In the case of the index being a DateTimeIndex, I think there's a nice easy default case for plotting a categorical column: Let each category assume one value on the y-axis (user-providable, or range(len(categories)) by default) and plot a symbol ('*' by default maybe) at every point in time the categorical value has an entry. The y-axis tick labels should be the categorical values of course. This can provide nice overviews in astronomy or other remote sensing sciences for example to indicate what kind of observation happened when and in what sequence. Or which of some set of lab experiments was executed when.
What do you guys think?
Comment From: jreback
can you show a (hacked) example of what you mean?
Comment From: michaelaye
So, I got through most of it, until I was stuck on how to control the ticklabels. Searching Google brought up this person's gist which basically exactly does what I asked for:
https://gist.github.com/amanahuja/0c2b6f086eed9d6c2fe0
My version is prettier (due to extra spaces for the y-axis and bigger symbols), but less general as I don't automatically determine the integers via a range over the set of categorical values:
http://nbviewer.ipython.org/gist/michaelaye/3ddb6fbc625617e6a663
Comment From: shoyer
It looks like seaborn.stripplot will handle plotting categorical data along one axis: https://github.com/mwaskom/seaborn/pull/410
Comment From: jreback
looks sexy!
Comment From: TomAugspurger
Going to close this in favor of more specific issues. The basics are working now (bar and scatter).