Fix selectbox and multiselect clearing selections for custom objects by lukasmasuch · Pull Request #13648 · streamlit/streamlit

lukasmasuch · 2026-01-20T15:13:40Z

Describe your changes

Fixes issue #13646 where st.multiselect() and st.selectbox clears user selections after each script rerun when using custom class objects without __eq__ implementation and a format_func parameter. This was a regression introduced in PR #13448.

Root Cause: The validation logic compared multiselect values using ==, which falls back to identity comparison for objects without __eq__. Since register_widget() deepcopies widget values, the new instances fail identity comparison with original options, causing valid selections to be filtered out.

Solution: Changed validation to compare values using format_func() by their formatted string representation instead of using ==. This is more robust and works correctly regardless of whether custom classes implement __eq__.

Testing Plan

Unit Tests: Added 4 comprehensive test cases covering custom objects without __eq__, partial matches, objects with __str__, and edge cases where format_func fails on incompatible types
E2E Tests: Added test_multiselect_custom_objects_without_eq to verify selections persist across script reruns with custom class objects
Backward Compatibility: The format_func parameter has a default value of str, so existing code continues to work without changes

Contribution License Agreement

By submitting this pull request you agree that all contributions to this project are made under the Apache 2.0 license.

…#13646) When using st.multiselect() with custom class objects and a format_func, selections would be cleared after each script rerun in v1.53.0. This was a regression introduced by PR #13448. The issue occurred because register_widget() deepcopies widget values, and the validation logic used == comparison which falls back to identity checks for objects without __eq__. Deepcopied instances fail identity comparison, causing valid selections to be filtered out. The fix changes the validation to use format_func() for comparisons instead, comparing by formatted string representation. This is more robust and works correctly with custom objects regardless of __eq__ implementation. The solution also includes error handling for edge cases where format_func fails on incompatible value types. - Modified validate_and_sync_multiselect_value_with_options() to accept and use format_func parameter - Added comprehensive unit tests for custom objects with and without __eq__ - Added edge case test for format_func failures - Added E2E test to verify selections persist across script reruns Co-Authored-By: Claude <[email protected]>

snyk-io · 2026-01-20T15:13:53Z

✅ Snyk checks have passed. No issues have been found so far.

Status	Scanner	Critical	High	Medium	Low	Total (0)
✅	Open Source Security	0	0	0	0	0 issues
✅	Licenses	0	0	0	0	0 issues

💻 Catch issues earlier using the plugins for VS Code, JetBrains IDEs, Visual Studio, and Eclipse.

github-actions · 2026-01-20T15:13:56Z

✅ PR preview is ready!

Name	Link
📦 Wheel file	https://core-previews.s3-us-west-2.amazonaws.com/pr-13648/streamlit-1.53.0-py3-none-any.whl
📦 `@streamlit/component-v2-lib`	Download from artifacts
🕹️ Preview app	pr-13648.streamlit.app (☁️ Deploy here if not accessible)

Copilot

Pull request overview

This PR fixes a regression where st.multiselect() would clear user selections after each script rerun when using custom class objects without __eq__ implementation. The root cause was that validation logic used == comparison, which falls back to identity comparison for objects without __eq__, causing deepcopied widget values to fail validation.

Changes:

Modified validation logic to compare values using their formatted string representation via format_func instead of direct equality comparison
Added comprehensive unit tests covering custom objects without __eq__, partial matches, and edge cases
Added E2E test to verify selections persist across script reruns with custom objects

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
`lib/streamlit/elements/lib/options_selector_utils.py`	Updated `validate_and_sync_multiselect_value_with_options` to compare values using `format_func` instead of `==`, with exception handling for format failures
`lib/streamlit/elements/widgets/multiselect.py`	Passed `format_func` parameter to validation function
`lib/tests/streamlit/elements/lib/options_selector_utils_test.py`	Added 4 comprehensive unit tests covering custom objects scenarios and edge cases
`e2e_playwright/st_multiselect.py`	Added test app with custom class objects without `__eq__`
`e2e_playwright/st_multiselect_test.py`	Added E2E test verifying selections persist across reruns

github-actions · 2026-01-20T15:23:54Z

📉 Frontend coverage change detected

The frontend unit test (vitest) coverage has decreased by 0.0000%

Current PR: 86.4600% (13347 lines, 1807 missed)
Latest develop: 86.4600% (13347 lines, 1807 missed)

✅ Coverage change is within normal range.

