replace a unknown string between two known strings with sed

Question

I have a file with the following contents:

WORD1 WORD2 WORD3

How can I use sed to replace the string between WORD1 and WORD3 with foo, such that the contents of the file are changed to the following?:

WORD1 foo WORD3

I tried the following, but obviously I'm missing something because that does not produce the desired results:

sed -i '' 's/WORD1.*WORD3/foo/g' file.txt

Could there be something like `WORD1 foo WORD3 bar WORD1 baz WORD3` in your string (i. e. more than one match per string)? — Tim Pietzcker, May 16 '12 at 07:18
Ah - good question, but no, it's not really necessary to accommodate for more than one match in this scenario. — Camden S., May 16 '12 at 12:49

score 44 · Accepted Answer · edited May 17 '12 at 14:07

44

sed -i 's/WORD1.*WORD3/WORD1 foo WORD3/g' file.txt

or

sed -i 's/(WORD1).*(WORD3)/\1 foo \2/g' file.txt

You might need to escape round brackets, depends on your sed variant.

edited May 17 '12 at 14:07

A.H.

answered May 16 '12 at 07:13

vyegorov

what should I do to allow WORD1, WORD2 and WORD3 be any strings? – Valeriy Van Sep 23 '13 at 13:01
@ValeriyVan, you need to escape delimiting chars in your WORDs (`/` in my example) as well as all other chars, that `sed` treats as controlling ones. – vyegorov Sep 24 '13 at 12:26
The second option results in: `sed: -e expression #1, char 30: invalid reference \2 on 's' command's RHS` on GNU bash, version 4.4.19 – user3751385 Oct 29 '18 at 09:28
2

@user3751385 escape the `(` and `)` – Sulli Feb 02 '19 at 19:42

score 7 · Answer 2 · answered May 16 '12 at 10:02

7

This might work for you:

sed 's/\S\+/foo/2' file

or perhaps:

sed 's/[^[:space:]][^[:space:]]*/foo/2' file

If WORD1 and WORD3 occur more than once:

echo "WORD1 WORD2 WORD3 BLA BLA WORD1 WORD4 WORD3" |
sed 's/WORD3/\n&/g;s/\(WORD1\)[^\n]*\n/\1 foo /g'
WORD1 foo WORD3 BLA BLA WORD1 foo WORD3

answered May 16 '12 at 10:02

potong

`[^[:space:]][^[:space:]]*` can be replaced by `[^[:space:]]\+` (as you have used in your first example) in some versions of `sed`. The `/2` is a nice touch. +1 – Dennis Williamson May 17 '12 at 04:47

score 3 · Answer 3 · edited May 23 '17 at 11:53

3

content of a sample file.txt

$ cat file.txt 
WORD1 WORD2 WORD3
WORD4 WORD5 WORD6
WORD7 WORD8 WORD9

(Correction by @DennisWilliamson in comment)
$ sed -e 's/$[^ ]\+$ \+$[^ ]\+$ \+$.*$/\1 foo \3/' file.txt

WORD1 foo WORD3
WORD4 foo WORD6
WORD7 foo WORD9

while awk is somehow simpler

$ awk -F' ' '{ print $1" foo "$3 }' file.txt

WORD1 foo WORD3
WORD4 foo WORD6
WORD7 foo WORD9

edited May 23 '17 at 11:53

Community

answered May 16 '12 at 08:24

Rony

1

It's not necessary to escape any of the spaces. And `[^\ ]` means "any character other than space or backslash. I would suggest `\+` instead of `*` since, as is, your command could replace an empty line. Also `g` is unnecessary - as is the redirection. – Dennis Williamson May 17 '12 at 04:40
Now it won't work at all. You misunderstood what I meant. `sed -e 's/$[^ ]\+$ \+$[^ ]\+$ \+$.*$/\1 foo \3/' file.txt` – Dennis Williamson May 17 '12 at 11:22

3 Answers3