Best Practice to deal with Apply on Pandas when the Dataframe is empty-CodePudding

A common pattern I follow in my code is creating a function which will be applied row-wise and return a single value. For example:

def apply_function(row):
   if row["A"] == "valueA":
      return "foo"
   elif row["B"] == "valueB":
      return "bar"
   else: 
      return "baz"

df["applied_column"] = df.apply(apply_function, axis=1)

This is fine as long as df is not empty. But if is, .apply returns instead a DataFrame and I get an exception, so I usually follow this boilerplate code:

if not df.empty:
    df["applied_column"] = df.apply(apply_function, axis=1)
else:
    df["applied_column"] = None

This works but I feel is not very elegant and not a good practice. Is it indeed the stardard way of dealing with this situation of Pandas, or there is better practices to achieve it?

CodePudding user response：

You can explicitly set return_type so Pandas does not infer anything from the return value of your apply_function.

E.g.

res = df.apply(apply_function, axis=1, result_type="reduce")

always returns a Series.