📊 View detailed coverage comparison

cursor

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}

cursor · 2026-01-20T15:25:51Z

lib/streamlit/elements/lib/options_selector_utils.py

+            # can be formatted successfully.
+            continue
+
+        if formatted_value in formatted_options_set:


Validation and serialization use inconsistent comparison methods

Low Severity

The new validation uses format_func comparison while the existing serialize method uses options.index() (which relies on ==). Since register_widget always deepcopies values, custom objects without __eq__ can pass validation but fail serialization. When serialization fails, the fallback assumes the value is a string (per the comment at line 129-130), but it's actually a custom object. This inconsistency can cause incorrect serialization if widget_state.value_changed is True (e.g., when using st.session_state[key] = value).

Additional Locations (1)

lib/streamlit/elements/widgets/multiselect.py#L125-L131

github-actions · 2026-01-20T16:08:06Z

Summary

This PR fixes issue #13646 where st.multiselect() was clearing user selections after script reruns when using custom class objects without __eq__ implementation and a format_func parameter. This was a regression introduced in PR #13448.

Root Cause: The validation logic in validate_and_sync_multiselect_value_with_options compared multiselect values using ==, which falls back to identity comparison for objects without __eq__. Since register_widget() deepcopies widget values, the new instances fail identity comparison with original options, causing valid selections to be filtered out.

Solution: Changed the validation to compare values using format_func() by their formatted string representation instead of using == directly.

Code Quality

The code changes are well-structured, minimal, and follow existing patterns in the codebase.

`lib/streamlit/elements/lib/options_selector_utils.py`

The new format_func parameter has a sensible default value (str), maintaining backward compatibility
The use of a set for formatted_options_set provides O(1) lookup, which is an improvement over the previous O(n) lookup via index_()
The exception handling with except Exception at line 385 is acceptable because:
- It only affects the specific value being validated (fails safe by filtering out the value)
- The comment clearly explains the edge case (format_func failing on incompatible types like raw strings from previous sessions)
- The # noqa: S112 annotation shows awareness of this design choice

`lib/streamlit/elements/widgets/multiselect.py`

Simple one-line change that passes format_func to the validation function - clean integration

Potential Edge Case (not a blocking issue)

If two different options have the same format_func output (e.g., two objects where format_func returns "same_label"), both values would be considered valid. This is consistent with the existing multiselect behavior where duplicate formatted options are allowed, so this is not a regression.

Test Coverage

Excellent test coverage. The changes are thoroughly tested with both unit tests and E2E tests.

Unit Tests (`lib/tests/streamlit/elements/lib/options_selector_utils_test.py`)

Four comprehensive test cases added in TestValidateMultiselectWithCustomObjects:

test_custom_objects_without_eq_using_format_func - Tests the primary fix scenario
test_custom_objects_partial_match_with_format_func - Tests filtering when some options are removed
test_default_format_func_with_custom_str - Tests objects with __str__ using default format_func
test_format_func_failure_filters_out_value - Tests edge case where format_func fails on incompatible types

All tests include:

Docstrings explaining the purpose (following best practices)
Both positive assertions (values kept) and negative/boundary assertions (values filtered, needs_reset flag)
Clear comments explaining the test scenario

E2E Tests (`e2e_playwright/st_multiselect_test.py`)

The new test_multiselect_custom_objects_without_eq test follows best practices:

Uses expect for auto-wait assertions (not assert)
Uses label-based locators via get_multiselect() helper
Has descriptive name and docstring with issue reference
Verifies initial state before interaction
Tests single selection, then multiple selections
Includes both value verification (expect_text) and UI verification (tag count)

The test app changes in e2e_playwright/st_multiselect.py properly:

Define the CustomOption class without __eq__ (with appropriate # noqa: B903)
Create new option instances on each script run (simulating the bug trigger)
Use format_func to display labels

Backwards Compatibility

Fully backward compatible. The format_func parameter has a default value of str, so:

Existing code continues to work without changes
The default behavior for simple types (strings, numbers) remains the same
Only affects multiselect widgets that use custom objects without __eq__

Security & Risk

Low risk. This is a targeted bug fix with:

No changes to user-facing APIs
No changes to data handling or serialization
The exception handling is fail-safe (invalid values are filtered out, not propagated)
The # noqa: S112 annotation on the broad exception is appropriate for this validation use case

Recommendations

No blocking issues. Minor suggestions for consideration:

Documentation consideration (optional): The PR description is excellent and explains the fix well. Consider adding a note about this fix in the changelog if Streamlit maintains one for bug fixes.
The MULTISELECT_COUNT update: The count was correctly updated from 20 to 21 to reflect the new test widget.

