It seems as though Bokeh’s high-level charting API, for instance bokeh.charts.Bar, does not respect the order inherent in a Pandas Categorical Index. One data series I was looking at contained Moody’s ratings (see https://en.wikipedia.org/wiki/Bond_credit_rating#Credit_rating_tiers for the scale). Alphabetical order would place the “A1” value first. When I use bokeh.charts.Bar, in fact, A1 is the first value on the left of the chart.
To solve this, I attempted to use an ordered Pandas categorical index for the series I pass in, which explicitly places “Aaa” first, and “Aa1” second, etc., but this does not seem to change anything in the resulting bar plot.
I will post example code and plots shortly, just wanted to start this discussion, to see what others’ experience has been.