profile.py

def row_processor(row):

    if row['lookupColumn'] == 1:
        # add all 3 columns
        value = row['colA'] + row['colB'] + row['colC']
        return value
    elif row['lookupColumn'] == 2:
        # add just 2 specific columns referenced by name
        value = row['colA'] + row['colB']
        return value
    else:
        value = row['colA']
        return value


# now use it in your apply. You don't have to use lambdas in apply i think...
df.apply(row_processor(row), axis=1)  # i think axis 1 is right. I forget which way is up! :)

might need extra handling of all-blank input here. If you go with this approach you could drop a debugger breakpoint on line 2 and inspect what's going on when the code runs.

Right. I also wrote something like this and will probably end up using it. :(