[Enesuygurs]

Link to issue

Fixes #5846

Description of change

plotly.colors.validate_colors and plotly.colors.convert_colors_to_same_type rewrite the list they are given in place.

• validate_colors: when called with None, it uses DEFAULT_PLOTLY_COLORS itself, so the module-level default colors become float tuples for the rest of the session. create_trisurf does this whenever colormap is left as None. When a user passes a colormap list instead, that list is converted to tuples.
• convert_colors_to_same_type: it aliases a list argument the same way, so passing a built-in sequence rewrites it:

import plotly.colors as pc

pc.convert_colors_to_same_type(pc.sequential.Inferno, colortype="rgb")
pc.sequential.Inferno[:2]      # ['rgb(0, 0, 4)', 'rgb(27, 12, 65)'] instead of ['#000004', '#1b0c41']

pc.validate_colors(None)
pc.DEFAULT_PLOTLY_COLORS[:1]   # [(0.1215..., 0.4666..., 0.7058...)] instead of ['rgb(31, 119, 180)']

Both functions now copy the list before converting it. The return values are unchanged.

Testing strategy

I added test_color_conversions_do_not_modify_input. It checks that DEFAULT_PLOTLY_COLORS and a user list are unchanged after calling both helpers. It fails on main and passes with this change.

Locally, tests/test_core/test_colors, tests/test_plotly_utils and tests/test_optional/test_figure_factory pass. The exception is the create_ternary_contour tests: they need scikit-image, which isn't installed here, and they fail the same way on main. ruff format --check . passes.

Additional information (optional)

This also fixes cross-test pollution in the existing suite. test_sample_colorscale calls convert_colors_to_same_type(colors.sequential.Inferno, ...), which rewrote Inferno. As a result, running tests/test_core/test_colors before tests/test_plotly_utils made test_colorscale_validator.py::test_acceptance_named[Inferno_r] fail on main. It passes with this change.

Guidelines

• I have reviewed the pull request guidelines and the Code of Conduct and confirm that this PR follows them.
• I have added an entry to the changelog if needed (not required for documentation PRs).

[KoolADE85]

Thanks for the PR! Can you please create an issue that describes the problem and how you encountered this issue in your day-to-day usage of plotly.py?

[Enesuygurs]

Thanks for taking a look! I opened #5846 with a description and a short reproduction, and linked it from this PR. To be transparent: I came across this while reading the color utilities rather than hitting it in a project. The most visible effect is that a built-in sequence such as plotly.colors.sequential.Inferno (or DEFAULT_PLOTLY_COLORS) stays modified for the rest of the session after one of these helpers is called on it. It also causes an order-dependent failure in the existing test suite.

[camdecoster]

Thanks for the fix! Could you do a pass to see if there are any other similar mutations that need to be addressed? It looks like convert_dict_colors_to_same_type would need the same update (and maybe others).

[camdecoster]

Suggested change

- Fix `plotly.colors.validate_colors` and `plotly.colors.convert_colors_to_same_type` modifying the list of colors passed to them in place, which also changed `plotly.colors.DEFAULT_PLOTLY_COLORS` (for example after calling `create_trisurf` without a `colormap`) and built-in color sequences such as `plotly.colors.sequential.Inferno` [[#5836](https://github.com/plotly/plotly.py/pull/5836)], with thanks to @Enesuygurs for the contribution!

- Fix color utilities mutating the list of colors passed to them [[#5836](https://github.com/plotly/plotly.py/pull/5836)], with thanks to @Enesuygurs for the contribution!

[Enesuygurs]

Thanks for the review! I went through every function in plotly.colors and found three more that changed their input in place: convert_dict_colors_to_same_type, validate_colors_dict and convert_colorscale_to_rgb. They now work on a copy as well. I added cases for them to test_color_conversions_do_not_modify_input, updated the changelog entry, and merged the latest main to resolve the changelog conflict.

[camdecoster]

Thanks for getting this fixed!