Custom Layout with Holoviews+Datashader

[gewitterblitz]

Hi, I want to create a layout with 3 rows and 2 columns and the subplot at location (3,2) being transposed (see attached sample plot below). You can ignore the subplot at location (2,2) for now.

All the subplots should be synced so that mouse scroll in one subplot zooms in/out in other subplots too. I could not find an easy solution to it although I did notice many folks asking for similar functionality.

What’s the easiest way to create such a custom layout in holoviews while using datashader functionality?
Here’s my code but it want to customize it by having only subplot in row1 covering two columns

import colorcet as cc

ltg_data = lma.get_active()  # returns a copy of the data, as a Pandas Dataframe
ltg_data = ltg_data[(ltg_data['alt'] <=20)]

points1 = hv.Points(ltg_data,kdims=['lon','alt'])
datashaded1 = hd.datashade(points1,cmap=cc.fire).redim.range(y=(0,20),x=(-87.6,-86.8))
plot1 = hd.dynspread(datashaded1, threshold=0.10, how='over').opts(height=300,width=600)

points2 = hv.Points(ltg_data,kdims=['lat','alt'])
datashaded2 = hd.datashade(points2,cmap=cc.fire).redim.range(y=(0,20),x=(-34,34.4))
plot2 = hd.dynspread(datashaded2, threshold=0.10, how='over').opts(height=300,width=600)

points3 = hv.Points(ltg_data,kdims=['time','alt'])
datashaded3 = hd.datashade(points3,cmap=cc.fire).redim.range(y=(0,20))
plot3 = hd.dynspread(datashaded3, threshold=0.10, how='over').opts(height=300,width=600)

points4 = hv.Points(ltg_data,kdims=['lon','lat'])
datashaded4 = hd.datashade(points4,cmap=cc.fire).redim.range(y=(-34,34.4),x=(-87.6,-86.8))
plot4 = hd.dynspread(datashaded4, threshold=0.10, how='over').opts(height=3![Screen Shot 2021-04-23 at 3.22.51 PM|418x500](upload://l3PFx0FWXP6XvIwfRdYYD21lu6G.png) 00,width=600)

hv.Layout([plot1,plot2,plot3,plot4]).cols(2)

[jbednar]

HoloViews Layouts are very limited, but because they are just a thin wrapper on top of Panel layouts, here I’d use Panel layouts instead. You can choose either hierarchical rows and columns like pn.Row(pn.Column(plot1, plot2, width=500), plot3, width=100), etc., or a Panel GridSpec. Panel doesn’t care whether the underlying plot objects are datashaded; should work fine either way.

[hmijail]

Thank you for the pointer.
I’d suggest that the documentation should be explicit about these limits, instead of just insisting on how easy it is to + and * things… so that we know when to switch to Panels.

In other layout-building tools a working .cols(2) layout could be treated as a unit that can be laid on a new .cols(1) layout, so I just spent some time trying to do just that - only to have them break down in Holoviews.