plotting with index of objects should use `FixedLocator`

[tacaswell]

As pointed out here (https://stackoverflow.com/questions/24468333/set-xlim-for-pandas-matplotlib-where-index-is-string/24477138#24477138) if the index of a data frame is objects (that is not, the index obviously convertible to scalars and the data is being plotted against range(len(df))), the ticklabels are set to be string representations of the objects (I presume via set_xticklabels) but the locator is still a AutoLocator which means if you change the xlimits or pan/zoom the axes, the ticklabels become de-coupled from the data.

A solution is to also use a fixed locator which will pin the ticks to only be on the index locations.

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.ticker as mticker

df = pd.DataFrame({'Foo': pd.Series([2,3,4], index=['2002', '2003', '2004'])})
fig, ax = plt.subplots()

df.plot(ax=ax)

ax.xaxis.set_major_locator(mticker.FixedLocator(np.arange(len(df))))
ax.xaxis.set_major_formatter(mticker.FixedFormatter(df.index))

I can make changes to the code if someone points me to where the relevant section is.

[danielballan]

I believe the change should be made in this method of the MPLPlot class, from which all plots (Line, Bar, Scatter, etc.) inherit.

https://github.com/pydata/pandas/blob/master/pandas/tools/plotting.py#L1045

[jreback]

closing as stale. pls reopen if still an issue.

[tacaswell]

This is still a bug and should be re-opened.

[jorisvandenbossche]

When someone tries to fix this, make sure to test whether #2980, #11465 and #15912 are then solved.

@TomAugspurger @sinhrks do you have an idea where this should best be handled? Somewhere in _make_plot?

[TomAugspurger]

Yeah, _make_plot is where the change is needed, but if we're fixing it in MPLPlot, then we'll need to do some refactoring.