Drawing Bokeh Curve from single data frame row

Datashader
1

Hi,
I am using Datashader and Bokeh to render a series of lines.
These are organised by row with the key columns being [x1, y1, x2, y2]
Is it possible to pass these to holoviews for interactive plotting?
I have managed to get it working by preprocessing to have in the HvCurve format where each entry is
[x1, y1]
[x2, y2]
[nan, nan]

But this is less efficient in terms of memory usage and requires pre-processing which impacts filtering.

If I try

            agg = HvCurve(ddf2, ('x1', 'x2'), ('y1', 'y2'))

Then I get a lot of spurious extra lines.

Essentially, I can render images using:

from datashader.utils import export_image

cvs = ds.Canvas(plot_width=  4*800,
                plot_height= 4*600,
               )

imbuf = cvs.line(ddf2,
                 x=('x1', 'x2'),
                 y=('y1', 'y2'),
                 axis=1, 
                 agg=ds.count('population'),
                 antialias=True
                )

np = tf.spread(tf.shade(imbuf, cmap=cc.fire), px=1)

filename = "plot.png"
export_image(np, filename, background="black")

can I do similar using Holoviews, with one row describing both points of the line?

2

Hello @sk2 , check out the Segments element.

import pandas as pd
import holoviews as hv
from holoviews.operation.datashader import datashade
hv.extension('bokeh')

df = pd.DataFrame(dict(x0=[0,1],y0=[0,1],x1=[1,5],y1=[4,2]))
datashade(hv.Segments(df))

bokeh_plot (1)
I don’t know how you could antialias the datashade version of it though.

1 Like