sed regular expression query

Question

To change the Tomcat password in the following line

<user username="user1" password="tomcat" roles="tomcat"/>

I'm trying

sed -i s/'username=\"user1\" password=\".*\"'/'username=\"user1\" password=\"NEWPASS\"'/g tomcat-users.xml

but the resulting line will be

<user username="user1" password="NEWPASS"/>

How do I change the regular expression to not cut off the last attribute? I want it to look like

<user username="user1" password="NEWPASS" roles="tomcat"/>

Answer 1

Less is more. Replace the g with a 1

sed -i s/tomcat/NEWPASS/1 tomcat-users.xml

Answer 2

Try the following substitution:

sed -i 's/password="[^"]*"/password="NEWPASS"/' tomcat-users.xml

If you want to do the substitution only in the line corresponding to user1, specify a regexp address:

sed -i '/username="user1"/ s/password="[^"]*"/password="NEWPASS"/' tomcat-users.xml

Explanation:

# for lines that match
# this regexp...
# ↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓↓
'/username="user1"/ s/password="[^"]*"/password="NEWPASS"/'
#                   ↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑↑
#                    ...execute this command.

Answer 3

Presumably, you only want to do this for lines that contain username="user1". So, you could do

sed -i '/username="user1"/{s/password="[^"]*"/password="NEWPASS"/}

The idea is to find any lines that contain username="user1" and run the script only on those. This is achieved by the /foo/{...} syntax. Then, you identify the password by looking for the longest stretch of non-" after password=" and replace the whole thing with the new version.

There is no need for g, that makes sed replace all occurrences in any single line ans is unnecessary here.

Answer 4

You could try the below sed command,

$ sed 's~\(<user username="user1" password="\)[^"]*~\1NEWPASS~g' file
<user username="user1" password="NEWPASS" roles="tomcat"/>