Sed replace words in quotes

Question

I have file which contains lines like below. Wanted to replace the words in quotes after -s and -e. Tried with sed but didn't work.

File Contents

abc -s "today" -e "today" def
abc -s "yesterday" -e "yesterday" def
abc -s "3 days ago" -e "3 days ago" def

Tried Command

sed 's|-s "*" -e "*"|-s "1" -e "2"|g' <filename>

Required output

abc -s "1" -e "2" def
abc -s "1" -e "2" def
abc -s "1" -e "2" def

Answer 1

The main issue with your own attempt is that "*" means "match zero or more double quotes, followed by yet another double quote". This is because * is a regular expression modifier that means "zero or more of the previous expression". This means that "*" would match strings like ", "", """ etc.

Apart from that, you needlessly use a different delimiter for the s/// command, and you use /g at the end even though you do not expect to do more than one replacement per line. But these are just issues relating to style.

---

There are several ways to do this with sed, but all of them must match a string between double quotes in one way or another. A string between double quotes can be match by the regular expression "[^"]*", i.e. "match a double quote, and then zero or more characters that are not double quotes, followed by another double quote character".

$ sed -e 's/-s "[^"]*"/-s "1"/' -e 's/-e "[^"]*"/-e "2"/' file
abc -s "1" -e "2" def
abc -s "1" -e "2" def
abc -s "1" -e "2" def

The above uses two separate expressions to match and replace the strings after the given option, which also means that

• it doesn't matter in which order the options are given in the input (-s "whatever" will always be replaced by -s "1" and -e "whatever" will always be replaced by -e "2"),
• it makes it indifferent to the number of spaces between the options (change e.g. -s "[^"]*" to -s *"[^"]*" to also allow multiple spaces between the option and the option argument),
• it doesn't matter whether or not there may be other options present (nor how these are ordered in relation to the -s and -e options), and
• if only one of -s and -e is present, but not the other, the option and the option-argument present would be replaced.

If you additionally need to do the replacements only on lines that start with abc:

sed -e '/^abc/{' -e 's/-s "[^"]*"/-s "1"/' -e 's/-e "[^"]*"/-e "2"/' -e '}' file

or without all those -e options in the sed command:

sed '/^abc/{ s/-s "[^"]*"/-s "1"/; s/-e "[^"]*"/-e "2"/; }' file

Change /^abc/ to /^abc.*def$/ if you additionally need to make sure that the lines you modify ends with def.

Answer 2

Sed uses regular expressions, not wildcard patterns. * means "repeat the preceding element zero or more times", and you need to repeat . which means "anything":

sed 's/-s ".*" -e ".*"/-s "1" -e "2"/'