Verdict

APPROVED: This is a well-implemented bug fix that properly addresses the regression. The code changes are minimal and targeted, the test coverage is comprehensive (4 unit tests + 1 E2E test), and the fix is fully backward compatible. The PR follows Streamlit's coding standards and testing best practices.

This is an automated AI review. Please verify the feedback and use your judgment.

Address PR review comment: The validation and serialization methods were using inconsistent comparison approaches. Validation used format_func while serialization used options.index() which relies on ==. This could cause incorrect serialization for custom objects without __eq__ when widget_state.value_changed is True. Changes: - Updated MultiSelectSerde to accept and store format_func parameter - Modified serialize() to use format_func for finding formatted options instead of options.index() which relies on == comparison - Added test for serializing deepcopied custom objects - Updated existing test to pass format_func to serde Co-Authored-By: Claude <[email protected]>

lib/streamlit/elements/widgets/multiselect.py

…bjects This extends the fix from #13646 to selectbox as well. The same issue existed where custom objects without __eq__ would fail validation and serialization after deepcopy because the old implementation used index_() which relies on == comparison. Changes: - Update SelectboxSerde to use format_func for serialization - Update validate_and_sync_value_with_options to use format_func - Pass format_func from selectbox widget to serde and validation - Add tests for selectbox with deepcopied custom objects Co-Authored-By: Claude <[email protected]>

For Enum values, use the original index_() approach which uses == comparison. This correctly handles enum class identity - enums from different classes (e.g., after script rerun) should NOT be considered equal, which is important for enum coercion to work correctly when coercion is disabled. For non-Enum values, use format_func comparison. This handles custom objects without __eq__ where widget values are deepcopied and the deepcopied instances would fail identity comparison with ==. Co-Authored-By: Claude <[email protected]>

e2e_playwright/st_multiselect.py

…header-height

lib/streamlit/elements/lib/options_selector_utils.py

kmcgrady

Code changes here are fine. Do we need to consider this for other options-based widgets? If not, can we just add to the PR description, why it's unnecessary?

st.radio
st.pills
st.segmented_control
st.select_slider
Maybe some column config like SelectboxColumn?

lukasmasuch · 2026-01-22T18:31:40Z

Code changes here are fine. Do we need to consider this for other options-based widgets?

st.radio -> already addressed in a recent PR
st.pills, st.segmented_control, st.select_slider are not impacted since they are still index-based for widget state, but are planned to be changed to support dynamic options support + potentially also required for binding query params.
Maybe some column config like SelectboxColumn? -> column types are not impacted since they work differently regarding the widget state

…13648) ## Describe your changes Fixes issue #13646 where `st.multiselect()` and `st.selectbox` clears user selections after each script rerun when using custom class objects without `__eq__` implementation and a `format_func` parameter. This was a regression introduced in PR #13448. **Root Cause**: The validation logic compared multiselect values using `==`, which falls back to identity comparison for objects without `__eq__`. Since `register_widget()` deepcopies widget values, the new instances fail identity comparison with original options, causing valid selections to be filtered out. **Solution**: Changed validation to compare values using `format_func()` by their formatted string representation instead of using `==`. This is more robust and works correctly regardless of whether custom classes implement `__eq__`. ## Testing Plan - Unit Tests: Added 4 comprehensive test cases covering custom objects without `__eq__`, partial matches, objects with `__str__`, and edge cases where `format_func` fails on incompatible types - E2E Tests: Added `test_multiselect_custom_objects_without_eq` to verify selections persist across script reruns with custom class objects - Backward Compatibility: The `format_func` parameter has a default value of `str`, so existing code continues to work without changes --- **Contribution License Agreement** By submitting this pull request you agree that all contributions to this project are made under the Apache 2.0 license. --------- Co-authored-by: Claude <[email protected]>

Copilot AI review requested due to automatic review settings January 20, 2026 15:13

Copilot started reviewing on behalf of lukasmasuch January 20, 2026 15:14 View session

lukasmasuch changed the title ~~Fix multiselect clearing selections for custom objects without __eq__~~ Fix multiselect clearing selections for custom objects Jan 20, 2026

lukasmasuch added security-assessment-completed change:bugfix PR contains bug fix implementation impact:users PR changes affect end users labels Jan 20, 2026

