通过将第一个数据帧的一列与第二个数据帧的两列匹配来合并两个数据帧

Question

I'm working with two data frames:

df1 = {'Metropolitan area': {0: 'New York City',
  1: 'Los Angeles',
  2: 'San Francisco Bay Area',
  3: 'Chicago',
  4: 'Dallas–Fort Worth'},
 'token_nhl': {0: 'Devils',
  1: 'Ducks',
  2: 'Sharks',
  3: 'Blackhawks',
  4: 'Stars'}}

df2 = {'NHL': {0: 'team1', 1: 'team2', 2: 'team3', 3: 'team4', 4: 'team5'},
 'token_nhl': {0: 'Devils', 1: 'Ducks', 2: 'x', 3: 'Stars', 4: 'Sharks'},
 'token_nhl1': {0: 'a', 1: 'b', 2: 'Blackhawks', 3: 'c', 4: 'd'}}

I'm trying to merge them, but I'd like to match the values of the 'token_nhl' columns in df1 with both 'token_nhl' and 'token_nhl1' in df2, so whenever I don't find a value in 'token_nhl', I go look for it in 'token_nhl1',and then the resulting data frame would be:

{'NHL': {0: 'team1', 1: 'team2', 2: 'team3', 3: 'team4', 4: 'team5'},
 'token_nhl_left': {0: 'Devils', 1: 'Ducks', 2: 'x', 3: 'Stars', 4: 'Sharks'},
 'token_nhl1_left': {0: 'a', 1: 'b', 2: 'Blackhawks', 3: 'c', 4: 'd'},
 'token_nhl_right': {0: 'Devils',1: 'Ducks',2: 'Blackhawks',3: 'Stars',4: 'Sharks'}}

Answer 1

For this you need to merge two times:

1: renaming columns, becuase after merge pandas is not giving two different columns

df1 = df1.rename(columns = {"token_nhl":"token_nhl_left"})
df2 = df2.rename(columns = {"token_nhl":"token_nhl_right"})
# creating variables
left_on = "token_nhl_left"
right_on1 = "token_nhl_right"
right_on2 = "token_nhl1"
left_columns = df1.columns

2. merge-1

    df_temp1 = pd.merge(left = df1, right = df2, left_on = left_on, right_on = right_on1, how = 'left')

3. merge-2

    df_temp2 = pd.merge(left = df_temp1[pd.isna(df_temp1[right_on1])][left_columns], right = df2, left_on = left_on, right_on = right_on2, how = 'left')

4. concat

    df_final = pd.concat([df_temp1[pd.notna(df_temp1[right_on1])], df_temp2])

Answer 2

My approach to this question involves two steps.

1 - Create a chunk of code to bring the desired information to a list:

lis = []
for (y,w) in zip(list(df2['token_nhl']), list(df2['token_nhl1'])):
    if y in list(df1['token_nhl']):
        lis.append(y)
    else:
        lis.append(w)

2 - Assign that list to a new data frame with all other needed data. After that, rename columns:

df3 = df2.assign(token_nhl_right=lis)
df3.rename(columns={'token_nhl':'token_nhl_left' ,'token_nhl1':'token_nhl1_left'})