dplyr可以连接多个列还是复合键？

Question

我意识到dplyrv3.0允许你加入不同的变量:

left_join(x, y, by = c("a" = "b")将匹配x.a到y.b

但是,是否可以连接变量组合或者我必须事先添加复合键？

像这样的东西:

left_join(x, y, by = c("a c" = "b d")匹配[ x.a和x.c]到[ y.b和y.d] 的串联

Answer 1

您可以将长度大于1的命名向量传递给以下by参数left_join():

library(dplyr)

d1 <- data_frame(
  x = letters[1:3],
  y = LETTERS[1:3],
  a = rnorm(3)
  )

d2 <- data_frame(
  x2 = letters[3:1],
  y2 = LETTERS[3:1],
  b = rnorm(3)
  )

left_join(d1, d2, by = c("x" = "x2", "y" = "y2"))

Answer 2

截至 2022 年 5 月，我们现在还可以选择使用join_by()，除了允许按特定列进行连接（如 Dave 的回答）之外，它还允许通过多种其他方式连接两个数据帧。

我们可以用：

• 平等条件：==
• 不等式条件：>=、>、<= 或 <
• 滚动助手：closest()
• 重叠助手： Between()、within() 或 Overlaps()

例子：

# first we create a dataset similar to Dave's one, but with a few more columns, which make it easy to demonstrate the other joins
library(tidyverse)
set.seed(0)

dfx <- tibble(
  id = 1:3,
  first_name = c("Alice", "Bob", "Charlie"),
  last_name = c("Adams", "Baker", "Chaplin"),
  a = rnorm(3),
  lb = 0.25,
  ub = 0.75)

dfy <- tibble(
  id = 1:3,
  first_name = c("Alice", "Bob", "Charlie"),
  last_name = c("Adams", "Baker", "Chaplin"),
  b = rnorm(3),
  other_range = 0,
  other_range2 = 1)

平等加入（OP要求的）：

left_join(dfx, dfy, join_by(id, first_name, last_name == last_name))

注意：如果两个数据框中要连接的列的名称相同，则不需要执行 col == col，只需使用 col，如上例中的前两列。

不等式连接：

left_join(dfx, dfy, join_by(a < b)) # join the rows where a < b

滚动连接：

left_join(dfx, dfy, join_by(closest(a < b))) # similar to above, but only take the closest match

重叠连接：

left_join(dfx, dfy, join_by(between(a, other_range, other_range2))) # join rows where a is between other_range and other_range2

left_join(dfx, dfy, join_by(overlaps(lb, ub, other_range, other_range2))) # join rows where the the ranges (lb to ub, and other_range to other_range2) overlap

left_join(dfx, dfy, join_by(within(lb, ub, other_range, other_range2))) # join rows where lb to ub is within other_range to other_range2

另请注意：join_by()假设您将在右列之前列出左列的列。如果您出于某种原因不想这样做，请用于x$左侧数据帧，并y$用于右侧数据帧，例如join_by(x$a < y$b)。

有关更多信息，请阅读文档。

更新：

我意识到我从未真正解决OP问题的核心：

像这样的东西：left_join(x, y, by = c("a c" = "b d")

您不能专门这样做，因为 dplyr 期望每个字符串都是列的名称。但是，如果您有两个字符串，其中包含由空格分隔的列，则可以执行以下操作：

j1 <- "id first_name last_name"
j2 <- j1 # let's pretend for the sake of argument they are different, as it doesn't change the answer

join_vec <- function(j1, j2) {
    setNames(str_split(j2, " ")[[1]], str_split(j1, " ")[[1]])
}

left_join(dfx, dfy, by = join_vec(j1, j2))