This works really nicely! A few things to add to this:

• I think the snap angle should be an rcParam rather than hardcoded to 5 degrees
• Please add a what's new entry for this
• We should document this behavior somewhere, I think a brief note here is probably the best place for right now: https://matplotlib.org/devdocs/api/toolkits/mplot3d.html

@scottshambaugh
I’m facing issues while passing the new parameter in rcParams. Could you please guide me on how to correctly register it?
Currently, I have:
Added "axes3d.snap_rotation": validate_float in the validators dictionary.
Added the following entry in the params list:
_Param(
"axes3d.snap_rotation",
default=0.0,
validator=validate_float,
description="Snap angle (in degrees) for 3D rotation when holding Control."
)

I have also made the corresponding changes in other required files axes3d.py and test file too.
step = mpl.rcParams["axes3d.snap_rotation"]
if (
step > 0
and getattr(event, "key", None)
and "control" in event.key
):
elev = step * round(elev / step)
azim = step * round(azim / step)
roll = step * round(roll / step)
in axes3d.py and monkeypatch.setitem(plt.rcParams, "axes3d.snap_rotation", 5.0) in test file.
However, I am getting an error saying that "axes3d.snap_rotation" is not a valid rcParams key.
Could you please let me know if I am missing any required step?

Could you push what you have right now so I can see the code changes?

@scottshambaugh
done now you can check that

@scottshambaugh
Thanks for your review and suggestion regarding required changes.
Please let me know if any additional changes are required.
I have removed the _snap_rotation flag and the related key press/release handlers, since snapping is now controlled through the axes3d.snap_rotation rcParam instead of using the previously hardcoded 5-degree value.

Hi @Chirag3841, I don't see those changes pushed up yet.

Hi @Chirag3841, I don't see those changes pushed up yet.

I have done that in the last commit which caused multiple errors where 2 nd pic is of test file and remaining of axes3d.py

FAILED lib/mpl_toolkits/mplot3d/tests/test_axes3d.py::test_ctrl_rotation_snaps_to_5deg - KeyError: 'elev'
Getting this error too

@scottshambaugh I resolved that issues in recent PR but still getting valid rc parameter error on local
And 12.3==10 error in elev

@scottshambaugh I changed the test file code and added what's a new doc too.

@scottshambaugh Thanks for your review and support.
Let me know if any changes are required or now .

Any Update ?

We require 2 maintainers to approve each PR that changes code, so it may take a bit for someone else to look at this.

We require 2 maintainers to approve each PR that changes code, so it may take a bit for someone else to look at this.

Thanks for the clarification

Should we support a tuple so that we can set individual snap angles for (elev / azim / roll)?

Also, would a special value make sense to lock that angle? That way, one could e.g. limit the rotation to azim and keep elev and roll fixed. But not sure whether this should be folded into snap.

@timhoffm That’s an interesting idea but per-angle snapping or locking could potentially be explored separately if we think it warrants expanding the API.

Well, it potentially touches a design decision: If we want per-angle snapping, we'd need to support three-component config

axes3d.snap_rotation: 5.0, 5.0, 5.0

Then we can decide whether we additional want to allow single-component and implicitly broadcast. This is a trade-off: convenience vs. a more complex API and handling logic. It's not worlds difference but something that we only have the freedom to decide before this is released.

I'm slightly on side of always requiring the three values. It's that cumbersome to write them out if one really cares to configure snap.

@timhoffm I feel that keeping it as a single float keeps the API simpler and likely covers the most common use case.
However, I understand the design considerations around long-term flexibility. If we think per-angle snapping is important from the start, I’m happy to update the PR to support a 3-component configuration.

My ask for making this angle an rcparam was more for wanting to avoid magic constants than anything else. I could see people wanting to make it e.g. 1 deg rather than 5, but think the per-axis snapping values is going to be used rarely if ever. IMO keeping this simple with a single value is best.

Ok, if the expectation is that individal angle steps are not needed, let's merge as is.

Thanks @scottshambaugh and @timhoffm for reviewing and supporting this PR.