Zoom reset when the slider of an Image is moved

[maximlt]

When an image has sliders, moving the sliders when zoomed in resets the zoom to the initial view.

import xarray as xr
import holoviews as hv
import hvplot.xarray

ds = xr.tutorial.open_dataset('air_temperature')
hvplot_dn_image = ds.hvplot('lon', 'lat')
hvplot_dn_image

This does not happen with this piece of code using Holoviews directly, i.e. the zoom is not changed when zoomed in and the slider is moved.

hvds = hv.Dataset(ds)
hv_dn_image = hvds.to(hv.Image, kdims=['lon', 'lat'], dynamic=True)
hv_dn_image

[ZZMitch]

Hello,

Is there an equivalent way to work around this bug if the Image is RGB (https://hvplot.holoviz.org/reference/xarray/rgb.html)? For example, I have an image time-series visualized with hvplot.rgb and the zoom resets when image slider for time is moved.

    tsPlot = imgCube.hvplot.rgb(x = 'x', 
                                y = 'y', 
                                bands = 'band', 
                                data_aspect = 1, 
                                frame_width = 600,
                                yaxis = None,
                                xaxis = None).opts(tools = ['crosshair'])

This works great (imgCube is an xarray.DataArray with 3 bands and various number of steps in time/x/y dimensions) and displays within a second or two even for large/long time-series - the only issue I have is with the zoom resetting described in this Issue.

I could not work out an equivalent set up to your second example with dataset.to(hv.Image ...) using hv.RGB instead. I found this example (https://stackoverflow.com/questions/60613662/holoviews-how-to-wrap-multiple-rgb-images-into-a-dataset-and-display-them-corre) that could work in theory, but it is very slow if the time-series has hundreds of images.

Thanks for your great work on this package!

[hoxbro]

To not update the axis in the original example, you can set

In the original example, the reset can be stopped by .opts(framewise=False). I would assume that would also be the case for RGB.

https://user-images.githubusercontent.com/19758978/233304871-30d5c422-4e11-4317-9918-7e213b508781.mp4

[ZZMitch]

That worked great, thanks for the quick response!

[iuryt]

Is this really a bug? Maybe setting up framewise = False by default would solve it?

[Evidlo]

This is not working for me, though the API I'm using is slightly different:

#!/usr/bin/env python3
# Very basic test with hvplot and fake data

import panel as pn
import xarray as xr
import hvplot.xarray
import hvplot
import numpy as np
import holoviews as hv

shape = (20, 100, 100)
dst = xr.Dataset(
    {
        'image': (['time', 'x', 'y'], np.random.random((5, 10, 10))),
        'profile': (['time', 'r'], np.random.random((5, 10)))
    },
    coords={
        'time': np.arange(5),
        'x': np.arange(10),
        'y': np.arange(10),
        'r': np.arange(10),
    }
)

slider = pn.widgets.DiscreteSlider(name='Time', options=list(dst.time.data))

im = dst.image.interactive.isel(time=slider).hvplot.image(
    colorbar=True,
    data_aspect=1
).opts(framewise=False)
li = dst.profile.interactive.isel(time=slider).hvplot.line(
    x='r'
)

row = pn.Row(im.panel(), li.panel())
app = pn.Column(im.widgets(), row)
app.show()