Weird IndexError for plotting after DF row selection

[michaelaye]

I'm doing this:

df[df.CHANNEL=='MUV'].loc['2015'].DET_TEMP.plot()

with the index being a DatetimeIndex and if I don't plot it all works fine:

df[df.CHANNEL=='MUV'].loc['2015'][['CHANNEL','DET_TEMP']].head()

but when I try the plotting I get this error:

---------------------------------------------------------------------------
IndexError                                Traceback (most recent call last)
<ipython-input-84-28500c6c30a0> in <module>()
----> 1 df[df.CHANNEL=='MUV'].loc['2015'].DET_TEMP.plot()

/usr/local/python3/miniconda/lib/python3.4/site-packages/pandas-0.16.2+409.g73dcb95-py3.4-linux-x86_64.egg/pandas/tools/plotting.py in plot_series(data, kind, ax, figsize, use_index, title, grid, legend, style, logx, logy, loglog, xticks, yticks, xlim, ylim, rot, fontsize, colormap, table, yerr, xerr, label, secondary_y, **kwds)
   2526                  yerr=yerr, xerr=xerr,
   2527                  label=label, secondary_y=secondary_y,
-> 2528                  **kwds)
   2529 
   2530 

/usr/local/python3/miniconda/lib/python3.4/site-packages/pandas-0.16.2+409.g73dcb95-py3.4-linux-x86_64.egg/pandas/tools/plotting.py in _plot(data, x, y, subplots, ax, kind, **kwds)
   2331         plot_obj = klass(data, subplots=subplots, ax=ax, kind=kind, **kwds)
   2332 
-> 2333     plot_obj.generate()
   2334     plot_obj.draw()
   2335     return plot_obj.result

/usr/local/python3/miniconda/lib/python3.4/site-packages/pandas-0.16.2+409.g73dcb95-py3.4-linux-x86_64.egg/pandas/tools/plotting.py in generate(self)
    940         self._make_legend()
    941         self._post_plot_logic()
--> 942         self._adorn_subplots()
    943 
    944     def _args_adjust(self):

/usr/local/python3/miniconda/lib/python3.4/site-packages/pandas-0.16.2+409.g73dcb95-py3.4-linux-x86_64.egg/pandas/tools/plotting.py in _adorn_subplots(self)
   1068                                 naxes=nrows * ncols, nrows=nrows,
   1069                                 ncols=ncols, sharex=self.sharex,
-> 1070                                 sharey=self.sharey)
   1071 
   1072         for ax in to_adorn:

/usr/local/python3/miniconda/lib/python3.4/site-packages/pandas-0.16.2+409.g73dcb95-py3.4-linux-x86_64.egg/pandas/tools/plotting.py in _handle_shared_axes(axarr, nplots, naxes, nrows, ncols, sharex, sharey)
   3333         layout = np.zeros((nrows+1,ncols+1), dtype=np.bool)
   3334         for ax in axarr:
-> 3335             layout[ax.rowNum, ax.colNum] = ax.get_visible()
   3336 
   3337         if sharex and nrows > 1:

IndexError: index 5 is out of bounds for axis 1 with size 3

System: Conda, all updates, Pandas: '0.16.2+409.g73dcb95' (73dcb95)

[michaelaye]

The error does not change when I apply the .loc operator in both cases:

df.loc[df.CHANNEL=='MUV'].loc['2015'].DET_TEMP.plot()

[michaelaye]

The selection works in this form:

df.loc[df.CHANNEL=='MUV', ['DET_TEMP']]['2015'].plot()

[jorisvandenbossche]

Would it be possible to provide some reproducible example? A part of the data (or dummy data) that shows the problem?

[sinhrks]

The error is raised by the same part as the problem fixed in #10879. But this looks to be a different problem caused by chained ops. Recommended to select by one op, such as

df.loc[df.index.year == 2015 & df.CHANNEL=='MUV', 'DET_TEMP'].plot()

[michaelaye]

I tried to set up a dummy example, but it does not fail. :(

import pandas as pd
import numpy as np
index = pd.date_range('2012', freq='3m', periods=20)
df = pd.DataFrame(np.random.rand(20,2), index=index, columns=['CASE_TEMP','DET_TEMP'])
df['CHANNEL'] = np.random.random_integers(low=1, high=2, size=20)
df.CHANNEL = df.CHANNEL.map(lambda x: 'MUV' if x==1 else 'FUV')
df.ix[4, 'DET_TEMP'] = np.NAN
df.ix[17, 'CHANNEL'] = np.NAN
df.ix[12, 'CASE_TEMP'] = np.NAN
df[df.CHANNEL=='MUV'].loc['2015'].DET_TEMP.plot()

[michaelaye]

@sinhrks I had to put parentheses around the row selectors, before it would not choke, but then after that with my data i get the same IndexError as above:

df.loc[(df.index.year == 2015) & (df.CHANNEL=='MUV'), 'DET_TEMP'].plot()

@jorisvandenbossche As I wrote above, without the plot() command I don't get an error, also the describe command does not show anything weird about the data:

count    18890.000000
mean       -23.543606
std          0.646694
min        -24.554800
25%        -23.805800
50%        -23.656000
75%        -23.356400
max        -14.368400
Name: DET_TEMP, dtype: float64

but if I store the selection into a Series, and call plot() on it it still fails.
I saved this Series into an hdf file with the handle 'df' and one can access it here:
https://dl.dropboxusercontent.com/u/139035/df_selection_data.h5
(It's only 307 KB).

[mroeschke]

Looks like the link to the data 404s, and we would need a minimal example to reproduce the bug. Going to close due to needing more info but happy to reopen if we get a self-contained example.