how to use str_detect within across when searching multiple columns for several search strings

user_4187056 0
Viewed 296

I'm looking to migrating my functions to the newly minted across.

A function I have searches several key words across several columns using filter_at.

However, I am struggling to replicate this using across as shown below:

library(tidyverse)

raw_df <- tibble::tribble(
  ~cust_name, ~other_desc, ~trans, ~val,
     "Cisco",   "nothing",    "a", 100L,
    "bad_cs",     "cisCo",    "s", 101L,
       "Ibm",   "nothing",    "d", 102L,
    "bad_ib",       "ibM",    "f", 102L,
    "oraCle",    "Oracle",    "g", 103L,
      "mSft",   "nothing",    "k", 103L,
      "noth",      "Msft",    "j", 104L,
      "noth",    "oracle",    "l", 104L
  )


search_string = c("ibm", "cisco")


# Done using `filter_at`
raw_df %>% 
  filter_at(.vars = vars(cust_name, other_desc),
            .vars_predicate = any_vars(str_detect(., regex(paste(search_string, collapse = "|"), ignore_case = TRUE)))
            
  ) %>% unique()
  
  
# Not able to replicate result with `across`
raw_df %>% 
  filter(across(
    .cols = c(cust_name, other_desc), 
    .fns = ~str_detect(.), regex(paste(search_string, collapse = "|"), ignore_case = TRUE)))



raw_df %>% 
  filter(str_detect,
         across(any_of(cust_name, other_desc),
         regex(paste(search_string, collapse = "|"), ignore_case = TRUE)))
across r
2 Answers

Combine across with Reduce to select rows which has any occurrence of the pattern.

library(dplyr)
library(stringr)

pat <- paste(search_string, collapse = "|")

raw_df %>% 
  filter(Reduce(`|`, across(c(cust_name, other_desc), 
        ~str_detect(., regex(pat, ignore_case = TRUE)))))

However, I think using if_any is more suitable here as it was build to handle such cases -

raw_df %>%
  filter(if_any(c(cust_name, other_desc), 
                ~str_detect(., regex(pat, ignore_case = TRUE))))

# cust_name other_desc trans   val
#  <chr>     <chr>      <chr> <int>
#1 Cisco     nothing    a       100
#2 bad_cs    cisCo      s       101
#3 Ibm       nothing    d       102
#4 bad_ib    ibM        f       102
user_39629140

While Ronak's solution is the one to use:

Here is an alternative with an additional trick. I think it is what if_any does: Using rowSums:

rowAny <- function(x) rowSums(x) > 0 


raw_df %>% 
    filter(rowAny(
        across(
            .cols = c(cust_name, other_desc),
            .fns = ~ str_detect(., regex("ibm|cisco", ignore_case = TRUE))
        )))

Output:

  cust_name other_desc trans   val
  <chr>     <chr>      <chr> <int>
1 Cisco     nothing    a       100
2 bad_cs    cisCo      s       101
3 Ibm       nothing    d       102
4 bad_ib    ibM        f       102
user_133216470
Related