[Matplotlib-users] Dynamically change X time format while zooming plot?

[Skip M. <sk...@po...>]

I have a small matplotlib app I wrote to plot columns of a CSV files.
The X axis is almost always time. Once displayed, I will often zoom in
on a small patch of a plot.  I'm currently selecting the strftime
format based on the original time range of the input.  As I zoom in,
however, that doesn't work so well.  Consider the attached plot.

How do I trigger a recalculation of the X axis format? I see this
class, which looks promising:

http://matplotlib.org/api/dates_api.html#matplotlib.dates.AutoDateFormatter

but when I use it in the obvious way, all I get is the current year
for all tick labels, despite the fact that the scaled attribute of the
formatter has keys which are much smaller than a year.

    formatter = matplotlib.dates.AutoDateFormatter(
                matplotlib.dates.AutoDateLocator())
    print formatter.scaled

The default scaled attribute displays as:

    {1.0: '%b %d %Y', 365.0: '%Y', 30.0: '%b %Y',
0.041666666666666664: '%H:%M:%S'}

I'm plotting data now with a range of about four hours, so I think the
last format (%H:%M:%S) should be used.

Thx,

Skip