ENH: Set default storage for strings globaly

[taoufik07]

Is your feature request related to a problem?

As far as I know, the only way to use the new string dtypes, is to set them manually either via the param dtype if it's supported or using astype otherwise the string will be stored as object.

Describe the solution you'd like

I'm aware that the string dtype is experimental, but it would be easy for us to play with it in large projects if there's a way to change the default behavior globally which can also be used in the future to set the storage for StringDtype.

[lithomas1]

Are you asking about using the StringArray by default when creating DataFrames or about using the setting the storage for StringArrays?
If you want to make StringArrays use the pyarrow based storage I think you can use pd.set_option(""string_storage", "pyarrow").

[taoufik07]

Yes that and the ability to infer string as string and not object

>>> df = pd.DataFrame(["1", "2", "3"])
>>> df
   0
0  1
1  2
2  3
>>>
>>> df.dtypes
0    object
dtype: object # string instead

[lithomas1]

Hmmm. I don't think we have a way to specify to infer the string datatype in the dataframe constructor.
I guess you could call .convert_dtypes() on the frame afterwards which would convert string columns to the string datatype and leave the other columns alone, if that's a suitable workaround.

[mroeschke]

There is future.infer_string so I think we can close this out