When visualizing large datasets with Bokeh, the interaction can become rather slow. To counter this, one can enable WebGL, which allows rendering some glyph types on graphics hardware.
WebGL is a JavaScript API that allows rendering content in the browser via the Graphics Processing Unit (GPU), without the need for plugins. WebGL is standardized and available in all modern browsers.
To enable WebGL, set the plot’s output_backend property to "webgl":
output_backend
"webgl"
p = Plot(output_backend="webgl") # for the glyph API p = figure(output_backend="webgl") # for the plotting API
Only a subset of Bokeh’s objects are capable of rendering in WebGL. Currently supported are the circle and line glyphs, and most markers supported by scatter() (asterisk, circle, square, diamond, triangle, inverted_triangle, cross, circle_cross, square_cross, diamond_cross, x, square_x, circle_x). You can safely combine multiple glyphs in a plot, even if some are rendered in WebGL, and some are not.
scatter()
The performance improvements when using WebGL varies per situation. Due to overhead in some places of BokehJS, we can currently not benefit from the full speed that you might expect from WebGL. This is also something we plan to improve over time.
Glyphs drawn using WebGL are drawn on top of glyphs that are not drawn in WebGL.
When the scale is non-linear (e.g. log), the system falls back to 2D rendering.
Making a selections of markers on Internet Explorer will reduce the size of the markers to 1 pixel (looks like a bug in IE).
import numpy as np from bokeh.plotting import figure, show, output_file N = 10000 x = np.random.normal(0, np.pi, N) y = np.sin(x) + np.random.normal(0, 0.2, N) output_file("scatter10k.html", title="scatter 10k points (no WebGL)") p = figure(output_backend="canvas") p.scatter(x, y, alpha=0.1) show(p)
import numpy as np from bokeh.plotting import figure, show, output_file N = 10000 x = np.random.normal(0, np.pi, N) y = np.sin(x) + np.random.normal(0, 0.2, N) output_file("scatter10k.html", title="scatter 10k points (with WebGL)") p = figure(output_backend="webgl") p.scatter(x, y, alpha=0.1) show(p)
import numpy as np from bokeh.plotting import figure, show, output_file N = 10000 x = np.linspace(0, 10*np.pi, N) y = np.cos(x) + np.sin(2*x+1.25) + np.random.normal(0, 0.001, (N, )) output_file("line10k.html", title="line10k.py example") p = figure(title="A line consisting of 10k points", output_backend="webgl") p.line(x, y, color="#22aa22", line_width=3) show(p)