Copilot AI reviewed Jan 20, 2026

View reviewed changes

cursor bot reviewed Jan 20, 2026

View reviewed changes

lukasmasuch added the ai-review If applied to PR or issue will run AI review workflow label Jan 20, 2026

github-actions bot removed the ai-review If applied to PR or issue will run AI review workflow label Jan 20, 2026

graphite-app bot reviewed Jan 20, 2026

View reviewed changes

lib/streamlit/elements/widgets/multiselect.py Outdated Show resolved Hide resolved

lukasmasuch and others added 2 commits January 20, 2026 20:19

graphite-app bot reviewed Jan 20, 2026

View reviewed changes

e2e_playwright/st_multiselect.py Show resolved Hide resolved

sfc-gh-lmasuch mentioned this pull request Jan 21, 2026

Allow dynamic changes to st.radio options when key is provided #13611

Merged

Merge remote-tracking branch 'upstream/develop' into lukasmasuch/fix-…

12d0e2a

…header-height

lukasmasuch changed the title ~~Fix multiselect clearing selections for custom objects~~ Fix selectbox and multiselect clearing selections for custom objects Jan 21, 2026

lukasmasuch added 2 commits January 21, 2026 23:17

Update

b53351d

Update

58f9f24

graphite-app bot reviewed Jan 21, 2026

View reviewed changes

lib/streamlit/elements/lib/options_selector_utils.py Show resolved Hide resolved

kmcgrady approved these changes Jan 22, 2026

View reviewed changes

lukasmasuch merged commit 8150720 into develop Jan 22, 2026
44 checks passed

lukasmasuch deleted the lukasmasuch/fix-header-height branch January 22, 2026 18:32

github-actions bot mentioned this pull request Jan 24, 2026

st.selectbox fails to update/change value when dynamicly adding new items #13686

Open

4 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix selectbox and multiselect clearing selections for custom objects#13648

Fix selectbox and multiselect clearing selections for custom objects#13648
lukasmasuch merged 7 commits intodevelopfrom
lukasmasuch/fix-header-height

lukasmasuch commented Jan 20, 2026 •

edited

Loading

Uh oh!

snyk-io bot commented Jan 20, 2026 •

edited

Loading

Uh oh!

github-actions bot commented Jan 20, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

github-actions bot commented Jan 20, 2026 •

edited

Loading

Uh oh!

cursor bot left a comment

Uh oh!

cursor bot Jan 20, 2026

Uh oh!

github-actions bot commented Jan 20, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

kmcgrady left a comment

Uh oh!

lukasmasuch commented Jan 22, 2026 •

edited

Loading

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

lukasmasuch commented Jan 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Describe your changes

Testing Plan

Uh oh!

snyk-io bot commented Jan 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅ Snyk checks have passed. No issues have been found so far.

Uh oh!

github-actions bot commented Jan 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅ PR preview is ready!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

github-actions bot commented Jan 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

📉 Frontend coverage change detected

Uh oh!

cursor bot left a comment

Choose a reason for hiding this comment

Uh oh!

cursor bot Jan 20, 2026

Choose a reason for hiding this comment

Validation and serialization use inconsistent comparison methods

Uh oh!

github-actions bot commented Jan 20, 2026

Summary

Code Quality

lib/streamlit/elements/lib/options_selector_utils.py

lib/streamlit/elements/widgets/multiselect.py

Potential Edge Case (not a blocking issue)

Test Coverage

Unit Tests (lib/tests/streamlit/elements/lib/options_selector_utils_test.py)

E2E Tests (e2e_playwright/st_multiselect_test.py)

Backwards Compatibility

Security & Risk

Recommendations

Verdict

Uh oh!

Uh oh!

Uh oh!

Uh oh!

kmcgrady left a comment

Choose a reason for hiding this comment

Uh oh!

lukasmasuch commented Jan 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

lukasmasuch commented Jan 20, 2026 •

edited

Loading

snyk-io bot commented Jan 20, 2026 •

edited

Loading

github-actions bot commented Jan 20, 2026 •

edited

Loading

github-actions bot commented Jan 20, 2026 •

edited

Loading

`lib/streamlit/elements/lib/options_selector_utils.py`

`lib/streamlit/elements/widgets/multiselect.py`

Unit Tests (`lib/tests/streamlit/elements/lib/options_selector_utils_test.py`)

E2E Tests (`e2e_playwright/st_multiselect_test.py`)

lukasmasuch commented Jan 22, 2026 •

edited

Loading