R dplyr full_join - 没有公共键，需要公共列混合在一起

Question

例如，我有这两个数据框：

dates = c('2020-11-19', '2020-11-20', '2020-11-21')
df1 <- data.frame(dates, area = c('paris', 'london', 'newyork'), 
                  rating = c(10, 5, 6),
                  rating2 = c(5, 6, 7))

df2 <- data.frame(dates, area = c('budapest', 'moscow', 'valencia'), 
                  rating = c(1, 2, 1))

> df1
       dates    area rating rating2
1 2020-11-19   paris     10       5
2 2020-11-20  london      5       6
3 2020-11-21 newyork      6       7
> df2
       dates     area rating
1 2020-11-19 budapest      1
2 2020-11-20   moscow      2
3 2020-11-21 valencia      1

使用 dplyr 执行外部连接时：

df <- df1 %>%
  full_join(df2, by = c('dates', 'area'))

结果是这样的：

       dates     area rating.x rating2 rating.y
1 2020-11-19    paris       10       5       NA
2 2020-11-20   london        5       6       NA
3 2020-11-21  newyork        6       7       NA
4 2020-11-19 budapest       NA      NA        1
5 2020-11-20   moscow       NA      NA        2
6 2020-11-21 valencia       NA      NA        1

即两个数据框中的评级列没有混合在一起，而是创建了两个单独的列。

如何获得这样的结果？

       dates     area rating   rating2 
1 2020-11-19    paris       10       5       
2 2020-11-20   london        5       6       
3 2020-11-21  newyork        6       7       
4 2020-11-19 budapest        1      NA        
5 2020-11-20   moscow        2      NA        
6 2020-11-21 valencia        1      NA

感谢@kybazzi 提供的解决方案，得到了想要的结果

df <- df1 %>%
  bind_rows(df2)

跟进

作为后续问题，我想将以下内容加入到已加入的数据框中：

df3 <- data.frame(dates, area = c('budapest', 'moscow', 'valencia'), 
                  rating2 = c(3, 2, 5))

用同样的方法，结果是这样的：

> df_final <- df %>%
+     bind_rows(df3)
> df_final
       dates     area rating rating2
1 2020-11-19    paris     10       5
2 2020-11-20   london      5       6
3 2020-11-21  newyork      6       7
4 2020-11-19 budapest      1      NA
5 2020-11-20   moscow      2      NA
6 2020-11-21 valencia      1      NA
7 2020-11-19 budapest     NA       3
8 2020-11-20   moscow     NA       2
9 2020-11-21 valencia     NA       5

如何得到这样的结果：

       dates     area rating   rating2 
1 2020-11-19    paris       10       5       
2 2020-11-20   london        5       6       
3 2020-11-21  newyork        6       7       
4 2020-11-19 budapest        1       3        
5 2020-11-20   moscow        2       2        
6 2020-11-21 valencia        1       5

Answer 1

您要查找的是 dplyr::bind_rows()，它将保留公共列并填充 NA 仅存在于其中一个数据框中的列：

> bind_rows(df1, df2)
       dates     area rating rating2
1 2020-11-19    paris     10       5
2 2020-11-20   london      5       6
3 2020-11-21  newyork      6       7
4 2020-11-19 budapest      1      NA
5 2020-11-20   moscow      2      NA
6 2020-11-21 valencia      1      NA

请注意，您也可以继续使用 full_join() - 但如果您不想拆分列，则必须确保将数据框之间的所有公共列作为键包括在内：

> full_join(
+   df1, df2,
+   by = c("dates", "area", "rating")
+ )
       dates     area rating rating2
1 2020-11-19    paris     10       5
2 2020-11-20   london      5       6
3 2020-11-21  newyork      6       7
4 2020-11-19 budapest      1      NA
5 2020-11-20   moscow      2      NA
6 2020-11-21 valencia      1      NA

dplyr 连接的文档提到：

Output columns include all x columns and all y columns. If columns in x and y have the same name (and aren't included in by), suffixes are added to disambiguate.

您也可以通过不指定 by 来避免此问题，在这种情况下 dplyr 将使用所有公共列。

> full_join(df1, df2)
Joining, by = c("dates", "area", "rating")
       dates     area rating rating2
1 2020-11-19    paris     10       5
2 2020-11-20   london      5       6
3 2020-11-21  newyork      6       7
4 2020-11-19 budapest      1      NA
5 2020-11-20   moscow      2      NA
6 2020-11-21 valencia      1      NA

据我所知，这两种方法都适合您的用例。事实上，我相信 full_join() 相对于 bind_rows() 的实际优势正是您要在此处避免的这种行为，即拆分不是键的列。

R dplyr full_join - 没有公共键，需要公共列混合在一起

R dplyr full_join - no common key, need common columns to blend together

r

outer-join

dplyr