Recode data frame based on an index column-CodePudding

I have the following data frame df:

 Index col_a col_b col_c col_d
1     4     c     v     g     j
2     1     x  <NA>     z     s
3  1, 3     k     j     n     y
4     2     q     t     o     i
5  2, 3     y     m     w     r
6     2     d     u     x     a
7  3, 4     n     y     k     g
8     4     h     d  <NA>     u
9     1     s     x     j     m

df <- structure(list(Index = c("4", "1", "1, 3", "2", "2, 3", "2", 
"3, 4", "4", "1"), col_a = c("c", "x", "k", "q", "y", "d", "n", 
"h", "s"), col_b = c("v", NA, "j", "t", "m", "u", "y", "d", "x"
), col_c = c("g", "z", "n", "o", "w", "x", "k", NA, "j"), col_d = c("j", 
"s", "y", "i", "r", "a", "g", "u", "m")), class = "data.frame", row.names = c(NA, 
-9L))

I wish to recode/update the values in the df to "Yes"/"No" based on the Index column, where 1 = col_a, 2 = col_b and so on.

This is my desired output:

  Index col_a col_b col_c col_d
1     4    No    No    No   Yes
2     1   Yes    No    No    No
3  1, 3   Yes    No   Yes    No
4     2    No   Yes    No    No
5  2, 3    No   Yes   Yes    No
6     2    No   Yes    No    No
7  3, 4    No    No   Yes   Yes
8     4    No    No    No   Yes
9     1   Yes    No    No    No

Currently I'm using a lapply approach to solve the problem, but I wonder if there's any simpler solution in just one or two lines of code.

do.call(rbind, lapply(1:nrow(df), \(x) {
  vec <- as.integer(unlist(strsplit(df[x, 1], ",")))   1
  df[x, vec] <- "Yes"
  df[x, -c(1, vec)] <- "No"
  df[x, ]
}))

CodePudding user response：

A base R option is to use row/column indexing should be faster for assignment than with rowwise looping

m1 <- cbind(rep(seq_len(nrow(df)), nchar(gsub("\\D ", "", df$Index))), 
    scan(text = df$Index, what = numeric(), sep=","))
df[-1][m1] <- "Yes"
df[-1][df[-1] != "Yes"|is.na(df[-1])] <- "No"

-output

> df
  Index col_a col_b col_c col_d
1     4    No    No    No   Yes
2     1   Yes    No    No    No
3  1, 3   Yes    No   Yes    No
4     2    No   Yes    No    No
5  2, 3    No   Yes   Yes    No
6     2    No   Yes    No    No
7  3, 4    No    No   Yes   Yes
8     4    No    No    No   Yes
9     1   Yes    No    No    No

Or another option is

i1 <- Reduce(`|`, lapply(read.csv(text = df$Index, 
    header = FALSE),\(x) col(df[-1]) == x))
df[-1] <- c("No", "Yes")[1 (i1 & !is.na(i1))]

-output

> df
  Index col_a col_b col_c col_d
1     4    No    No    No   Yes
2     1   Yes    No    No    No
3  1, 3   Yes    No   Yes    No
4     2    No   Yes    No    No
5  2, 3    No   Yes   Yes    No
6     2    No   Yes    No    No
7  3, 4    No    No   Yes   Yes
8     4    No    No    No   Yes
9     1   Yes    No    No    No

Or with tidyverse

library(dplyr)
library(stringr)
df %>%
  mutate(across(starts_with("col_"),
   ~ case_when(str_detect(Index, as.character( match(cur_column(), 
      names(df))-1)) ~ "Yes", TRUE ~ "No")))

-output

  Index col_a col_b col_c col_d
1     4    No    No    No   Yes
2     1   Yes    No    No    No
3  1, 3   Yes    No   Yes    No
4     2    No   Yes    No    No
5  2, 3    No   Yes   Yes    No
6     2    No   Yes    No    No
7  3, 4    No    No   Yes   Yes
8     4    No    No    No   Yes
9     1   Yes    No    No    No

CodePudding user response：

What you have already seems pretty good. We could make a few minor improvements as shown:

Index <- strsplit(df[[1]], ",")
do.call("rbind", lapply(1:nrow(df), function(i) {
  dfi <- df[i, ]
  dfi[-1] <- "No"
  replace(dfi, as.integer(Index[[i]])   1L, "Yes")
}))