fix(pyi): incremental .pyi regen now follows the import graph by FarhanAliRaza · Pull Request #6422 · reflex-dev/reflex

FarhanAliRaza · 2026-04-29T18:22:15Z

Fixed

Fixed partial generation, before it dropped the unchanged in files.
fixed the inherited changes that were not happening before.

All Submissions:

Have you followed the guidelines stated in CONTRIBUTING.md file?
Have you checked to ensure there aren't any other open Pull Requests for the desired changed?

Type of change

Please delete options that are not relevant.

Bug fix (non-breaking change which fixes an issue)

New Feature Submission:

Does your submission pass the tests?
Have you linted your code locally prior to submission?

…l files Replace the .pyi_generator_last_run/.pyi_generator_diff sentinel files with the last commit that touched pyi_hashes.json as the regen baseline, and expand the change set along the import graph so editing a parent class also regenerates subclass stubs. Set pass_filenames: false on the pre-commit hook and rewrite scan_all's pyi_hashes.json update as a per-file evidence-based merge so partial-target runs no longer truncate the file.

codspeed-hq · 2026-04-29T18:24:35Z

Merging this PR will degrade performance by 4.29%

❌ 1 regressed benchmark
✅ 16 untouched benchmarks
⏩ 2 skipped benchmarks¹

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	`BASE`	`HEAD`	Efficiency
❌	`test_get_all_imports_single_pass[_complicated_page]`	1.6 ms	1.6 ms	-4.29%

_{Comparing FarhanAliRaza:fix-pyi-generation (4062d22) with main (76ff38c)}

2 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. ↩

greptile-apps · 2026-04-29T18:28:20Z

Greptile Summary

This PR overhauls the incremental .pyi stub regeneration pipeline: make_pyi.py is rewritten to use git log on pyi_hashes.json as the baseline commit, then expands the dirty set along the AST import graph so parent-class changes cascade to subclasses. pyi_generator.py fixes a mutable class-level written_files list and rewrites the hash-merge logic to correctly preserve unscanned entries while pruning stale ones.

Confidence Score: 4/5

Safe to merge; all findings are P2 style/hardening suggestions with no correctness-breaking defects on the happy path.

The logic is sound and well-tested. Three P2 notes: missing ast.TryStar handling (minor Python 3.11+ gap), the uncommitted-hashes loop (usability only), and the single-anchor heuristic replacing LCA (low risk with one root-level hashes file). No P0/P1 issues found.

scripts/make_pyi.py and packages/reflex-base/src/reflex_base/utils/pyi_generator.py contain all substantive logic changes.

Important Files Changed

Filename	Overview
scripts/make_pyi.py	Major rewrite: replaces SHA-file-based change detection with git-log baseline + AST import-graph expansion; ast.TryStar gap and uncommitted-hashes loop concern noted.
packages/reflex-base/src/reflex_base/utils/pyi_generator.py	Fixes mutable class-level written_files (moved to init), resolves file targets, and rewrites hash-merge logic; anchor-from-first-file regression noted.
tests/units/reflex_base/utils/pyi_generator/test_hash_merge.py	New test file covering key hash-merge scenarios: partial runs, dropped entries, missing hash file, and incremental merges. All cases appear logically correct.
.pre-commit-config.yaml	Adds pass_filenames: false so the hook determines its own change set via git — consistent with the new make_pyi.py logic.
.gitignore	Removes now-deleted .pyi_generator_last_run and .pyi_generator_diff transient files — correct cleanup.
AGENTS.md	Updates .pyi stub instructions to reflect the new incremental workflow; removes stale references to .pyi_generator_last_run.

_{Reviews (1): Last reviewed commit: "fix(pyi): regenerate stubs from pyi_hash..." | Re-trigger Greptile}

… races Add --no-update-hashes flag to pyi_generator and use it from both the top-level and per-package hatch build hooks. The shared hash file is a dev-tooling artifact; reading/writing it during parallel workspace package builds caused races and isn't needed at wheel-build time.

Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>

The hash file is workspace-scoped, so walking up from the first entry in written_files made the resolved path depend on iteration order and could latch onto a package-local file before reaching the repo root. Also warn when pyi_hashes.json exists locally but isn't tracked in git, since that forces a full regen on every run until it's committed.

…/reflex into fix-pyi-generation

adhami3310 · 2026-04-29T21:11:44Z

    modules: list = []
    root: str = ""
    current_module: Any = {}
-    written_files: list[tuple[str, str]] = []
+
+    def __init__(self) -> None:
+        """Initialize per-instance scan state."""
+        self.written_files: list[tuple[str, str]] = []


shouldn't the same be done to modules and current module?

