How to create new column with multiple conditions in python

Answering first question:

• First copy all prices into price_new
• Then use df["price_new"] as the default value in np.where calls

df["price_new"] = df["price"]
df["price_new"] = np.where(condition, value, df["price_new"])

Answering second question:

You can reduce code by putting all conditions and corresponding values into a suitable data structure (here list of named tuples), and then use it to generate boolean masks (condlist) used by np.select that will compute your result.

Note that I use dicts and dict.setdefault when computing masks so masks that appear more than once are only computed once.

from collections import namedtuple

Condition = namedtuple("Condition", ["year", "type1", "type2"])
Mapping = namedtuple("Mapping", ["condition", "value"])

mappings = [
    Mapping(Condition(year=2015, type1="apple", type2="natural"), value=25),
    Mapping(Condition(year=2016, type1="apple", type2="natural"), value=26),
    Mapping(Condition(year=2015, type1="apple", type2="organic"), value=20),
    Mapping(Condition(year=2016, type1="apple", type2="organic"), value=22),
]

conditions, values = zip(*mappings)

years, types1, types2 = {}, {}, {}
condlist = [
    (
        years.setdefault(year, df["year"] == year)
        & types1.setdefault(type1, df["type1"] == type1)
        & types2.setdefault(type2, df["type2"] == type2)
    )
    for year, type1, type2 in conditions
]

df["price_new"] = np.select(condlist, values, default=df["price"])

Check if it works as expected:

expected = [25, 20, 26, 22, 20, 15, 20, 18]
print(f"{np.all(df['price_new'] == expected) = }")  # True

Old answer:

You can put all your conditions and matching values into a suitable data structure, and then process with computation.

Here I choose a list of tuples (named tuples for clarity) where first element of the tuple is the condition and second element is the corresponding value.

from collections import namedtuple

Mapping = namedtuple("Mapping", ["where", "value"])

mappings = [
    Mapping(
        where=(
            (df["year"] == 2015)
            & (df["type1"] == "apple")
            & (df["type2"].isin(["natural"]))
        ),
        value=25,
    ),
    Mapping(
        where=(
            (df["year"] == 2016)
            & (df["type1"] == "apple")
            & (df["type2"].isin(["natural"]))
        ),
        value=26,
    ),
    Mapping(
        where=(
            (df["year"] == 2015)
            & (df["type1"] == "apple")
            & (~df["type2"].isin(["natural"]))
        ),
        value=20,
    ),
    Mapping(
        where=(
            (df["year"] == 2016)
            & (df["type1"] == "apple")
            & (~df["type2"].isin(["natural"]))
        ),
        value=22,
    ),
]

From there you can either:

• Rewrite the previous code as a for loop:

df["price_new_1"] = df["price"]
for where, value in mappings:
    df["price_new_1"] = np.where(where, value, df["price_new_1"])

• Use np.select to compute all at once:

condlist, choicelist = zip(*mappings)
df["price_new_2"] = np.select(condlist, choicelist, default=df["price"])

Check if it works as expected:

expected = [25, 20, 26, 22, 20, 15, 20, 18]
print(f"{np.all(df['price_new_1'] == expected) = }")  # True
print(f"{np.all(df['price_new_2'] == expected) = }")  # True