I am really having a problem when I am trying to plot a 3D scatter plot with multiple points, and I need to look into a particular area to analyse it, but I am not able to zoom in on that point or cluster of points using the scroll wheel and need to use the magnifier that also creates some problem in zooming properly.
Proposed solution
Users would love a scroll wheel zoom because it works well in 2D plots, but implementing it in 3D plots can lack precision. So alternatively, we can use a Box Zoom, where we will draw a rectangle on that area, having areas and boundaries at the perfect points under which we need, and it will zoom to that particular area. This data-centric method provides unambiguous, precise zooming, perfectly framing the desired cluster for detailed inspection
To clarify, are you looking for a scroll-wheel zoom? 3D axes were excluded when that feature was added in #30405, but they could be included.
Or are you looking for the axes to be stretched to precisely fit the area inside the zoom box to what's on screen? That was considered as part of adding zoom behavior to 3D axes, but we ultimately decided against it because it can cause the axes to be stretched in unpredictable ways when you project the 2D box to 3D coordinates. It looks fine when you do the initial zoom, but when you rotate the view you can see the distortion. #23449
I really appreciate that you decided to add the scroll zoom option, but according to me it might not be sufficient as in cluster analysis scroll zoom will not isolate the required points from the non required ones and this might create some further problems. But this can be useful in normal day to day small projects so kindly consider adding it.
According to me the box zoom feature would create distortion only if there is problem in adjusting the aspect ratio. To fix that we would need to develop something that when a person would draw a small rectangle in 2D, it would select the x,y,z axes and adjust the aspect ratio to reduce distortion.It's just a suggestion that I thought could be the reason but real problem could be identified after using it only.
These two features together can be really beneficial. Please feel free to discuss further.
I coded up that behavior (box zoom with changing aspect ratios) when implementing 3D zooming initially, and it worked deterministically and as expected from the starting view angle. The code change to implement it is here: #23449 (comment)
But the stretching of the axes meant that the data looked unexpectedly distorted when the view was rotated. IMO it would be a downgrade in user experience from the current implementation, but I'm open to having my mind changed on that if someone wants to record a demo.
Activity
scottshambaugh commented on Nov 23, 2025
To clarify, are you looking for a scroll-wheel zoom? 3D axes were excluded when that feature was added in #30405, but they could be included.
Or are you looking for the axes to be stretched to precisely fit the area inside the zoom box to what's on screen? That was considered as part of adding zoom behavior to 3D axes, but we ultimately decided against it because it can cause the axes to be stretched in unpredictable ways when you project the 2D box to 3D coordinates. It looks fine when you do the initial zoom, but when you rotate the view you can see the distortion. #23449
DataDetective-Neel commented on Nov 24, 2025
Thanks for your reply.
I really appreciate that you decided to add the scroll zoom option, but according to me it might not be sufficient as in cluster analysis scroll zoom will not isolate the required points from the non required ones and this might create some further problems. But this can be useful in normal day to day small projects so kindly consider adding it.
According to me the box zoom feature would create distortion only if there is problem in adjusting the aspect ratio. To fix that we would need to develop something that when a person would draw a small rectangle in 2D, it would select the x,y,z axes and adjust the aspect ratio to reduce distortion.It's just a suggestion that I thought could be the reason but real problem could be identified after using it only.
These two features together can be really beneficial. Please feel free to discuss further.
scottshambaugh commented on Nov 24, 2025
I coded up that behavior (box zoom with changing aspect ratios) when implementing 3D zooming initially, and it worked deterministically and as expected from the starting view angle. The code change to implement it is here: #23449 (comment)
But the stretching of the axes meant that the data looked unexpectedly distorted when the view was rotated. IMO it would be a downgrade in user experience from the current implementation, but I'm open to having my mind changed on that if someone wants to record a demo.