How to Use Dplyr::Select Functions within Purrr:Map2 All Within a Pipe

Question

library(tidyverse)

Using the example data (bottom), I'm attempting something like the code below, but I'm stuck on how to use this strategy of selecting columns by name using dplyr::select functions within map2. This particular example uses regex, but I want an approach that could also use "contains", or "ends_with" or "starts_with" from dplyr::select.

I realize I could use indexes, below...which works...

map2(Df[8:12],Df[3:7],~ if_else(.x != 3, recode(.y,`1`=0,`2`=0,`3`=0,`4`=1,`5`=1),88)) %>%
as.data.frame %>%
rename_all(paste0,"_new") %>%
cbind(Df,.)

But since my real dataset has many column names I just want to use "select" functions to name them, or use regex. I've tried variations of the code below, but none seem to work. How can I correct this?

The code below first subsets since I only want the variables I'm working with, then pipes into map2.

Df<-Df%>%select(Code,Pet,matches("^q.*s$"),matches("^q.*i$"))%>%
map2(Df(matches("^q.*i$")), Df(matches("^q.*s$")), ~ if_else(.x != 1,  
recode(.y,`1`=0,`2`=0,`3`=0,`4`=1,`5`=1),88)) %>%
as.data.frame %>%
rename_all(paste0,"_new") %>%
cbind(Df,.)

Example Data:

q25i<-c(2,1,88,2,1,2,2,2)
q26i<-c(2,88,88,88,2,2,2,1)
q27i<-c(2,2,1,1,1,1,1,2)
q28i<-c(88,1,1,2,2,2,2,88)
q29i<-c(1,1,1,2,2,1,88,2)
q25s<-c(3,5,88,4,1,4,4,5)
q26s<-c(4,4,5,5,1,4,4,3)
q27s<-c(3,3,4,1,4,5,5,3)
q28s<-c(4,5,88,1,3,2,2,2)
q29s<-c(88,88,3,4,4,3,3,2)
q25U<-c(2,4,4,4,4,4,5,4)
q26U<-c(5,4,6,5,4,3,6,7)
q27U<-c(4,3,2,3,3,3,2,1)
q28U<-c(4,3,2,3,3,2,3,1)
q29U<-c(4,3,5,5,4,3,3,2)
Code<-c("P1","AB","AB","P1","P1","CD","AAA","CD")
Pet<-c("Dog","Cat","Dog","Fish","Dog","Cat","Rabbit","Fish")

 Df<-data.frame (Code,Pet,q25U,q26U,q27U,q28U,q29U,q25i,q26i,q27i,q28i,q29i,q25s,q26s,q27s,q28s,q29s)

Answer 1

Perhaps this is what you're looking for:

library(dplyr)
library(purrr)

list(select(Df, matches("^q.*i$")), select(Df, matches("^q.*s$"))) %>%
  pmap( ~ if_else(.x != 1, recode(.y,`1`=0,`2`=0,`3`=0,`4`=1,`5`=1),88)) %>%
  as.data.frame %>%
  rename_all(paste0,"_new") %>%
  cbind(Df,.)

Result:

  Code    Pet q25U q26U q27U q28U q29U q25i q26i q27i q28i q29i q25s q26s q27s q28s q29s
1   P1    Dog    2    5    4    4    4    2    2    2   88    1    3    4    3    4   88
2   AB    Cat    4    4    3    3    3    1   88    2    1    1    5    4    3    5   88
3   AB    Dog    4    6    2    2    5   88   88    1    1    1   88    5    4   88    3
4   P1   Fish    4    5    3    3    5    2   88    1    2    2    4    5    1    1    4
5   P1    Dog    4    4    3    3    4    1    2    1    2    2    1    1    4    3    4
6   CD    Cat    4    3    3    2    3    2    2    1    2    1    4    4    5    2    3
7  AAA Rabbit    5    6    2    3    3    2    2    1    2   88    4    4    5    2    3
8   CD   Fish    4    7    1    1    2    2    1    2   88    2    5    3    3    2    2
  q25i_new q26i_new q27i_new q28i_new q29i_new
1        0        1        0        1       88
2       88        1        0       88       88
3       88        1       88       88       88
4        1        1       88        0        1
5       88        0       88        0        1
6        1        1       88        0       88
7        1        1       88        0        0
8        1       88        0        0        0

Notes:

This matches with your result if you use the correct indices:

map2(Df[8:12],Df[13:17],~ if_else(.x != 1, recode(.y,`1`=0,`2`=0,`3`=0,`4`=1,`5`=1),88)) %>%
  as.data.frame %>%
  rename_all(paste0,"_new") %>%
  cbind(Df,.)

The reason for using pmap instead of map2 is because pmap takes a list of inputs whereas map2 takes exactly two inputs. For instance, the following uses map2 instead of pmap:

list(select(Df, matches("^q.*i$")), select(Df, matches("^q.*s$"))) %>%
  {map2(.[[1]], .[[2]], ~ if_else(.x != 1, recode(.y,`1`=0,`2`=0,`3`=0,`4`=1,`5`=1),88))} %>%
  as.data.frame %>%
  rename_all(paste0,"_new") %>%
  cbind(Df,.)

This is less convenient, IMO, since you have to specify the inputs manually and you have to wrap map2 with {} to override the %>% default of piping into the first argument.