如何加快在 Bokeh 中创建的图的页面加载速度？

Question

After I run my code, the page opens up and for 13 seconds a blank page shows up, and after that, the expected plots.

Is there a way to speed this up? (13 seconds, user wise is sadly too long)

The code produces 44 plots, in two columns(2n loop), differ by parameter (1st loop) and then differ by tool name (3 rd loop)

Also if bokeh, isnt the right tool for this, would happily open to hear a way to plot interactive plots, preferably with python

This is the code:

dateparse = lambda x: pd.datetime.strptime(x, '%Y-%m-%dT%H:%M:%S')

df = pd.read_csv("LUSU.csv",parse_dates=['PM_START_DATE'], date_parser=dateparse)

df.head()
print("time elapsed: {:.2f}s".format(time.time() - start_time))
x = df['PM_START_DATE']
y = df['TASK_VALUE']
tool=df['ENTITY']
tool_list=df['ENTITY'].unique()
param_list=df['PARAMETER'].unique()
#
print("time elapsed: {:.2f}s".format(time.time() - start_time))
colors = itertools.cycle(Spectral11)
output_file('LUSU.html', mode="cdn")
for i in range(0,44,2):
    row = []
    for _ in range(2):
        p = figure(title=param_list[i], x_axis_label='date', y_axis_label='chart value', x_axis_type="datetime", toolbar_location="below")
        for j in range(len(tool_list)):
            df1=((df['PARAMETER']==param_list[i] )& (df['ENTITY']==tool_list[j] ))
            source = ColumnDataSource(data=dict(x=x.loc[df1], y=y.loc[df1], tool=tool.loc[df1]))
            p.line(x='x', y='y',legend='tool', source=source)
            p.scatter(x='x', y='y',legend='tool',size=10,color=next(colors), source=source)
        p.add_tools(HoverTool(tooltips=[("Entity", "@tool"), ("Chart Value", "@y{%0.2f}"), ("Date", "@x{%F}")], formatters={"x": "datetime", "y": 'printf'}))
        #p.xaxis.formatter = DatetimeTickFormatter(days=["%m/%d/%Y"])
        p.legend.location = "top_right"
        p.legend.click_policy = "mute"
        row.append(p)
        i = i + 1
    grid.append(row)

fig=layout(grid)
reset_output()

show(fig)

Answer 1

You could try to use webGL to speed it up. This allows glyphs to render via the GPU.

figure(title=param_list[i], x_axis_label='date', y_axis_label='chart value', x_axis_type="datetime", toolbar_location="below", output_backend="webgl")

More information: https://docs.bokeh.org/en/latest/docs/user_guide/webgl.html .

Further reading about Bokeh performance issues with a lot of plots: https://github.com/bokeh/bokeh/issues/6294 .