection via materialize_registered_memo_bodies, and passthrough-auto-memo exclusion. All scenarios are well-covered.
tests/units/test_testing.py	Updates the existing real-app harness test to assert reload=False and MEMOS preservation; correctly mocks initialize_frontend_dependencies for the new code path.
news/+harness-real-app-no-reload.bugfix.md	Changelog entry describing the AppHarness real-app no-reload fix.
packages/reflex-base/news/+memo-harness-reregister.bugfix.md	Changelog entry describing the memo re-registration fix.
pyi_hashes.json	Hash update for reflex/experimental/memo.pyi reflecting the new materialize_registered_memo_bodies export.

_{Reviews (3): Last reviewed commit: "more memo harness fixes" | Re-trigger Greptile}

[codspeed-hq]

Merging this PR will not alter performance

✅ 26 untouched benchmarks
⏩ 8 skipped benchmarks¹

---

_{Comparing benedikt-bartscher:fix-memo-harness (384dec7) with main (f79d011)}

Footnotes

1. 8 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 40fed7de7d

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".