使用 lambda 函数应用 dplyr::starts_with()

Question

I have below implementation

library(dplyr)
library(tidyr)
dat = data.frame('A' = 1:3, 'C_1' = 1:3, 'C_2' = 1:3, 'M' = 1:3)

Below works

dat %>% rowwise %>% mutate(Anew = list({function(x) c(x[1]^2, x[2] + 5, x[3] + 1)}(c(M, C_1, C_2)))) %>% ungroup %>% unnest_wider(Anew, names_sep = "")

However below does not work when I try find the column names using dplyr::starts_with()

dat %>% rowwise %>% mutate(Anew = list({function(x) c(x[1]^2, x[2] + 5, x[3] + 1)}(c(M, starts_with('C_'))))) %>% ungroup %>% unnest_wider(Anew, names_sep = "")

Any pointer on how to correctly apply starts_with() in this context will be very helpful.

PS: This is continuation from my earlier post Apply custom function that returns multiple values after dplyr::rowwise()

Answer 1

starts_with must be used within a selecting function so we can write this. across is also a selecting function so we could alternately use across(M | starts_with('C_')) in place of select(...) . c_across is also a selecting function but it does not preserve names.

dat %>%
  rowwise %>%
  mutate(Anew = list({function(x) c(x[1]^2, x[2] + 5, x[3] + 1)}
    (select(cur_data(), M, starts_with('C_'))))) %>%
  ungroup %>%
  unnest_wider(Anew, names_sep = "")
## # A tibble: 3 × 7
##       A   C_1   C_2     M AnewM AnewC_1 AnewC_2
##   <int> <int> <int> <int> <dbl>   <dbl>   <dbl>
## 1     1     1     1     1     1       6       2
## 2     2     2     2     2     4       7       3
## 3     3     3     3     3     9       8       4

Here group_modify would also work and allow the use of formula notation to specify an anonymous function. The indexes in the anonymous function have been reordered to correspond to the order in the input.

dat %>%
  group_by(A) %>%
  group_modify(~ cbind(.x, Anew = c(.x[3]^2, .x[1] + 5, .x[2] + 1))) %>%
  ungroup
## # A tibble: 3 × 7
##       A   C_1   C_2     M Anew.M Anew.C_1 Anew.C_2
##   <int> <int> <int> <int>  <dbl>    <dbl>    <dbl>
## 1     1     1     1     1      1        6        2
## 2     2     2     2     2      4        7        3
## 3     3     3     3     3      9        8        4

Answer 2

If we wrap the starts_with in c_across and assuming there is a third column that starts with C_ , then the lambda function on the fly would work

library(dplyr)
library(tidyr)
dat %>%
  rowwise %>%
   mutate(Anew = list((function(x) c(x[1]^2, x[2] + 5, x[3] + 
      1))(c_across(starts_with("C_"))))) %>%
  unnest_wider(Anew, names_sep = "")

-output

# A tibble: 3 × 8
      A   C_1   C_2   C_3     M Anew1 Anew2 Anew3
  <int> <int> <int> <int> <int> <dbl> <dbl> <dbl>
1     1     1     1     1     1     1     6     2
2     2     2     2     2     2     4     7     3
3     3     3     3     3     3     9     8     4

---

Or instead of doing rowwise , we may create a named list of functions and apply column wise with across (would be more efficient)

fns <- list(C_1 = function(x) x^2, C_2 = function(x) x + 5, 
      C_3 = function(x) x + 1)
dat %>%
   mutate(across(starts_with("C_"), 
    ~ fns[[cur_column()]](.x), .names = "Anew{seq_along(.fn)}"))

-output

   A C_1 C_2 C_3 M Anew1 Anew2 Anew3
1 1   1   1   1 1     1     6     2
2 2   2   2   2 2     4     7     3
3 3   3   3   3 3     9     8     4

data

dat <- data.frame('A' = 1:3, 'C_1' = 1:3, 'C_2' = 1:3, C_3 = 1:3, 'M' = 1:3)