Link z axis of holoviews plot inside panel

Panel
1

Hello,

I have a problem to share the z axis of dynamic QuadMesh inside a panel
Here is an example:

import panel as pn
import numpy as np
import holoviews as hv
from functools import partial
hv.extension("bokeh")
pn.extension()

radius = np.r_[np.linspace(100, 200, 50)]
theta = np.r_[np.linspace(0, 2*np.pi, 100)]
Radius,Theta = np.meshgrid(radius, theta)
X,Y = Radius*np.cos(Theta), Radius*np.sin(Theta)
data1 = np.random.rand(*X.shape, 10)
data2 = np.random.rand(*X.shape, 10) * (np.arange(10)+1)[None, None,:]
def make_quad(index, X, Y, data):
    quad = hv.QuadMesh((X,Y,data[:,:,index]))
    return quad.opts(frame_width=400, colorbar=True, aspect=1)

index_selector = pn.widgets.IntSlider(value=5, start=0, end=9)
dmap1 = hv.DynamicMap(pn.bind(partial(make_quad, X=X, Y=Y, data=data1), index_selector))
dmap2 = hv.DynamicMap(pn.bind(partial(make_quad, X=X, Y=Y, data=data2), index_selector))
pn.Column(index_selector, pn.Row(pn.pane.HoloViews(dmap1, linked_axes=True), pn.pane.HoloViews(dmap2, linked_axes=True)))

share_z

I’d like both quadmesh share the same colorbar
I tried to use a hv Layout for both dynamicmaps
pn.Column(index_selector, dmap1 + dmap2)
but it works only on the first render, if the value of the slider is changed colorbar are not shared anymore
share_z_layout

2

I could get it to work with the following:

import panel as pn
import numpy as np
import holoviews as hv
from functools import partial
hv.extension("bokeh")
pn.extension()

radius = np.r_[np.linspace(100, 200, 50)]
theta = np.r_[np.linspace(0, 2*np.pi, 100)]
Radius,Theta = np.meshgrid(radius, theta)
X,Y = Radius*np.cos(Theta), Radius*np.sin(Theta)
data1 = np.random.rand(*X.shape, 10)
data2 = np.random.rand(*X.shape, 10) * (np.arange(10)+1)[None, None,:]

def make_quads(index):
    quad1 = hv.QuadMesh((X,Y,data1[:,:,index]))
    quad1.opts(frame_width=350, colorbar=True, aspect=1)
    quad2 = hv.QuadMesh((X,Y,data2[:,:,index]))
    quad2.opts(frame_width=350, colorbar=True, aspect=1)
    return (quad1 + quad2)

index_selector = pn.widgets.IntSlider(value=5, start=0, end=9)
pn.Column(index_selector, pn.bind(make_quads, index_selector))

But if I change the last line to pn.Column(index_selector, hv.DynamicMap(pn.bind(make_quads, index_selector))) it breaks the colorbar sync.

Hope this helps.

3

Actually I’d like to have the DynamicMaps since it’s more reactive, and I’d like to be able to use panel layout to put results in different tabs

4

my solution would be:

import param
import panel as pn
import numpy as np
import holoviews as hv
from functools import partial
hv.extension("bokeh")
pn.extension()

radius = np.r_[np.linspace(100, 200, 50)]
theta = np.r_[np.linspace(0, 2*np.pi, 100)]
Radius,Theta = np.meshgrid(radius, theta)
X,Y = Radius*np.cos(Theta), Radius*np.sin(Theta)
data1 = np.random.rand(*X.shape, 10)
data2 = np.random.rand(*X.shape, 10) * (np.arange(10)+1)[None, None,:]
def make_quad(index, X, Y, data):
    quad = hv.QuadMesh((X,Y,data[:,:,index]))
    return quad.opts(frame_width=400, colorbar=True, aspect=1)

index_selector = pn.widgets.IntSlider(value=5, start=0, end=9)
clim = pn.widgets.RangeSlider()
@pn.depends(index_selector, watch=True)
def update_clim(index):
    d = np.stack((data1[:,:,index], data2[:,:,index]))
    cmin, cmax = d.min(), d.max()
    with param.batch_watch(clim):
        clim.start = cmin
        clim.end = cmax
        clim.value = (cmin, cmax)
        clim.step = (cmax-cmin)/100
update_clim(index_selector.value)
dmap1 = hv.DynamicMap(pn.bind(partial(make_quad, X=X, Y=Y, data=data1), index_selector)).apply.opts(clim=clim.param.value)
dmap2 = hv.DynamicMap(pn.bind(partial(make_quad, X=X, Y=Y, data=data2), index_selector)).apply.opts(clim=clim.param.value)
pn.Column(index_selector, clim, pn.Row(pn.pane.HoloViews(dmap1), pn.pane.HoloViews(dmap2)))

in the same time I can adjust the z axis but I see this more like a workaround than the expected solution

5

@xavArtley Could you file a Panel issue? We should investigate actually reusing the same color mapper instances.

1 Like
6

I properly wasn’t that clear in my last post, but what I wanted to say is it works as you would expect without DynamicMap but breaks the colorbar sync when you add it. So it is highly likely a more advanced layout with separate plots and in taps will still not work.