bokeh: In python3, rectangle does not appear when x axis is type datetime

I’ve been having trouble plotting rectangular glyphs with bokeh in Python 3. I can’t find this issue documented anywhere, or any efforts to correct it.

I’m trying to plot rectangles whose x-axis extent is given as a timedelta, and whose positions are datetimes.

This is a minimum example that produces the error. I generate a set of times via numpy’s datetime64, convert them to datetime objects, and plot the set of rectangles. ~No rectangles appear in the plot if the times are close to the current UTC, but if I shift them back 50 years they appear just fine.~ This was in error – The rectangles appear, but their sizes vary. If I make them narrower by a factor of 10, they disappear altogether.

#!/bin/env python
from datetime import datetime, timedelta  
import numpy as np
from bokeh.plotting import figure, show
from bokeh.models import DatetimeTickFormatter
  
## The line below is the key -- If I subtract 50 years from the current date, the rectangles appear just fine.
## If i don't subtract them, the rectangles are not rendered.
t0 = datetime.utcnow() - timedelta(weeks=52 * 50)
 
times = t0.timestamp() * 1e3 + np.linspace(0, 5, 100, endpoint=False) * 3600 * 1e3
times = times.astype('datetime64[ms]')
  
plot = figure(
      output_backend='svg',
      plot_height=500,
      plot_width=1280,
      x_axis_label='UTC Time',
      y_axis_label="",
      y_axis_type='linear'
)
  
plot.xaxis.formatter = DatetimeTickFormatter(
      hours=['%H:%M'],
      minutes=['%H:%M'],
      seconds=['%H:%M']
)
  
dm = np.linspace(0, 100, 20)
width = (np.diff(times)[0] * 0.8).astype(timedelta)
times, dm = np.meshgrid(times, dm)
plot.rect(
      x=times.astype(datetime).flatten(),
      y=dm.flatten(), width=width,
      height=100 / 20, name='rect'
)
  
show(plot)

Note the indicated line right after the import statements – it seems that bokeh is having trouble with recent dates in Python3, probably because their unix times in milliseconds are very large. If I shift back to 1970, as shown here, the error goes away.

I suspect that this function may be part of the problem. The internal serialization in bokeh converts all datetimes to milliseconds since the unix epoch, which might be overflowing something.