how to parse xml as an argument to a shell script

Question

this is the sample xml file I have, I need to get max_size as an output?

I tried using xlimit and xpath , both not working.

<?xml version="1.0" encoding="UTF-8"?>
<weblogic-service>
   <service_info name="dummyapp" version="5.0"
   <object name="WebApplications">
      <item Location="dummyapp/ear/dummyapp.ear" name="dummy"/>
   </object>
   <object name="jdbc_oracle_nonxa">
      <item db_name="MYDB1" db_user="dummyapp_user" global_trans="None" initial_size="10" jndi_name="dummyapp-aty-ds" max_size="25" name="dummyapp-aty-ds" statement_cache_size="10"/>
</object>

Answer 1

Assuming the XML is well formed, such as

<?xml version="1.0" encoding="UTF-8"?>
<weblogic-service>
  <service_info name="dummyapp" version="5.0"/>
  <object name="WebApplications">
    <item Location="dummyapp/ear/dummyapp.ear" name="dummy"/>
  </object>
  <object name="jdbc_oracle_nonxa">
    <item db_name="MYDB1" db_user="dummyapp_user" global_trans="None" initial_size="10" jndi_name="dummyapp-aty-ds" max_size="25" name="dummyapp-aty-ds" statement_cache_size="10"/>
  </object>
</weblogic-service>

(I've closed tags that were not closed in the question)

... then the following will output 25 (for this example):

xml sel -t -v '//item/@max_size' -nl file.xml

This is using XMLStarlet to get the value of the max_value attribute of all item nodes. The -nl at the end inserts a newline after the output.

If you need to be more specific and only look at the item node in the object node whose name is jdbc_oracle_nonxa:

xml sel -t -v '//object[@name="jdbc_oracle_nonxa"]/item/@max_size' -nl file.xml

Note that XMLStarlet is installed as xmlstarlet rather than xml on some systems.

---

Using xmllint (no newline at end of output):

xmllint --xpath 'string(//object[@name="jdbc_oracle_nonxa"]/item/@max_size)' file.xml

Answer 2

Often, you are much faster, but less secure, with sed or grep:

sed -E -n 's/.*max_size="([0-9]+)".*/\1/p' sample-4.xml 
25

or grep:

grep -Eo 'max_size="[0-9]+" ' sample-4.xml | grep -Eo "[0-9]+"
25

It doesn't validate the xml and might pick commented code without recognizing. If you look at the screen, and get two results, you might notice, but if it is used in an automated process, it is error prone.

Sometimes you have written the program, generating the xml, so you know, whether it might output comments or max_size="..." in a different, unwanted context. If you know the limits, there is nothing wrong with sed, grep or awk, except that they prevent you from learning the beautiful xmlstarlet-syntax. :)