pythonic way to detect specific pandas column type

Question

what is the most pythonic way to select columns based on there dtype? (only needing the columns and not the entire df as with select_dtypes ) .
assuming we are looking for datetime columns, currently we are using :

date_cols = [col for col in df.columns if "datetime" in str(df[col].dtype]

another option is

date_cols = [col for col in df.columns if isinstance(df[col].dtype, pd.core.dtypes.dtypes.XXXX)]

but could not find the correct dtype for dtype('<M8[ns]').
np.datetime64 does not detect it either.

Answer 1

Use pandas.api.types.is_datetime64_dtype, for list of all possible functions check general_utility_functions:

rng = pd.date_range('2017-04-03', periods=10)
df = pd.DataFrame({'Date': rng, 'a': range(10)})  
print (df)

date_cols = [col for col in df.columns if pd.api.types.is_datetime64_dtype(df[col])]
print (date_cols)
['Date']

Simplier is DataFrame.select_dtypes with columns:

date_cols = df.select_dtypes('datetime64').columns
print (date_cols)
Index(['Date'], dtype='object')

Or:

date_cols = df.columns[df.dtypes.map(pd.api.types.is_datetime64_dtype)]
print (date_cols)
Index(['Date'], dtype='object')

Answer 2

I would do:

from pandas.api.types import is_datetime64_dtype

ts = df.dtypes.apply(is_datetime64_dtype)
ts[ts].index