categorical fixup 3

[jankatins]

• check that plot does the right thing, i.e. ordering of the categories
• Make a utils.testing.makeCategorical() Series(np.array(list('abcdefghij')).take(np.random.randint(0,10,size=len(df)))).astype('category') + NaN + type + size + unique kword?
• check for self.values.xxx in Series and check that the methods using such call work when values is a Categorical. See WIP: searchsorted implementation #7447
• Apply and Categoricals: BUG/ENH: categorical returned during a transform #8065
• Categorical.searchsorted() categorical needs a searchsorted implmentation #8420
• docs: "[...] number of possible values (categories; -categories-+levels+ in R). " DOC/FIX: In R it's still levels... #8561
• move assert_categorical_equal to testing.py and fix "categories" failure msg
• check whatsnew: In the note change "levels" to "categories" and s/Schultz/Schulz/
• docstring in API: update pd.cut with the new categorical integration #8077 (comment)
• add a link to "comparison with R" section from http://pandas-docs.github.io/pandas-docs-travis/comparison_with_r.html
• check that the memory info is doing the right thing for categorical (ENH: dataframe memory usage #7619)

[jreback]

https://github.com/ContinuumIO/dynd-python/blob/master/docs/source/type.rst

@JanSchulz FYI ndt is libdynd essentially a modern version of numpy, supporting among things: categorical, missing values in int, var lenstrings

going to integregrate some of this for 0.16 (it coexists with numpy)

[jankatins]

IMO this Bug can be set to 0.15.1 or later, the more urgent ones are done in the individual PRs.

[jreback]

why don't u close this one and roll a new issue for the remaining items

[jreback]

@JanSchulz what's left on this one?

[jankatins]

IMO the rest of the ToDos from above can wait and categorical is ready for
a release if nothing new pops up. I'm away from my computer for the next
week, so hopefully that wont happen...

[jreback]

@JanSchulz optimistic :) hahah

[jankatins]

"move assert_categorical_equal to testing.py" is done in 56dbb8c

Nice :-) Thanks to @immerrr

[shoyer]

Just read the v0.15.0 release notes and noticed that they still use "levels" in a couple of spots instead of "categories" :(.

see that third warning message: http://pandas.pydata.org/pandas-docs/stable/whatsnew.html#v0-15-0-october-18-2014

[jreback]

@jorisvandenbossche can you address @shoyer comment when you repush the docs? thxs

@shoyer joris is rebuilding the docs (and repushing) anyhow...thxs

[shoyer]

also see the second bullet here: http://pandas.pydata.org/pandas-docs/stable/whatsnew.html#categoricals-in-series-dataframe

[jorisvandenbossche]

@shoyer thanks for noting! I added fixes for them in #8586, and also added a bullet point for the rename levels -> categories (which was actually missing as far as I could see)

[TomAugspurger]

Most of these have been addressed / aren't a priority now.