How to set crs of datashader image; Or how to stop skewing of plot

GeoViews
,
1

When I use holoviews.streams.Buffer and hv.DynamicMap(gv.Points, streams=[dfstream]) to stream point geometries onto a geoviews.tile_source it works just fine. However what I really want to do is datashade the points before plotting on the map but this seems to have a strange effect on the plot projection.

Edit to add this version info:
Geoviews: 1.8.1
Holoviews: 1.13.3

Here is an example of streaming the point geometries.

import param, geoviews as gv, holoviews as hv
import pandas as pd, numpy as np
from datashader.utils import lnglat_to_meters as webm
from holoviews.operation.datashader import datashade

gv.extension('bokeh',logo=False)

class Mapview(param.Parameterized):
    tiles         = gv.tile_sources.CartoEco()
    template_df   = pd.DataFrame({'lng': [], 'lat': []}, columns=['lng', 'lat'])
    dfstream      = hv.streams.Buffer(template_df, index=False, length=100000, following=False)
    points        = hv.DynamicMap(gv.Points, streams=[dfstream])
    map_layout    = tiles * points
   #map_layout = tiles * datashade(points) 

    def show_map(self):
        self.map_layout.opts(
            gv.opts.WMTS(global_extent=True,width=800,height=400,show_grid=False,xaxis=None,yaxis=None),
            gv.opts.Points(size=5, color='green', fill_alpha=0.3, line_alpha=0.4)
        )        
        return self.map_layout.opts(data_aspect=0.5)
    
m = Mapview()
m.show_map()

And then I can stream data onto it like so:

lng = np.random.uniform(-120.24150,-70.796140, 10000)
lat = np.random.uniform(14.6256105,60.7169661, 10000)
df = pd.DataFrame(list(zip(lng,lat)),columns=['lng','lat'])
m.dfstream.send(df[['lng','lat']])

And it works just fine; it looks like this.

image
image851×413 123 KB

If you switch the commented map_layout declaration in the Mapview() class to get a datashaded image on the map and run it again things go a bit strange. On first glance it appears to have worked:

image
image815×403 120 KB

However, using the zoom tools on the plot results in strange skewing which does not happen with the non-datashaded points.

image
image854×414 79.5 KB

I tried converting the x,y values to meters before streaming to the map

df.loc[:,'x_meters'], df.loc[:,'y_meters'] = webm(df.lng, df.lat)

…and changing the template_df columns in Mapview() appropriately but then nothing at all appears on the map.

How do I prevent the map from skewing in this scenario?

2

Solved. Specifying width and height arguments to datashader resolves the problem of the map plot skewing/stretching. Just make sure the width & height provided to datashader match the values for the map plot. Using my example above, this now works:

map_layout = tiles * datashade(points,width=800,height=400)

And remove data_aspect=0.5 from the plot.opts so the plot is produced like this instead.

return self.map_layout