如何使用优先级键将两个键上的熊猫数据帧连接在一起？

Question

How can I left join two pandas dataframes ( df1 , df2 ) on two keys ( bla1 , bla2 ), where the bla2 key should be used whenever it's not null (see last two rows in df1 )?

Pseudo-code

if bla2 is not null then join bla on bla2

else join bla on bla1

Dataframes

df1

| bla1 | bla2 | a   | b   |
|------|------|-----|-----|
| 1    |      | ... | ... |
| 2    |      | ... | ... |
| 3    |      | ... | ... |
| 4    | 7    | ... | ... |
| 5    | 8    | ... | ... |

+ df2

| bla | x   | y   | z   |
|-----|-----|-----|-----|
| 1   | ... | ... | ... |
| 2   | ... | ... | ... |
| 3   | ... | ... | ... |
| 7   | ... | ... | ... |
| 8   | ... | ... | ... |

= df3

| bla1 | bla2 | a   | b   | x   | y   | z   |
|------|------|-----|-----|-----|-----|-----|
| 1    |      | ... | ... | ... | ... | ... |
| 2    |      | ... | ... | ... | ... | ... |
| 3    |      | ... | ... | ... | ... | ... |
| 5    | 7    | ... | ... | ... | ... | ... |
| 4    | 8    | ... | ... | ... | ... | ... |

Answer 1

[First create a new column to combine both columns.

df1["new_column"] = df1.bla2.fillna(df1.bla1);

Then join both frames and drop extra created columns.

df3 = pd.merge(df1, df2, how="inner", left_on="new_column", right_on="bla").drop(["new_column", "bla"], axis=1)