Implementation of altair backend.

[r3kste]

Summary of Changes

1. Added the line_plot and grid_line_plot function for altair.
2. Added a small example in the documentation to show the altair plot.

Plot Comparision

Plotly	Altair

Checklist

• Documentation
• Tests
• Comparision of plot outputs

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
src/optimagic/config.py	100.00% <100.00%> (ø)
src/optimagic/visualization/backends.py	99.59% <100.00%> (+0.17%)	⬆️
src/optimagic/visualization/convergence_plot.py	95.65% <ø> (ø)
src/optimagic/visualization/history_plots.py	95.78% <ø> (ø)
src/optimagic/visualization/profile_plot.py	91.52% <ø> (ø)
src/optimagic/visualization/slice_plot.py	96.12% <ø> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[r3kste]

The failing tests are still due to

ERROR: Could not install packages due to an OSError: [Errno 28] No space left on device

Looks like bayes_optim library requires torch>=2.0.1. But on linux, CI installs the cuda version of torch which also attempts to install large nvidia/cuda libraries.

In a fresh environment without bayes_optim, installing the cpu version of torch before installing bayes_optim skips the installation of the cuda libraries. Indeed, all Windows and Macos tests are passing as they install the cpu version of torch. So, I believe this could be the ideal solution. However, I wasn't able to specify/control the order in which the libraries are installed by pip. Moreover, this doesn't explain why these failures have started appearing recently (even in other PRs).

[timmens]

Great investigative work! Have you tried adding the conda-forge version pytorch-cpu to the conda environment? The conda enviromment should be solved and installed before the pip environment. However, it could be that pip simply ignores this and downloads the torch dependency again.

[r3kste]

Great! Adding pytorch-cpu as a conda dependency seems to have worked. Would you prefer it to be done in a separate PR instead of this one?

The two tests failing currently are unrelated but they seem to persist even after multiple re-runs. One of them is due to comparison of positive zero with negative zero similar to #642.

File "how_to/how_to_constraints.md", line 327, in how_to_constraints.md
Failed example:
    res.params.round(3)
Expected:
    array([ 1. ,  0.8,  0.6,  0.4,  0.2, -0. ])
Got:
    array([1. , 0.8, 0.6, 0.4, 0.2, 0. ])

The other one is due to the actual not being close enough to the expected. I'm not entirely sure what causes this.

ACTUAL: array([0.188, 0.006, 0.011])
DESIRED: array([0.19 , 0.006, 0.011])

[timmens]

Yes, please open another PR to fix this. We will see whether the two mentioned issues persist there. Thanks!

[r3kste]

@timmens Sure. I have opened #656.