R dplyr::c_across() rowSums中的奇怪行为

Question

我正在尝试了解如何将rowSums()仅应用于特定的列。这是一个重述：

df <- tibble(
  "ride" = c("bicycle", "motorcycle", "car", "other"),
  "A" = c(1, NA, 1, NA),
  "B" = c(NA, 2, NA, 2)
)

我可以通过索引2:3得到想要的结果

df %>%
  mutate(total = rowSums(.[2:3], na.rm = TRUE))

# A tibble: 4 × 4
  ride           A     B total
  <chr>      <dbl> <dbl> <dbl>
1 bicycle        1    NA     1
2 motorcycle    NA     2     2
3 car            1    NA     1
4 other         NA     2     2

但是，如果我尝试按名称指定列，则会出现奇怪的结果

df %>%
  mutate(total = sum(c_across(c("A":"B")), na.rm = TRUE))

# A tibble: 4 × 4
  ride           A     B total
  <chr>      <dbl> <dbl> <dbl>
1 bicycle        1    NA     6
2 motorcycle    NA     2     6
3 car            1    NA     6
4 other         NA     2     6

我做错了什么？

我可以通过这样的方式实现我想要的：

df %>%
  mutate_all(~replace(., is.na(.), 0)) %>%
  mutate(total = A + B)

但是我想通过传递一个向量来指定列名，这样将来我就可以更改为不同的列名组合。这就是我想要实现的东西：

cols_to_sum <- c("A","B")

df %>%
  mutate(total = sum(across(cols_to_sum), na.rm = TRUE))

Answer 1

您可以使用select指定要sum的列。

library(dplyr)

cols_to_sum <- c("A","B")

df %>%
  mutate(total = rowSums(select(., all_of(cols_to_sum)), na.rm = TRUE))

#  ride           A     B total
#  <chr>      <dbl> <dbl> <dbl>
#1 bicycle        1    NA     1
#2 motorcycle    NA     2     2
#3 car            1    NA     1
#4 other         NA     2     2

c_across与rowwise协同工作-

df %>%
  rowwise() %>%
  mutate(total = sum(c_across(all_of(cols_to_sum)), na.rm = TRUE)) %>%
  ungroup