Indexes support the set operations, union, intersection, difference, and symmetric difference:

>>> c1 = columns[:4]
>>> c1
Index(['INSTNM', 'CITY', 'STABBR', 'HBCU'], dtype='object')

>>> c2 = columns[2:6]
>>> c2
Index(['STABBR', 'HBCU', 'MENONLY'], dtype='object')

>>> c1.union(c2) # or `c1 | c2`
Index(['CITY', 'HBCU', 'INSTNM', 'MENONLY', 'RELAFFIL', 'STABBR'], dtype='object')

>>> c1.symmetric_difference(c2) # or `c1 ^ c2`
Index(['CITY', 'INSTNM', 'MENONLY'], dtype='object')

Indexes share some of the same operations as Python sets. Indexes are similar to Python sets in another important way. They are (usually) implemented using hash tables, which make for extremely fast access when selecting rows or columns from a DataFrame. As they are implemented using hash tables, the values for the Index object need to be immutable such as a string, integer, or tuple just like the keys in a Python dictionary.