ENH: support Categorical hist plotting

[jreback]

s = Series(list('abbc')).astype('category')

This raises

s.hist()

This works

s.value_counts().plot()

I think this should work, beyond this, anyone have ideas on what/how to plot categorical data?

cc @JanSchulz
cc @TomAugspurger
cc @jorisvandenbossche

[jorisvandenbossche]

I don't know if that should work. The natural way to plot a categorical in something hist-like is using a barplot with the frequency of each category I think (s.value_counts().plot(kind='bar')). But, that is not exactly what a histogram is (or at least, not the 'classical' definition). On the other hand: practicallity beats purity?

Also, in R:

> x <- c('a', 'a', 'b', 'c')
> hist(x)
Error in hist.default(x) : 'x' must be numeric
> hist(as.factor(x))
Error in hist.default(as.factor(x)) : 'x' must be numeric

So maybe we should rather have a better error message here? (which also holds True for string series):

In [14]: s.astype(object).hist()
...
TypeError: cannot concatenate 'str' and 'float' objects

So maybe we should check for numeric dtype in hist?

[wesm]

What do you think about making s.hist() do s.value_counts().plot.bar()?

[wcneill]

Hi, newb here. I was going to submit a bug report, but I found this. It does seem the same, but before I walk away without reporting, I wanted to make sure that my issue is actually different:

The docs for DataFrame.hist() say:

This function calls matplotlib.pyplot.hist(), on each series in the DataFrame, resulting in one histogram per column.

However, what is displayed is a set of histograms for numerical data only. This lead me to believe that maybe matplotlib.pyplot.hist() only works on numeric data. However, I tried it anyway and found that

matplotlib.pyplot.hist(df['categorical column with strings/objects'])

results in a beautiful histogram.

I also tried calling .hist directly on a series with categorical data like so:

df['categorical string data].hist()

And also got great results.

Again, not trying to step on any toes, just a new programmer trying to contribute. Let me know if this issue is already covered here :)