They need to be removed. They are not used.

adhami3310 · 2026-04-29T21:12:33Z

+        if use_json and (file_paths or file_targets):
+            file_paths = list(map(Path, file_paths))


this might be a good time to take PYI hash generation into its own function

Move the pyi_hashes.json merge logic out of PyiGenerator.scan_all into a module-level _update_pyi_hashes_file() so scan_all reads as scan + ruff fixup + (optional) hash update. Remove unused class attributes (modules, root, current_module) that were never read or written, and collapse a duplicate _relative_to_pwd() call in the file walk. Also fix test_tailwind flake: the stylesheet is now written to Path.cwd() / assets so it survives AppHarness's importlib.reload() across pytest reruns (the cached __file__ pinned writes to the prior tmp_path while the compiler read from the current cwd).

masenf

i might be reading it wrong, but are we regenerating modules that are imported by modules that changed? the other fixes seem legit.

masenf · 2026-05-01T07:28:23Z

+    path.write_text(json.dumps(mapping, indent=2, sort_keys=True) + "\n")
+
+
+def _make_workspace(root: Path) -> Path:


this should be a fixture that returns the workspace path based on the tmp_path fixture

FarhanAliRaza · 2026-05-01T10:09:01Z

For example if react_player is changed, in the main branch now it only generates react_player.pyi because it only sees this in git diff. But it should have also changed audio.pyi and video.pyi because they import react player amd inherit from it . Before partial generation missed it. Only full regneration catches that.

So we are regenerating that imports the changed modules.

FarhanAliRaza requested a review from a team as a code owner April 29, 2026 18:22

greptile-apps Bot reviewed Apr 29, 2026

View reviewed changes

Comment thread scripts/make_pyi.py

Comment thread scripts/make_pyi.py

Comment thread packages/reflex-base/src/reflex_base/utils/pyi_generator.py Outdated

FarhanAliRaza and others added 4 commits April 29, 2026 23:55

Update scripts/make_pyi.py

2844d73

Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>

Merge branch 'fix-pyi-generation' of https://github.com/FarhanAliRaza…

502a677

…/reflex into fix-pyi-generation

adhami3310 reviewed Apr 29, 2026

View reviewed changes

masenf reviewed May 1, 2026

View reviewed changes

test(pyi): use shared workspace fixture in hash-merge tests

4062d22

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(pyi): incremental .pyi regen now follows the import graph#6422

fix(pyi): incremental .pyi regen now follows the import graph#6422
FarhanAliRaza wants to merge 7 commits intoreflex-dev:mainfrom
FarhanAliRaza:fix-pyi-generation

FarhanAliRaza commented Apr 29, 2026

Uh oh!

codspeed-hq Bot commented Apr 29, 2026 •

edited

Loading

Uh oh!

greptile-apps Bot commented Apr 29, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

adhami3310 Apr 29, 2026

Uh oh!

FarhanAliRaza Apr 30, 2026 •

edited

Loading

Uh oh!

adhami3310 Apr 29, 2026

Uh oh!

FarhanAliRaza Apr 30, 2026

Uh oh!

masenf left a comment

Uh oh!

masenf May 1, 2026

Uh oh!

FarhanAliRaza commented May 1, 2026 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

		if use_json and (file_paths or file_targets):
		file_paths = list(map(Path, file_paths))

		path.write_text(json.dumps(mapping, indent=2, sort_keys=True) + "\n")


		def _make_workspace(root: Path) -> Path:

Conversation

FarhanAliRaza commented Apr 29, 2026

All Submissions:

Type of change

New Feature Submission:

Uh oh!

codspeed-hq Bot commented Apr 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merging this PR will degrade performance by 4.29%

Performance Changes

Footnotes

Uh oh!

greptile-apps Bot commented Apr 29, 2026

Greptile Summary

Confidence Score: 4/5

Important Files Changed

Uh oh!

Uh oh!

Uh oh!

Uh oh!

adhami3310 Apr 29, 2026

Choose a reason for hiding this comment

Uh oh!

FarhanAliRaza Apr 30, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

adhami3310 Apr 29, 2026

Choose a reason for hiding this comment

Uh oh!

FarhanAliRaza Apr 30, 2026

Choose a reason for hiding this comment

Uh oh!

masenf left a comment

Choose a reason for hiding this comment

Uh oh!

masenf May 1, 2026

Choose a reason for hiding this comment

Uh oh!

FarhanAliRaza commented May 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

codspeed-hq Bot commented Apr 29, 2026 •

edited

Loading

FarhanAliRaza Apr 30, 2026 •

edited

Loading

FarhanAliRaza commented May 1, 2026 •

edited

Loading