Index class ignores dtype parameter

[CRP]

Apparently Index always forces dtype to object, irrespective of what is specified as parameter.

Pandas 0.4.3

[wesm]

This is intentional for now. The only two supported types are object and integer (int64) at the moment-- is it preventing you from doing something? You are actually the first person who's ever raised this as an issue.

[CRP]

Not really, just something I noticed while trying (unsuccessfully) to Index a dataframe with floats.

BTW, I just started using pandas a few days ago and love it, great job!

[wesm]

Ah, gotcha. Well they'll still be floats inside the object array and lookups will work and all that.

You know, it would be cool to make a Float64Index class- now that Int64Index has been done and integrated so it will play nice with object Indexes it should be largely a copy-paste job. I'll create a separate github issue about this and tag you in it

[lodagro]

Somehow i was completely ignorant of the fact that there are only two supported types (object and int64) for an index.
This breaks my idea to use dtype of an index to check if it holds numeric data.

from pandas/core/index.py Index class

   def is_numeric(self):
        return issubclass(self.dtype.type, np.number)

The only place this little method is used is in Series.plot() and DataFrame.plot().
So what is an eficent way to check of the index holds numeric data? Fixing this will close #954

[adamklein]

There is an inferred_type property that I'm using in the timeseries branch (b/c datetime64 is technically an int64), maybe can use this.

[wesm]

inferred_type is present in master and save to use for this case

In [1]: Index(randn(10)).inferred_type
Out[1]: 'floating'