Tjyylli

I have a sequence of variables in a dataframe (over 100) and I would like to create an indicator variable for if particular text patterns are present in any of the variables. Below is an example with three variables. One solution I've found is using tidyr::unite() followed by dplyr::mutate(), but I'm interested in a solution where I do not have to unite the variables.

c1<-c("T1", "X1", "T6", "R5")

c2<-c("R4", "C6", "C7", "X3")

c3<-c("C5", "C2", "X4", "T2")



df<-data.frame(c1, c2, c3)



  c1 c2 c3

1 T1 R4 C5

2 X1 C6 C2

3 T6 C7 X4

4 R5 X3 T2



code.vec<-c("T1", "T2", "T3", "T4") #Text patterns of interest

code_regex<-paste(code.vec, collapse="|")



new<-df %>% 

  unite(all_c, c1:c3, remove=FALSE) %>% 

  mutate(indicator=if_else(grepl(code_regex, all_c), 1, 0)) %>% 

  select(-(all_c))



  c1 c2 c3 indicator

1 T1 R4 C5 1

2 X1 C6 C2 0

3 T6 C7 X4 0

4 R5 X3 T2 1

Above is an example that produces the desired result, however I feel as if there should be a way of doing this in tidyverse without having to unite the variables. This is something that SAS handles very easily using an ARRAY statement and a DO loop, and I'm hoping R has a good way of handling this.

The real dataframe has many additional variables besides from the "c" fields to search, so a solution that involves searching every column would require subsetting the dataframe to first only contain the variables I want to search, and then joining the data back with the other variables.

We can use tidyverse

library(tidyverse)

df %>%

    mutate_all(str_detect, pattern = code_regex) %>%

    reduce(`+`) %>% 

    mutate(df, indicator = .)

#  c1 c2 c3 indicator

#1 T1 R4 C5         1

#2 X1 C6 C2         0

#3 T6 C7 X4         0

#4 R5 X3 T2         1

Or using base R

Reduce(`+`, lapply(df, grepl, pattern = code_regex))

#[1] 1 0 0 1

Using base R, we can use sapply and use grepl to find pattern in every column and assign 1 to rows where there is more than 0 matches.

df$indicator <- as.integer(rowSums(sapply(df, grepl, pattern = code_regex)) > 0)



df

#  c1 c2 c3 indicator

#1 T1 R4 C5         1

#2 X1 C6 C2         0

#3 T6 C7 X4         0

#4 R5 X3 T2         1

If there are few other columns and we are interested to apply it only for columns which start with "c" we can use grep to filter them.

cols <- grep("^c", names(df))

as.integer(rowSums(sapply(df[cols], grepl, pattern = code_regex)) > 0)

Using dplyr we can do

library(dplyr)



df$indicator <- as.integer(df %>%

              mutate_at(vars(c1:c3), ~grepl(code_regex, .)) %>%

              rowSums() > 0)

Base R with apply

apply(df[cols], 1, function(x) sum(grepl(code_regex, x)))

# [1] 1 0 0 1