for my figure I wanted to use the 'datetime' option in bokeh as following:
top = figure(width=900, height=500, x_axis_type='datetime')
My data for the x-axis is in the datetime.time format.
x_time = [datetime.time(0, 0), datetime.time(0, 0, 3), datetime.time(0, 0, 13), datetime.time(0, 0, 23), datetime.time(0, 0, 26)]
However it produces the following error when trying to add:
top.image_url(x=datetime.time(0,0,3), y= 10 url = [some_url]]
and
top.add_layout(Arrow(x_start=datetime.time(0,0,0), y_start=5,
x_end=datetime.time(0,0,3), y_end=10)
ValueError: expected an element of either String, Dict(String, Either(String, Instance(Transform), Instance(ColorMapper), Float)) or Float, got datetime.time(0, 0)
As suggested from Rutger Kassies I transformed my data to microseconds, now it only shows seconds: Change from seconds to minutes
It seems the Bokeh annotations only take numbers, not Datetime
or Time
objects. A workaround is to convert your times to microseconds and use those to plot.
An example:
from bokeh.plotting import figure, show, output_notebook
from bokeh.models import Arrow
import datetime
def time_to_microseconds(t):
dmin = datetime.datetime.min
dummy_tdelta = (datetime.datetime.combine(dmin, t) - dmin)
return dummy_tdelta.total_seconds()*1000
x_time = [datetime.time(0,0,1),
datetime.time(0,0,2),
datetime.time(0,0,3),
datetime.time(0,0,4),
datetime.time(0,0,5)]
top = figure(width=300, height=300, x_axis_type='datetime')
# a line works fine with time objects
top.line(x_time, range(len(x_time)))
# layout needs numbers
top.add_layout(Arrow(x_start=time_to_microseconds(datetime.time(0,0,2)),
y_start=3,
x_end=time_to_microseconds(datetime.time(0,0,3)),
y_end=2))
edit:
You can change the tick-formatting with:
from bokeh.models import DatetimeTickFormatter
top.xaxis.formatter = DatetimeTickFormatter(seconds=["%M:%S"],
minutes=["%M:%S"],
minsec=["%M:%S"],
hours=["%M:%S"])