3d Visualizations of optimization problems

[shammeer-s]

Sandbox Version: These changes are added as a sandbox version of slice_plot. It can be imported using om.sandbox.slice_plot_3d

Summary of changes

1. slice_plot_3d now supports surface and contour projections.
2. Introduction of layout_kwargs and make_subplot_kwargs for easier layout and subplot control. Users can adjust figure size, spacing, shared axes, and grid structure effortlessly.
3. plot_kwargs allows users to customize visual aspects (line color, scatter size, surface colorscale, contour smoothness) for each plot type individually.
4. Added a new notebook, how_to_slice_plot_3d.ipynb, with detailed examples and use cases to assist with adoption.

Future Improvements

1. Improved plot headings, refined 3D surface scene settings, customized axis ticks/labels, and subplot titles for clearer grids.
2. Reducing redundant grid evaluations to improve computational efficiency.

Note: An extensive parameter usage is provided in the attached notebook (Inside the .zip), advanced_slice_plot_3d_usage.zip. Example outputs are also included below in this comment.

[timmens]

Hey,

Thanks for the PR!

Before I start a thorough review of your changes, lets tackle some high-level issues:

1. Can you move the code from plot_data.py into the slice_plot.py module, unless it is also used in other plotting functions.
2. Can you re-order the functions such that the function that is exported by the module (slice_plot) is at the beginning, then the private slice plot functions, and lastly the plotting data code.
3. Can you create a small notebook where you showcase the usage of the new features (this does not need to pushed, you can just upload it in a comment here).

Thanks a lot! If you have any questions, let me know here or on Zulip!

[shammeer-s]

Hey @timmens,

Thanks for the feedback!

Sure, I’ll move the code from plot_data.py into slice_plot.py as suggested, since it’s not being used elsewhere. I’ll also re-order the functions so that the exported slice_plot function comes first.

I'll work on a small notebook to demonstrate the new features and share it here once it’s ready.

Also, I’ll take this opportunity to improve the modularity of the plotting logic a bit—cleaning it up should make the flow easier to understand.

Let me know if there's anything else I should add or keep in mind!

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codecov]

Codecov Report

Attention: Patch coverage is 17.22973% with 245 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/optimagic/visualization/slice_plot_3d.py	16.38%	245 Missing ⚠️

Files with missing lines	Coverage Δ
src/optimagic/__init__.py	93.33% <100.00%> (ø)
src/optimagic/sandbox.py	100.00% <100.00%> (ø)
src/optimagic/visualization/slice_plot_3d.py	16.38% <16.38%> (ø)

🚀 New features to boost your workflow:

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

[timmens]

Thanks so much—this is looking great already! A few things that need to be addressed before we continue:

General

• Type annotations: Mypy is complaining because there are missing annotations. Could you add argument and return type hints wherever the types are clear? For the exported function interfaces, feel free to silence the error (if you are unsure) with # type: ignore[no-untyped-def].
• Pre-commit hooks: Ruff is flagging some issues. Please fix those and make sure all pre-commit hooks run and pass before pushing.
• Tests: Add test_slice_plot_3d.py to cover slice_plot_3d.py (use test_slice_plot.py as a template). Ideally include:
  • End-to-end tests calling slice_plot_3d() with various inputs.
  • Unit tests for each helper function.
• Code: To gain a better understanding how your code differs from our current slice_plot implementation, could you please align your code with that of slice_plot. Currently you are doing two things at the same time: (1) you are adding the 3D feature, and (2) you are refactoring the slice_plot code. It is hard to focus on (1) when you are also doing (2). For example, the function evaluate_func does not exist in the original implementation, and it is hard for me to see whether you added functionality there.

Visualization comments

These comments arose after looking at the how-to notebook

• 1D Slice Plot
  • Remove the plot title "Slice Plot".
  • Remove the legend entry "trace 1".
• 3D Surface Plot (Two parameters)
  • Remove the plot title "Slice Plot".
  • Ensure both axes are fully visible (currently cropped, especially in the lower center).
  • Rename the axes from "x"/"y" to "alpha"/"beta".
• 2D Contour Plot (Two parameters)
  • Remove the plot title "Slice Plot".
  • Label the axes "alpha" and "beta".
  • Add a color legend so it’s clear which color matches each contour level.
• Grid View (Multiple parameters)
  • Use a selector to focus on the first three parameters.
  • Once the individual plots are fixed, also ensure:
  • No plot title "Slice Plot".
  • No legend "trace 1".
  • All axes visible in the 3D plots (no cropping).
  • Display parameter names outside the grid—once along the left edge and once along the bottom, similar to sns.pairplot.
• Behavior change for projection:
  • If projection is a string (or Enum), e.g., "surface":
    • Diagonal: univariate slice plots
    • Lower triangle: 2D/3D plots based on that string
    • Upper triangle: empty
  • If projection is a dictionary {"lower": "surface", "upper": "contour"}:
    • Diagonal: univariate slice plots
    • Lower triangle: 3D surface plots
    • Upper triangle: 2D contour plots
  • I’d recommend parsing projection early into {"lower":…, "upper":…}, where None means no plot (e.g. "surface" → {"lower":"surface","upper":None}). Given your implementation the dictionary values can be parsed to the Projection Enum at this point as well.

Thanks a lot! If you have any questions let me know!

[timmens]

Suggested change

	slice_plot_3d = slice_plot_3d
	__all__ = ["slice_plot_3d"]

[timmens]

Please perform explicit checks for is None.

[timmens]

I'd prefer UNIVARIATE.

[timmens]

Suggested change

	class Projection(Enum):
	SLICE = auto()
	class Projection(str, Enum):
	SLICE = "slice"

Same for the other attributes.

[timmens]

Suggested change

	def is_single(self):
	def is_univariate(self):

[timmens]

Suggested change

	def is_multiple(self):
	def is_bivariate(self):

[timmens]

Please use regular markdown bullet points, instead of unicode bullet points.