Allow to zoom secondary ranges/scales independently

[mattpap]

This is functionally ready. This PR allows for wheel zoom of individual axes (zoom_on_axis="individual") and on linked (or perhaps a better term is cross) axes (zoom_on_axis="individual_linked"). The original behavior of zoom_on_axis with a boolean value is left unchanged. Additionally ZoomInTool and ZoomOutTool can be configured with renderers, which are used to determine which ranges will be affected by zoom:

This example shows WheelZoomTool(zoom_on_axis="individual"), and ZoomInTool and ZoomOutTool configured with a single renderer:

Screencast_00001.mp4

TODO:

• fix unit tests
• add tests for the new behavior

fixes #12829

[bryevdv]

What are you envisioning for the API / properties (independent vs synced will need to be configurable somehow)?

[codecov]

Codecov Report

Merging #13049 (e09eb3e) into branch-3.2 (8f38eeb) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@             Coverage Diff             @@
##           branch-3.2   #13049   +/-   ##
===========================================
  Coverage       92.43%   92.43%           
===========================================
  Files             315      315           
  Lines           20072    20078    +6     
===========================================
+ Hits            18554    18560    +6     
  Misses           1518     1518           

[jlstevens]

Thanks!

I'll test out and review the functionality shortly.

[jlstevens]

I've done some initial tests and it seems to be working great!

@mattpap My main worry is that I really would love to see this is 3.2!

[mattpap]

Unit tests are update. I still need to finish new tests.

[ianthomas23]

I've only had a quick look and I can review properly tomorrow. But my first observation is that if I use "individual_linked" in the example then I not seeing any different behaviour. Perhaps I am misunderstanding the difference here. Anyway, is it possible to post a video of the different behaviour to make it clearer?

[bryevdv]

"individual_linked" is definitely not evocative to me, I had to go look into the code more closely. IIUC these are the cases:

• False — nothing zooms
• True — every axis zooms together
• individual — just the one single axis zooms independently of anything else
• individual_linked — the one axis and its "partner" axis in the frame zoom together, but independently of anything else.

If we keep this then I think "individual_frame" would be more comprehensible. But I'm also not averse to splitting these dimensions into more properties. i.e. zoom_on_axis stays a simple bool and controls whether there is any zooming at all, and some separate enum property or property control the behaviour, if zooming is turned on. I think that might be cleaner.

E.g. just spitballing, zoom_together = "none" | "frame" | "all" where "none" would just zoom the one axis itself and nothing else, "frame" would also zoom the "cross" axis for the frame, and "all" would be the current "link everything" behavior.

[jlstevens]

I don't have any strong opinions on this but I think a new property might indeed be slightly cleaner as there are two behaviors here: 1. whether to have per-axis zooming at all 2. how this links to other axes when enabled. Even if a new property is not introduced, I do agree that "individual_linked" doesn't convey much to me either.

As the word 'linked' has already been used (and is somewhat intuitive to me), how about a property called zoom_linking (or similar)? The options @bryevdv suggest for this property seems reasonable to me!

[mattpap]

@jlstevens, would you prefer to test this in holoviews as-is or wait for this PR to be refined?

[jlstevens]

@mattpap I assume the tweaks suggested to the API won't change the behavior at all: I'm happy to test with whichever API in holoviews and the sooner the better!

If my assumption is wrong and you think the available set of behavior will change, then in that case, I think it would be best to wait.

[mattpap]

• with individual_linked (both axes either red or blue zoom):

Screencast_00000.mp4

• with individual (only a single axis in the given dimension either red or blue zooms; cross axis is unaffected):

Screencast_00001.mp4

• with True (both red and blue axes in the given dimension zoom; cross axes are unaffected):

Screencast_00002.mp4

[mattpap]

I updated the API, so that zoom_on_axis is left unchanged and new property zoom_together is added. I chose "cross" in favor of "frame". My proposal is to go ahead and merge this, to allow @jlstevens test this in holoviews. We can further refine the API before 3.2. I will separately add more tests for new features.

[mattpap]

I will need some someone to approve this if the plan is sound.

[mattpap]

Test failure is unrelated and fixed in PR #13190 (f5f44b9).

[ianthomas23]

I updated the API, so that zoom_on_axis is left unchanged and new property zoom_together is added. I chose "cross" in favor of "frame". My proposal is to go ahead and merge this, to allow @jlstevens test this in holoviews. We can further refine the API before 3.2. I will separately add more tests for new features.

Thanks for the videos and expanded example, it is all much clearer to me. Functionality and code look fine to me, so let's proceed and, as you say, refine the API before 3.2.

[github-actions]

This pull request has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.