Using awk/sed to parse/replace XML inline

Question

I'm trying to modify multiple version numbers from a pom.xml file in a single shell script, where I'm unsure of what the version number of the component may be. I can parse down to version, however, cannot replace that version inline.

The pom.xml looks something like this:

...
        <dependency>
            <groupId>abc.efg</groupId>
            <artifactId>service1</artifactId>
            <version>0.0.10</version>
        </dependency>
        <dependency>
            <groupId>abc.efg</groupId>
            <artifactId>service2</artifactId>
            <version>0.1.5</version>
        </dependency>
        <dependency>
            <groupId>abc.efg</groupId>
            <artifactId>service3</artifactId>
            <version>0.0.1</version>
        </dependency>
...

I'm completely new to shell scripting so forgive me if this is a little sloppy, but this far I can pull the service version stringing grep like;

$ grep -A 1 "service1" pom.xml | grep -Eo '[.0-9]*'
$ 0.0.10

but using awk to replace inline is proving challenging

$ grep -A 1 "service" pom.xml | grep -Eo '[.0-9]*' | sed 's/[,0-9]/3.2.1/g'
$ 3.2.1.3.2.1.3.2.1

Output I'm looking for is:

0.0.10 -> 3.2.1

$ grep -A 1 "service" pom.xml | grep -Eo '[.0-9]*' | #sedsyntax
$ 3.2.1

Any advice or pointers would be appreciated at this point

Use an XML/HTML parser (xmllint, xmlstarlet ...). – Cyrus Apr 21 '17 at 21:23 — Cyrus, Apr 21 '17 at 21:23

George Vasiliou · Accepted Answer · 2017-04-21T22:26:51.980

You can do it all at once with sed. Search for the pattern, go to next line , replace the version number.

$ sed -r '/service1/!b;n;s/([0-9][.]?)+/1.1.11/' file
<xml>
  <dependency>
    <groupId>abc.efg</groupId>
    <artifactId>service1</artifactId>
    <version>1.1.11</version>
  </dependency>
  <dependency>
    <groupId>abc.efg</groupId>
    <artifactId>service2</artifactId>
    <version>0.1.5</version>
  </dependency>
  <dependency>
    <groupId>abc.efg</groupId>
    <artifactId>service3</artifactId>
    <version>0.0.1</version>
  </dependency>
</xml>

Explanation:

sed -r '/service1/!b;n;s/([0-9][.]?)+/1.1.11/' file
        |________||___||_____________________|      
            |       |               |
Pattern-----|       |               |
Go next line--------|               |
Replace Operation-------------------|
-r switch (or -E) = Extended Regexp Support

George this is spot on, thank you very much for the detailed explanation and taking the time out to help. — Gerard_14, Apr 21 '17 at 22:50
Better match `>service1<` in view of the `service12` (not shown in the example pom). — Walter A, May 21 '17 at 08:16

score 0 · Answer 2 · answered Apr 21 '17 at 21:31

0

The immediate problem in your sed expression is not capturing all of the characters in one go.

$ echo "0.0.10" | sed 's/[.0-9]*/3.2.1/g'
3.2.1

I highly recommend doing this operation with an XML tool like xmlstarlet.

answered Apr 21 '17 at 21:31

Andrew Domaszek

651
6
18

Using awk/sed to parse/replace XML inline

2 Answers2