Using grep in R to delete rows from a data.frame

I have a dataframe such as this one:

    d <- data.frame(cbind(x=1, y=1:10,    z=c("apple","pear","banana","A","B","C","D","E","F","G")), stringsAsFactors = FALSE)

I'd like to delete some rows from this dataframe, depending on the content of column z:

    new_d <- d[-grep("D",d$z),]

This works fine; row 7 is now deleted:

    new_d
     x  y      z
  1  1  1  apple
  2  1  2   pear
  3  1  3 banana
  4  1  4      A
  5  1  5      B
  6  1  6      C
  8  1  8      E
  9  1  9      F
  10 1 10      G

However, when I use grep to search for content that is not present in column z, it seems to delete all content of the dataframe:

    new_d <- d[-grep("K",d$z),]
    new_d
    [1] x y z
    <0 rows> (or 0-length row.names)

I would like to search and delete rows in this or another way, even if the character string I am searching for is not present. How to go about this?

Answers

You can use TRUE/FALSE subsetting instead of numeric.

grepl is like grep, but it returns a logical vector. Negation works with it.

 d[!grepl("K",d$z),]
   x  y      z
1  1  1  apple
2  1  2   pear
3  1  3 banana
4  1  4      A
5  1  5      B
6  1  6      C
7  1  7      D
8  1  8      E
9  1  9      F
10 1 10      G

Here's your problem:

> grep("K",c("apple","pear","banana","A","B","C","D","E","F","G"))
integer(0)

Try grepl() instead:

d[!grepl("K",d$z),]

This works because the negated logical vector has an entry for every row:

> grepl("K",d$z)
 [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE
> !grepl("K",d$z)
 [1] TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE

For completeness, since R 3.3.0, grep and friends come with an invert argument:

new_d <- d[grep("K", d$z, invert = TRUE)]

You want to use grepl in this case, e.g., new_d <- d[! grepl("K",d$z),].

Need Your Help