BUG: grouped hist raises AttributeError with single group

[sinhrks]

Includes 3 minor fixes for hist.

• DataFrame.hist raises AttributeError when the target column specifiedwith by kw only contains a single value.

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import pandas.util.testing as tm

n=100
gender = tm.choice(['Male', 'Female'], size=n)
classroom = tm.choice(['A', 'B', 'C'], size=n)
single = tm.choice(['S'], size=n)

df = pd.DataFrame({'gender': gender,
                   'classroom': classroom,
                   'single': single, 
                   'height': np.random.normal(66, 4, size=n),
                   'weight': np.random.normal(161, 32, size=n),
                   'category': np.random.randint(4, size=n)})

df.hist(by='single')
# AttributeError: 'AxesSubplot' object has no attribute 'ravel'

• hist can accept rot kw only when by is specified, but rot actually does nothing. I understand rot value should be used for xticklabels rotation if xrot is not specified?

df.hist(rot=45)
# AttributeError: Unknown property rot
# -> This is OK because we can use xrot kw

df.hist(by='classroom', rot=45)
# It raises no error, but ticks are NOT rotated.

• hist always displays xticklabels on all axes even if specifying sharex=True. Other plots only draws xticklabels on bottom axes if sharex is True.

[jreback]

@TomAugspurger ?

[TomAugspurger]

For me, df.hist(by='classroom', rot=45) does rotate the xlabels.

I'd say rot should be a synonym for xrot in hist. If you one of them, use that for thexticklabels. If you see both raise aValueError. Make sure it's consistent forhist_frame,hist_series, andgrouped_hist`.

[TomAugspurger]

We probably don't need both rot and xrot in the function signature.

I'd just keep it as rot (even though its inconsistent with hist), and in the function body check

xrot = kwargs.get('xrot') or rot

[sinhrks]

Strange. The latest master seems to output following result, and both grouped_hist and _grouped_plot doesn't have rot logic. I'm not sure on what part xticks can be rotated in current code.

axes = df.hist(by='classroom', rot=30)

Whatever, how about remove rot fix from this PR? I only cared to get grouped_hist worked assuming it is a special case of grouped_hist. To make plot functions consistent, it is better to use same logic as theplot and boxplot (not only hist). I'll open separate issue once organized.

[jreback]

@sinhrks pls rebase

[jreback]

this looks ok, @TomAugspurger ?