[Bimpebabs]

Closes #48.

Modernizes docrepr's vendored front-end assets, which dated from 2015, addressing each point raised in the issue. Split into four focused commits so each can be reviewed independently.

What changed

• HTML5 doctype — replaced the XHTML 1.0 Transitional doctype in rich_repr.html with <!DOCTYPE html>, added lang="en" and the HTML5 <meta charset> form.
• MathJax — the template loaded MathJax 2 from https://cdn.mathjax.org, which was retired in 2017. It emitted a console warning and never loaded; and because current Sphinx emits v3-style \(...\) delimiters, the math would not have rendered even if the CDN were alive. Now loads MathJax 3 from the officially recommended jsDelivr CDN (tex-mml-chtml component) with an inline v3 config. math_config.js is reduced to the jQuery-only monospace fallback used when math rendering is off (the old MathJax.Hub SVG config was dead code, gated on an undefined right_sphinx_version template variable).
• Pygments — regenerated pygments.css from the same sphinx style with the current Pygments, keeping the existing bare-class scheme. The old file was missing token classes newer Pygments emits (.fm, .vm, .ch, .cpf, .sa, .dl, .mb, .ges), which left some tokens unstyled.
• Bootstrap + jQuery + copy button — jQuery 1.11.3 -> 3.7.1 and Bootstrap 3.3.5 -> 3.4.1 (CSS + JS). Kept the jQuery-based track because all of docrepr's scripts depend on jQuery; Bootstrap 3.4.1 is API-compatible with the tab markup used here and is a security/maintenance release over 3.3.5. copy_button.js was rewritten for modern jQuery: it relied on the two-argument \$.fn.toggle(fn, fn) helper removed in jQuery 1.9 (so it did nothing), and its selector only matched .highlight-python/.highlight-python3 wrappers, so with current Sphinx the button no longer appeared at all. It now uses a click handler that tracks its own state and matches any div.highlight block containing prompts.

Testing

• cd docrepr && python -bb -X dev -W error -m pytest — 16 passed (the CI-faithful command from CONTRIBUTING.md, warnings-as-errors).
• Rendered the representative output cases in a browser and confirmed, at runtime: the dead-CDN console warning is gone; MathJax 3 loads and typesets the math; the copy button now appears and toggles the prompts/output; and the Bootstrap tabs still switch. No console errors or failed network requests on any page.
• python -m build + twine check --strict pass, and the built wheel contains all the updated assets; pip check reports no broken requirements.

Visual regression / reference screenshots

This changes the rendered output, so the reference screenshots for the visual regression tests need regenerating. I'll leave a Please update reference screenshots comment so the CI regenerates them per-platform.

Notes

• Vendored files are the unminified distributions, matching the previous convention.
• Bootstrap 3.4.1 also resolves the XSS advisories that affected 3.3.5, which may interest the security-conscious deployments mentioned in the issue. If the maintainers want this on the 0.2.x release branch as well, happy to backport.

[ccordoba12]

Hey @Bimpebabs, thanks for your contribution! Don't worry about the failing tests, that's because our CI integration is quite outdated.

We'll try to fix it as soon as possible.

[Bimpebabs]

Thanks @ccordoba12, much appreciated!

Since the CI came up: I took a look and put together a small, tested branch that gets the suite green again on current Python and dependencies. Happy to open it as a separate PR if that'd be useful to you (totally fine if you'd rather handle it yourselves). In short it:

• switches setup.cfg to the SPDX license field and declares packages/package_data explicitly, so python -bb -X dev -W error -m build passes with current setuptools;
• sets asyncio_mode/asyncio_default_fixture_loop_scope in pytest.ini for modern pytest-asyncio;
• makes the IPython Inspector/object_info calls in the tests version-compatible (IPython 9 needs theme_name, and name/found);
• bumps the workflow action versions and the Python matrix to 3.9-3.13.

I verified build, twine check --strict, pip check, and python -I -bb -X dev -W error -m pytest (16 passed) all pass on Python 3.12 with current IPython/Sphinx/pytest, no dependency pinning.

One heads-up related to this PR specifically: conftest.py's visual-regression fixture waits for #MathJax_Message, which is a MathJax 2 element. Since this PR moves to MathJax 3 (no such element), that wait will need updating so the visual-regression job doesn't hang -- I'm happy to include that fix here.

[CAM-Gerlach]

First off, just to note for the record the tests are failing pending resolution of #50 , which should be a straightforward upgrade (the visual regression tests will likely need additional work, but that can be deferred for now).

Additionally, it uses a still-ancient, long-EOL Bootstrap release from 2019, so it isn't really an improvement there.

However, unfortunately there's a much bigger issue with this PR, especially given it was explicitly intended to be for a good first issue for contributor new to open source such as yourself: Not only the code but the PR description and even the replies to (human) reviewers appears to be 100% AI generated (and likely agent-posted), without any disclosure.

This directly contravenes the AI usage policy that our project along with the others involved in the sustainability series program have adopted, which you were explicitly reminded of in your orientation emails, in both office hours, and in our Discord. Specifically, using LLMs to assist in coding is allowed but must be disclosed, which it was not; that contributors must understand the changes and describe them in their own words in the PR description, which they did not; and that under no circumstances should LLMs be used to generate text for nor replace genuine human to human communication when replying to a reviewer, which it was).

Beyond the policy, having an LLM do the entire contribution process for you unfortunately kinda defeats the purpose of a "good first issue" to begin with in terms of learning how to contribute and communicate in an open source project. To be sure, LLMs can be useful tools in this regard toward gaining understanding of a new codebase, suggesting approaches to solve an issue and reviewing your work. However, the problem arises when someone just points the LLM at an issue and tells it to make a PR, even having it read and respond to the human reviewers for them. The whole purpose of the endeavor is for you to learn and gain experience with the process, which using a LLM agent to do the whole thing denies you.

Finally, just to note as of last week on our Discord we've pinged and reminded everyone in this program to please only claim one issue at a time until your PR is merged, so there are enough to go around for everybody--I see you also opened spyder-ide/spyder-docs#441 at the same time.

[CAM-Gerlach]

For the reasons explained above, I unfortunately will have to close this PR, as to actually accomplish the purpose of a "good first issue" you'll need to actually go through and experience the contributing process yourself. Let us know (here or on Discord) if you have any questions, and we'll be happy to assist.

Just to note--to avoid any possible confusion going forward, I've merged PR #60 which adds an explicit statement of the AI policy in this repo's Contributing Guide as well as an AGENTS.md with explicit instructions to LLMs of what to do and not to do (although it should not be relied upon as an end-all, be all, and you are ultimately responsible for their actions), in addition to the previous mentions in the Spyder main repo CONTRIBUTING.md, our governance and guidelines and elsewhere.