Negative lookahead checking for absence of multiple patterns-CodePudding

I am currently trying to index rows that should and shouldn't be kept based on whether the regex-values in the pattern-column exists in the description-column, in the data below.

data <- data.frame(id = c(1,2,2,3,3,4), 
                   old_levels = c(0,1,1,1,1,2),
                   levels = c(1,2,3,2,3,4),
                   description = c("vegetable", "fruit", "fruit",
                                   "meat", "meat", "soda"),
                   pattern = c("vegetable",
                               "fruit", 
                               "?!(vegetable|fruit)", 
                               "fruit",
                               "?!(vegetable|fruit)", 
                               NA))

Using dplyr I figured that the below example should work:

data %>% rowwise() %>% mutate(matches = grepl(pattern, description))

However, this yields:

# A tibble: 6 x 6
# Rowwise: 
     id old_levels levels description pattern             matches
  <dbl>      <dbl>  <dbl> <chr>       <chr>               <lgl>  
1     1          0      1 vegetable   vegetable           TRUE   
2     2          1      2 fruit       fruit               TRUE   
3     2          1      3 fruit       ?!(vegetable|fruit) FALSE  
4     3          1      2 meat        fruit               FALSE  
5     3          1      3 meat        ?!(vegetable|fruit) FALSE  
6     4          2      4 soda        NA                  NA

The NA is expected and is working as intended, however I'm struggling to get the negative lookahead to work, as matches in row 5 should be TRUE...

Any help would be appreciated!

CodePudding user response：

The lookahead syntax is (?!...), not ?!(...).

Besides, grepl with the default TRE library does not support lookarounds, you need to pass perl=TRUE.

You can use

data <- data.frame(id = c(1,2,2,3,3,4), 
                   old_levels = c(0,1,1,1,1,2),
                   levels = c(1,2,3,2,3,4),
                   description = c("vegetable", "fruit", "fruit",
                                   "meat", "meat", "soda"),
                   pattern = c("vegetable",
                               "fruit", 
                               "^(?!.*(?:vegetable|fruit))", 
                               "fruit",
                               "^(?!.*(?:vegetable|fruit))", 
                               NA))

data %>% rowwise() %>% mutate(matches = grepl(pattern, description, perl=TRUE))

Output:

> data %>% rowwise() %>% mutate(matches = grepl(pattern, description, perl=TRUE))
# A tibble: 6 x 6
# Rowwise: 
     id old_levels levels description pattern                    matches
  <dbl>      <dbl>  <dbl> <chr>       <chr>                      <lgl>  
1     1          0      1 vegetable   vegetable                  TRUE   
2     2          1      2 fruit       fruit                      TRUE   
3     2          1      3 fruit       ^(?!.*(?:vegetable|fruit)) FALSE  
4     3          1      2 meat        fruit                      FALSE  
5     3          1      3 meat        ^(?!.*(?:vegetable|fruit)) TRUE   
6     4          2      4 soda        <NA>                       NA