sed: match between quotes, first instance only

Question

I've read several related posts, but I haven't got my sed string right yet. How can I find the first line that contains myVariable="something", and change what is between the quotation marks to some variable (this is in a shell script)?

I've tried:

sed -i '/myVariable/s/"\([^"]*\)"/"${1}"/' /file

But this inserted $1 into all instances of myVariable="".

If I put "0," before the command like I've seen other posts suggest, then it no longer matches on lines with "myVariable". Instead it replaces the first instance of quotes on every line.

Followup questions: will it make a difference if I want to match a variable with a space in it? What if I want to replace the nth instance of myVariable=""?

Example file.txt:

<blah myVariable="this" />
<blah myVariable="that" />
<blah myVariable="the other thing" />

if I pass in apple to:

sed -i '/myVariable/s/"\([^"]*\)"/"'"${1}"'"/' /file.txt

file.txt becomes:

<blah myVariable="apple" />
<blah myVariable="apple" />
<blah myVariable="apple" />

but I want:

<blah myVariable="apple" />
<blah myVariable="that" />
<blah myVariable="the other thing" />

edit 2:

I tried

sed -i -e '0,/myVariable/s/"\([^"]*\)"/"${1}"/' /file

but that gave me something like:

<foo bar="${1}" />
<blah myVariable="${1}" />
<blah myVariable="${1}" />
<blah myVariable="${1}" />
<alpha bravo="${1}" charlie="notchanged"/>

Note: I don't care about the $1, that can be changed by escaping correctly. I'm more worried that instead of changing the first instance of myVariable, I've now changed the first instance of "" on every single line (regardless of myVariable).

Show sample input and your desired output for that sample input. — Cyrus, May 01 '15 at 13:20
Shell variables don't expand in single quotes. You need to double-quote the variable expansion in the replacement (outside of a single-quoted string): `'/myVariable/s/"$[^"]*$"/"'"${1}"'"/'` — Etan Reisner, May 01 '15 at 13:24
Do you have to use `sed` ? I think this is more doable with `perl` . — tivn, May 01 '15 at 17:44
See http://stackoverflow.com/questions/29613304/is-it-possible-to-escape-regex-metacharacters-reliably-with-sed/29626460#29626460 — Ed Morton, May 01 '15 at 18:41
Thanks I'll read through that link. @tivn, I'd be interested in knowing the perl solution as well. — Ryan, May 01 '15 at 19:23
seems quite strange. do you use gnu version of sed? with mine one this example works fine: ` $ sed '0,/myVariable/s/"$[^"]*$"/"'"${test}"'"/' /tmp/file.txt` — Dieselist, May 01 '15 at 22:55

score 2 · Accepted Answer · answered May 01 '15 at 20:01

2

Just use awk:

$ awk -v new="apple" '!f && sub(/myVariable="[^"]+/,"myVariable=\""new){f=1} 1' file       
<blah myVariable="apple" />
<blah myVariable="that" />
<blah myVariable="the other thing" />

answered May 01 '15 at 20:01

Ed Morton

188,023
17
78
185

score 1 · Answer 2 · answered May 02 '15 at 03:16

This is how we can do that with Perl:

perl -pe 'if (/(myVariable)="[^"]*"/) { s//$1="apple"/ if ++$i==1 }' -i a.txt

This will works whether or not there is whitespace inside the quote pair. For only the n-th regex match, replace 1 in ++$i==1 with n.

score 0 · Answer 3 · answered May 01 '15 at 13:27

0

Break in and out of single quotes

sed -i '/myVariable/s/"\([^"]*\)"/"'"${1}"'"/' /file

answered May 01 '15 at 13:27

RTLinuxSW

822
5
9

Thanks for the comment. I've edited my question to explain a little more. – Ryan May 01 '15 at 13:38

sed: match between quotes, first instance only

3 Answers3