Read a word in XML between elements using sed command

Question

I want to read a word between two xml elements using sed command.

For e.g. in below xml, I want to read the number 1234567.

 <ns1:account>
    <ns2:name>Corporation</ns2:name>
    <address>
        <StrtNm>NewYork</StrtNm>
        <BldgNb>3</BldgNb>
        <PstCd>230300</PstCd>
        <Ctry>USA</Ctry>
    </address>
</ns1:account>
<ns3:details>
<ns4:accnum>
    <ns5:info>
        <nd6:accnum>1234567</nd6:accnum>
    </ns5:info>
</ns4:accnum>
</ns3:details>

I was able to do this using a combination of grep and sed commands as below,

grep -oz '<.*details>\s*<.*accnum>\s*<.*info>\s*<.*accnum>[0-9]*</.*accnum>' test.xml |sed -n 's:.*<.*accnum>\(.*\)</.*accnum>.*:\1:p'

but I read that grep -oz is not good for performance since it treats the entire file as a single line. So I tried with two sed commands but it only works if the file is properly formatted as the one shown above. It doesn't work if the xml comes as a single line without pretty printing. This is what I tried:

sed -n '/.*details>/,/<\/.*accnum>/p' test.xml |sed -n 's:.*<.*accnum>\(.*\)<.*accnum>:\1:p'

Challenges:

1. The file can come with or without namespace prefixes in the elements.
2. The file is pretty large, about 100Mb or more.
3. The file contents can come as a properly formatted xml or as the entire xml as a single line.

I haven't tried awk command yet since there are existing scripts in our application which use the commands listed above, and I was hoping to get the same working.

Answer 1

I've had to edit your XML to make it a well-formed document (adding the <root/> element and declaring the namespaces):

<?xml version="1.0"?>
<root xmlns:ns1="urn:ns1" xmlns:ns2="urn:ns2" xmlns:ns3="urn:ns3" xmlns:ns4="urn:ns4" xmlns:ns5="urn:ns5" xmlns:nd6="urn:nd6">
  <ns1:account>
    <ns2:name>Corporation</ns2:name>
    <address>
      <StrtNm>NewYork</StrtNm>
      <BldgNb>3</BldgNb>
      <PstCd>230300</PstCd>
      <Ctry>USA</Ctry>
    </address>
  </ns1:account>
  <ns3:details>
    <ns4:accnum>
      <ns5:info>
        <nd6:accnum>1234567</nd6:accnum>
      </ns5:info>
    </ns4:accnum>
  </ns3:details>
</root>

Having done that I can use xmlstarlet to parse the XML file and extract precisely the element you need

xmlstarlet sel -t -v '//nd6:accnum' -n x.xml
1234567

You can modify the XPath to be more precise, as necessary. For example /root/ns3:details/ns4:accnum/ns5:info/nd6:accnum would be an extreme option.

If you don't have xmlstarlet available I strongly recommend you install it. If the system is not yours to manage, make it a prerequisite of the project you're on. Trying to parse an XML file with sed and awk will work in the short-term but it's setting up technical debt down the road, particularly if you have little control over the precise layout of the XML document (whitespace, newlines, comments, etc.).

Answer 2

Using xidel and a valid xml input (see @roaima answer), we can:

xidel   -se  '//nd6:accnum/text()'  file.xml

where

• //nd6:accnum/text() is a XPath expression meant to -- look for element "nd6:accnum" anywhere and select its text.

Answer 3

This one-liner perl command will print the expected result:

perl -lne 'print "$1" if /<nd6:accnum>(\w+)</' file.xml
1234567