ENH: set_index drops index name

[jseabold]

Index

df = pd.util.testing.makeDataFrame()
df.index.name = 'name'
df.set_index(df.index)
df.set_index(df.index).index.names

MultiIndex

from statsmodels.datasets import grunfeld
data = sm.datasets.grunfeld.load_pandas().data
data = data.set_index(['firm', 'year'])
data.set_index(data.index).index.names

Version

[~/]
[33]: pd.version.version
[33]: '0.13.1-254-g150f323'

[jseabold]

I guess I shouldn't be using set_index to do this. Use reindex. Might want to either raise an error or pass along to reindex under the hood?

[jreback]

This is technically not valid according to the doc-string, as it accepts arrays which is sufficiently ambiguous, but 'works' with Series/ndarray (and index happens to be a ndarray subclass currently). So I think this could easily work; marking as an enhancement.