Cursor x axis value is not shown when using regular timeseries

[lesteve]

Hi,

you should be able to paste the following snippet in a ipython --pylab session

import pandas
index = pandas.date_range(start='2001-1-1', end='2001-1-10', freq='D')
ts = pandas.TimeSeries(index=index, data=index.day)
figure()
ts.plot()
figure()
ts[:3].append(ts[6:]).plot()

Although the labelling is definitely nicer in the regular timeseries case, the cursor x axis value is not shown (bottom right corner with the tk pylab backend)
http://imgur.com/WCe6P,o8jge#0:

The irregular timeseries with the standard matlplotlib plot_date labelling does show a value for the cursor x axis value:
http://imgur.com/WCe6P,o8jge#1

I tried with the qt backend and I get the same results so it appears to be backend independent.

Not a huge deal but I find the cursor x axis value quite useful from time to time when interactively inspecting a plot (zooming, panning, etc...).

[lesteve]

I see this has been tagged as Community, I'd be more than happy to take a look at this if I could get a few pointers to get started.

I actually started looking in matplotlib.axes.AxesSubplot.format_coord (let me know if this was the wrong place to start from) and ended up in pandas.tseries.converter.TimeSeries_DateFormatter.__call__ (ax.fmt_xdata is None in my case but that might be set to a function if we don't want to use the formatter in this case).

    def __call__(self, x, pos=0):
        if self.formatdict is None:
            return ''
        else:
            fmt = self.formatdict.pop(x, '')
            return Period(ordinal=int(x), freq=self.freq).strftime(fmt)

In my case the formatter formatdict is an empty dict so __call__ will always return an empty string. I am not sure I understand the reason for the pop (getting the string for a x coordinate may work only once) or the structure of the formatdict (do you really want one format string per timestamp), although I am guessing this is used in the scikits.timeseries style tick labelling.

Final question what is x supposed to be in this case, is it just an the underlying int64 value of the pandas.Timestamp? What's the best way to get it from a pandas.Timestamp object?

[willfurnass]

A simple, possibly crude solution would be to use something like the following at the end of tsplot(series, plotf, **kwargs) in pandas/tseries/plotting.py

ax.format_coord = lambda t, y : "t = {:%Y-%m-%d %H-%M}    y = {}".format(\
                  datetime.fromtimestamp(t, series.index.to_timestamp().tz)), y)

[jreback]

merged in 0.11