Pandas groupby, contains and sum

Question

                             BUILDING                 UNITS
0                      01 ONE FAMILY DWELLINGS        168.0
1                      02 TWO FAMILY DWELLINGS        269.0
2                    03 THREE FAMILY DWELLINGS        234.0
3                        04 TAX CLASS 1 CONDOS          0.0
4                       06 TAX CLASS 1 - OTHER          0.0
5               07 RENTALS - WALKUP APARTMENTS      10232.0
6             08 RENTALS - ELEVATOR APARTMENTS      15136.0
7                 09 COOPS - WALKUP APARTMENTS          0.0
8               10 COOPS - ELEVATOR APARTMENTS          0.0
9                            11A CONDO-RENTALS          0.0
10               12 CONDOS - WALKUP APARTMENTS          0.0
11             13 CONDOS - ELEVATOR APARTMENTS          0.0
12                      14 RENTALS - 4-10 UNIT        534.0
13           15 CONDOS - 2-10 UNIT RESIDENTIAL          0.0

I want to be able to group things that contain a string. For example, here, I'd want to group all the condos example into one category with a a sum of the units, the same, rentals would be grouped as rentals. The remaining ones that don't fall into any related category would be grouped as other.

I know that this selects those which contain condo, but how do you group them and sum their units as one category?

df[df['BUILDING'].str.contains("CONDO")]

I'm looking for something like so:

Building Units
Other      #
Condos     #sum condos
Rentals    #sum rentals

Answer 1

Use numpy.select to create a marker for building units, and use it for groupby operations

df["BUILDING UNITS"] = np.select(
    condlist=[
        df["BUILDING"].str.contains("CONDO"),
        df["BUILDING"].str.contains("DWELLINGS"),
    ],
    choicelist=[
        "Condos", "Rentals"
    ],
    default="Other"
)

df.groupby("BUILDING UNITS").sum()

---

                  UNITS
BUILDING UNITS
Condos              0.0
Other           25902.0
Rentals           671.0