Rescale / transform y axis of DynamicMap scatter

HoloViews
1

I have an app with streaming data displayed in a scatter plot using hv.DynamicMap. The y data of the scatter is in units of J/mol, but I’d like the users to interact with the data in units of kJ/mol.

Currently I’m rescaling the data with a factor 1e3 before plotting it. The data is plotted with a colormap so I’m also rescaling the clim opt.
However, in other parts of the app I’m also using the same data and colormap and as a consequence the step of rescaling appears in many parts of my app.

I’m trying to rescale/transform the data just before plotting such as described in this hvplot issue here.

How can I do the equivalent thing in HoloViews? I’ve been playing with hv.dim but I seem to be unable to get that to work.

Also, it it possible to somehow pass the norm object to opts directly instead of clim ?

MWE of what I’m currently doing with rescaling:


import holoviews as hv
import numpy as np
import pandas as pd
import panel as pn
from functools import partial
import param
from holoviews.streams import Pipe
from matplotlib.colors import Normalize

hv.extension('bokeh')

class Controls(param.Parameterized):
    cmap = param.Selector(default='turbo', objects=['turbo', 'inferno', 'viridis'])

x = np.arange(50)
y = 10000*(np.random.rand(50)+0.2)
sclf = 1e-3

df = pd.DataFrame({'x': x, 'y': y*sclf})

ymin, ymax = 2000, 12000
norm = Normalize(vmin=ymin, vmax=ymax)
clim = (norm.vmin*sclf, norm.vmax*sclf)

stream = Pipe(data=df)
controls = Controls()

func = partial(hv.Scatter, kdims=[None, 'y'])
plot = hv.DynamicMap(hv.Scatter, streams=[stream])
transform = {'y': hv.dim('y')*1e-3}
s = plot.apply.opts(cmap=controls.param['cmap'], color='y', clim=clim, colorbar=True)

hv_pane = pn.pane.HoloViews(s)
app = pn.Row(hv_pane, pn.panel(controls.param.cmap))

app.servable()

App:

image

2

I’m not sure I’m following all the details here, but a DynamicMap can do any arbitrary function, so if you’re already using DynamicMap, seems like you can fold any rescaling you want into that:

import holoviews as hv
import numpy as np
import pandas as pd
import param
from holoviews.streams import Pipe
from matplotlib.colors import Normalize

hv.extension('bokeh')

x = np.arange(50)
y = 10000*(np.random.rand(50)+0.2)
sclf = 1e-3

df = pd.DataFrame({'x': x, 'y': y})

ymin, ymax = 2000, 12000
norm = Normalize(vmin=ymin, vmax=ymax)
clim = (norm.vmin*sclf, norm.vmax*sclf)

stream = Pipe(data=df)

def scaled_scatter(data): 
    d = data.copy()
    d.y *= sclf
    return hv.Scatter(d)

plot = hv.DynamicMap(scaled_scatter, streams=[stream])
plot.apply.opts(cmap='turbo', color='y', clim=clim, colorbar=True)

image
image622×632 25.4 KB

3

Thanks, I’ve updated my code and I’m rescaling the data before sending it to the plot as you suggest.

I guess my question wasn’t too clear but I’ll try to rephrase. I was expecting the following to work:

ydim = hv.dim('y')*1e-3
plot = hv.Scatter(df, kdims=['x'], vdims=[ydim])
s = plot.apply.opts(cmap=controls.param['cmap'], color=ydim, clim=clim, colorbar=True)

This does not work as a hv.dim is not a valid type for vdims, it must be tuple, str, dict or Dimension. However I can pass the hv.dim to the color keyword argument of opts and does rescale the data before applying the colors.

My question: is there some other construction I can use which I can pass to both vdims and color to rescale both the ydata and the color data?