Problems with R's grep function

Question

I thought I understood well the use of grep, it finds the matches and I create a filter with them, to select or delete the rows that contain the matches. But I have in some cases I don't get the expected result. I give examples:

T_fil [grep('\b(vp)\b', T_fil$Int),] 
# ok, leave the row that has vp

T_fil [-grep('\b(vp)\b', T_fil$Int),] 
# ok, remove the ones with vp

T_fil [grep('\b(q)\b', T_fil$Int),] 
# good, don't select any rows because none have q

T_fil [-grep('\b(q)\b', T_fil$Int),] 
# error, it deletes all the rows for me, and I shouldn't delete any, because none have q

T_fil [grep('\b()\b', T_fil$Int),] 
# good, select all, I don't understand why, but being empty I'm interested in not making any changes

T_fil [-grep('\b()\b', T_fil$Int),] 
# error, it deletes all the rows and what interests me is that being empty does not make any changes

Can someone explain this behavior to me and what can I do to make the result correct?

Konrad Rudolph · Answer

The issue is that, if grep finds no hit, it returns an empty vector. A negative empty vector is still an empty vector. You then use that to select rows, and nothing gets selected.

In reality, grep is bad for selecting elements for this reason. A better function that works almost the same is grepl, which returns a logical vector instead, and the result of which can be inverted with !:

T_fil [grepl('\b(q)\b', T_fil$Int),] 
# good, don't select any rows because none have q

T_fil [!grepl('\b(q)\b', T_fil$Int),] 
# good, select all rows

Alternatively, you could also pass invert = TRUE to grep to obtain the same result. That is: do not use -grep(…) to invert the result of a query, it’s unreliable. Instead, either use grep(…, invert = TRUE) or use !grepl(…).

Problems with R's grep function

1 Answers