Is there a way to nest sed substitutions?

Question

I want to change the format of every line in a file, and also run another substitution on one of the matched groups.

As an example, I could have a simple three-column CSV file (without escapes) that I wanted to put in another format.

sed -r 's/^([a-z]+),([a-z]+),([a-z]+)$/\1: \3 (\2)'
# would turn "comma,separated,values"
# into       "comma: values (separated)"

However, I also want to remove all the vowels from the second column, i.e. run s/[aeiou]//g on \2.

Is there a nice solution to this problem of nesting substitutions? Please assume that both substitutions are complicated, and that this was just an example.

My sed is not GNU sed version 4.0.

Answer 1

I think you're better off with awk for this one, to make it easier to manipulate the fields.

echo "gaviidae,gruidae,picidae" | awk -F "," '{gsub(/[aeiou]/, "", $2); printf("%s: %s (%s)\n", $1, $3, $2)}'

outputs:

gaviidae: picidae (grd)

Answer 2

This is cryptic (like almost any non-trivial sed) but it'll do the job. It takes advantage of the fact that you want the modified word to go at the end-- the same trick will work if you want it to go to some other position, but the command will be a little longer.

sed 'h;s/.*,([a-z]+),.*/(\1)/;s/[aeiou]//g;x;s/([a-z]+),[a-z]+,([a-z]+)/\1: \2 /;G;s/\n//'

Or in English: "save a copy in the hold buffer, kill all but the second word, strip out the vowels, swap the buffers, rearrange the words (dropping the middle one), add the other buffer onto the end, remove the new-line".