[luth2]

SUMMARY

Replaces #45821 with a reviewable, line-ending-clean diff. Related to #45019.

The previous PR included repository-wide CRLF/LF churn. This branch is rebuilt from the same original base and preserves its substantive changes without that churn: 49 files changed, 680 insertions, 135 deletions, in one commit. The original branch has not been rewritten.

• Add experimental Python 3.15 source compatibility while keeping Python 3.11 as the minimum supported version and leaving the default container interpreter and support classifiers unchanged.
• Select NumPy >=2.5.4,<2.6 and pandas >=3.1.0,<3.2 on Python 3.15 using dependency markers; preserve the existing ranges for Python 3.11-3.14.
• Update pandas post-processing, result-set handling, SQL parsing and related compatibility paths, with regression coverage.
• Add a Python 3.11-3.15 compatibility workflow covering focused tests, dependency consistency, a fresh SQLite migration and application health.
• Use package metadata for the lean/superset Docker dependency stage on Python 3.15, retaining the requirements-based path for older interpreters. Build read-only mounted sources in a disposable writable copy and test cleanup on both success and failure.
• Declare rich directly for CLI imports and document installation, validation and known limitations.

The frontend Node pin remains v24.16.0 with LF line endings, matching the original base. This avoids carrying the CRLF-corrupted version string that blocked the previous Netlify docs preview. No Netlify configuration or Node version change is needed.

Beyond removing line-ending churn, this replacement adds an explicit union type annotation in the container regression test and the shell script's final newline, as required by local checks.

BEFORE/AFTER SCREENSHOTS OR ANIMATED GIF

Not applicable. No intentional UI changes.

TESTING INSTRUCTIONS

Create an isolated Python 3.15 environment and install from package metadata, not the generated Python 3.11 requirement files:

python -m venv .venv
source .venv/bin/activate
python -m pip install -e ./superset-core -e .
python -m pip check
export FLASK_APP=superset
export SUPERSET_HOME="$(mktemp -d)"
export SUPERSET_SECRET_KEY="$(python -c 'import secrets; print(secrets.token_hex(32))')"
python -m flask db upgrade
python -c 'from superset.app import create_app; assert create_app().test_client().get("/health").status_code == 200'

Run the new compatibility workflow and the container regression tests:

PYTEST_DISABLE_PLUGIN_AUTOLOAD=1 python -m pytest --noconftest -o addopts= \
  tests/unit_tests/docker_app_dependencies_test.py \
  tests/unit_tests/scripts/docker_release_target_test.py \
  tests/unit_tests/docker_compose_image_tag_test.py -q
bash -n docker/install-app-dependencies.sh

Validation of this replacement commit:

• All 18 container-related regression tests passed on Python 3.15.2; shell syntax and whitespace checks passed.
• Pylint passed for the changed backend files, using the repository configuration and a single worker.
• Pre-commit was run on all staged changes. All applicable hooks passed except MyPy, which reports an unused type: ignore at superset/db_engine_specs/mysql.py:680. The same finding was reproduced on the unchanged original base with the same selection of modified existing files. The new test's MyPy error was corrected before this commit.
• The complete published diff was checked against the previous branch: substantive content is preserved, apart from the two check corrections described above.

Remaining limits: this is experimental compatibility, not production certification. The complete official image build remains unverified because the reference host ran out of memory during frontend minimization, although the isolated Python 3.15 dependency stage passed. The documented Gunicorn gthread shutdown limitation remains unresolved. The new Netlify preview and PR CI still need to run; /health is liveness, not dependency readiness.

ADDITIONAL INFORMATION

• Has associated issue: [SIP-234] Add Support for Python 3.14 #45019; replaces feat(python): add experimental Python 3.15 compatibility #45821.
• Required feature flags:
• Changes UI
• Includes DB Migration
• Introduces new feature or API
• Removes existing feature or API

[netlify]

✅ Deploy Preview for superset-docs-preview ready!

Name	Link
🔨 Latest commit	6a830c5
🔍 Latest deploy log	https://app.netlify.com/projects/superset-docs-preview/deploys/6ae5f5bbcd81fbf347afd017
😎 Deploy Preview	https://deploy-preview-45823--superset-docs-preview.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.
🤖 Make changes	Run an agent on this branch

---

To edit notification comments on pull requests, go to your Netlify project configuration.

[codeant-ai-for-open-source]

  CodeAnt PR Risk

🟢  Low Risk

• The PR appears safe to merge; the reviewed datetime-import, pivot-shaping, and boxplot-conversion concerns are fixed in the current code.

Assessed commit: 6a830c53dab6

[luth2]

@sadpandajoe All five latest review findings are fixed in 6b78fdb and their threads have been answered and resolved. I also merged the latest master (1d0fc23) in 7acde2c; GitHub now confirms the branch is up to date and conflict-free.

Post-merge local validation: 630 focused tests passed on Python 3.15/pandas 3.1.0, including temporal imports, joins, date normalization, columnar readers, pandas post-processing, result sets and the static OpenAPI tests. Scoped pre-commit hooks (including MyPy/Ruff and workflow security checks) and the POT drift check pass. Before this master merge, the four modules affected by the latest fixes also passed all 346 tests on each of Python 3.11/pandas 3.0.6 and Python 3.15/pandas 3.1.0.

The remaining merge gates are the required code-owner approval and 24 workflows awaiting maintainer approval. Could a maintainer please approve the CI runs and re-review the updated PR? These local results are not a substitute for the pending hosted CI.

[sadpandajoe]

Wrapping except ValueError around the vectorized pd.to_datetime sends every parse error into the per-row fallback, not just out-of-bounds ones. pandas now infers one format from the first value and raises when a later value doesn't match, so a column like ["01/02/2020", "05/03/2020", "13/02/2020"] used to fail the import with a format mismatch, but now imports as 2020-01-02, 2020-05-03, 2020-02-13: the first two month-first and the third day-first, with no warning since the log only counts out-of-bounds NaTs. Could the fallback be limited to OutOfBoundsDatetime and the "Mixed timezones" case, re-raising other ValueErrors?

[sadpandajoe]

This sort_index(axis=1) sorts every column level, whereas the old stack(level=0).unstack() kept the order pivot_table produced. With marginal_distributions=True and combine_value_with_metric=True, a pivot_table result of (m1,AU) (m1,US) (m1,All) (m2,AU) ... used to come back as (AU,m1) (AU,m2) (US,m1) (US,m2) (All,m1) (All,m2) and now comes back as (AU,m1) (AU,m2) (All,m1) (All,m2) (US,m1) (US,m2), so the totals column lands in the middle instead of last. test_pivot_combine_preserves_column_pairs sorts both frames before comparing, so it can't see this. Can the reorder avoid sorting the non-metric levels?