Iterating variables over a pipeline with for loop in R-CodePudding

I have a dataset like the following simplified one:

x_1 <- c(1, NA, 2, 3, NA, 4, 5)
x_2 <- c(2, 1, NA, NA, NA, 4, 6)
y_1 <- c(2, 4, 6, 8, NA, 10, NA)
y_2 <- c(NA, 4, NA, 8, 10, 11, 13)
df <- data.frame(x_1, x_2, y_1, y_2)

  x_1 x_2 y_1 y_2
1   1   2   2  NA
2  NA   1   4   4
3   2  NA   6  NA
4   3  NA   8   8
5  NA  NA  NA  10
6   4   4  10  11
7   5   6  NA  13

The goal is to coalesce each of the two corresponding variables (x and y) and to replace the values that are not the same (e.g. first row of x_1 and x_2) with NA. I did this with the following:

df <- df %>%
  mutate(x = coalesce(x_1, x_2)) %>%
  mutate(x = ifelse(!is.na(x) &
                    !is.na(x_2) &
                    x != x_2,
                    NA,
                    x)) %>%
 select(!c(x_1, x_2))

Now, I have to do this with 21 variables so I thought I put the variables in a list and feed them through the pipeline with a for loop like this:

cols <- c("x", "y")

for(i in cols){
  var_1 <- paste(i, "1", sep = "_")
  var_2 <- paste(i, "2", sep = "_")
  
  df <- df %>%
    mutate(i = coalesce(var_1, var_2)) %>%
    mutate(i = ifelse(!is.na(i) &
                      !is.na(var_2) &
                      i != var_2,
                      NA,
                      i)) %>%
    select(!c(var_1, var_2))
}

What happens is that the code is executed, but instead of the new variables there is only the variable "i" with empty values. It seems as if R does not recognise the "i" in the pipeline as the iterator, however it does recognize "var_1" and "var_2" (because they are being removed from the dataset).

Does anyone know why that is and how I can fix it?

Thanks a lot in advance.

CodePudding user response：

fun <- function(x, var) {

  var_1 <- sym(paste(var, "1", sep = "_"))
  var_2 <- sym(paste(var, "2", sep = "_"))

  x %>%
    mutate(!!var := ifelse((!!var_1 != !!var_2) %in% TRUE,
                           NA, coalesce(!!var_1, !!var_2))) %>%
    select(!c(var_1, var_2))
}

cols <- c("x", "y")

Reduce(fun, cols, init = df)

#    x  y
# 1 NA  2
# 2  1  4
# 3  2  6
# 4  3  8
# 5 NA 10
# 6  4 NA
# 7 NA 13

CodePudding user response：

If you want to avoid rlang:

library(tidyverse)
library(stringr)

x_1 <- c(1, NA, 2, 3, NA, 4, 5)
x_2 <- c(2, 1, NA, NA, NA, 4, 6)
y_1 <- c(2, 4, 6, 8, NA, 10, NA)
y_2 <- c(NA, 4, NA, 8, 10, 11, 13)
df <- data.frame(x_1, x_2, y_1, y_2)

my_coalesce <- function(d) {
  vec_1 <- select(d, 1) %>% pull()
  vec_2 <- select(d, 2) %>% pull()
  res <- coalesce(vec_1, vec_2)
  res[vec_1 != vec_2] <- NA
  res
}

cols <- c("x", "y")

map(cols, ~df %>%
      select(starts_with(.x)) %>% # or:
      #select(str_c(.x, "_", 1:2)) %>% 
      my_coalesce()) %>%
  set_names(cols) %>%
  as_tibble()