[T4rk1n]

Drops Python 3.9. Dash now requires Python 3.10, and CI tests on 3.10 and 3.14 (was 3.9 and 3.12).

Changes

• setup.py: python_requires=">=3.10", classifiers 3.10 through 3.14.
• testing.yml: every ["3.9", "3.12"] matrix is now ["3.10", "3.14"]. Percy moves with the upper version (3.14). Single-version jobs (build, typing, table, benchmarks) stay on 3.12.
• Lint (main and DCC) runs on 3.10 only: the pinned black 22.3.0 and pylint 3.0.3 crash on 3.14. A linter bump is now unblocked but needs a repo-wide reformat, so it is left for its own PR.
• Removed 3.9 workarounds: pylint --ignored-modules=mcp, the mimesis version split, MCP test skips, the mypy test gate and its python_version>="3.12" install marker, a dead 3.8 fallback in the component generator.

Fix found along the way

On Python 3.10, isinstance(list[str], type) is True, so issubclass(list[str], Component) raised TypeError for MCP tools with annotations like list[str]. CI never caught it because MCP tests were skipped on 3.9 and only ran on 3.12. test_mcpd009 and test_mcpm003 fail on 3.10 before the fix and pass after.

Notes

• jupyterlab<4 pulls y-py, which has no 3.14 wheel. pip builds it from source with Rust (present on GitHub runners), so the 3.14 installs are a bit slower.
• Verified locally: unit tests on 3.10 and 3.14.10, MCP and callback integration tests with dash_duo on both, compliance tests and lint on 3.10.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[github-actions]

Dash performance benchmarks

✅ all within thresholds

	scenario	metric	p90 (ms)	median	growth	baseline p90	note
✅	callback_chain	chain_ms	442.7	425.1	0.97x	499.1
✅	callback_chain	graph_ms	2.5	2.5	1.0x	2.4
✅	callback_fanout	fanout_ms	83.1	78.2	0.86x	92.5
✅	deep_nesting	render_ms	71.7	63.9	1.05x	56.8
✅	full_children_replace	replace_ms	5343.3	1992.1	18.32x	4697.1
✅	initial_render_large	render_ms	663.3	575.5	0.95x	694.4
✅	initial_render_small	render_ms	101.3	93.9	0.97x	104.0
✅	patch_append_nested	append_ms	129.8	92.3	2.54x	192.3
✅	patch_append_toplevel	append_ms	129.6	86.6	2.41x	140.2
✅	patch_scalar_update_large	update_ms	154.7	140.2	1.03x	202.9
✅	wildcard_all_resolve	wildcard_ms	311.0	293.6	0.93x	313.6
✅	wildcard_all_resolve	graph_ms	1.1	1.1	1.0x	1.3

growth = late-third / early-third per-op time; ~1 is flat, a large value means the per-op cost scales with accumulated state.

machine scale vs baseline: 0.98x - divided out of the baseline ratios so they compare like for like (the absolute warn/fail ceilings are left un-scaled); calibrated on initial_render_small.

[KoolADE85]

This also closes #4038