Re: [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....

> 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.

I kind of got this working.  I had to associate the locator with the X
axis and call it's autoscale() method before calling plot.show().  Now
I get %H:%M:%S formatting for everything, even when I'm zoomed way out
on a data set containing two-days worth of time series data.  I
suppose I can fiddle with the AutoDateFormatter's scaled attribute,
but the default looks like it ought to work.  Any thoughts on what I'm
(still) missing?

The locator/formatter code looks like this:

    locator = matplotlib.dates.AutoDateLocator()
    formatter = matplotlib.dates.AutoDateFormatter(locator)
    ...
    left_plot = figure.add_subplot(111)
    left_plot.set_title(title)
    left_plot.set_axisbelow(True)
    left_plot.yaxis.set_major_formatter(pylab.FormatStrFormatter('%g'))
    ...
    locator.set_axis(left_plot.xaxis)
    left_plot.xaxis.set_major_formatter(formatter)
    locator.autoscale()
    pylab.show()

This works fine except for the lack of dynamic scaling and apparently
incorrect choice of labels on plots over large time scales.

Skip