我如何遍历xml输出中的不同项目，然后使用bash / linux shell命令为每个项目打印出不同的值

Question

I have the following XML output (produced by using curl to create a SOAP call to a WSDL):

<?xml version="1.0"?>
<env:Envelope xmlns:env="http://schemas.xmlsoap.org/soap/envelope/">
  <env:Header/>
  <env:Body>
    <ns1:getNodesResponse xmlns:ns1="http://node.sdk.nms.ov.hp.com/">
      <return>
        <item>
          <created>2013-04-22T12:48:06.676Z</created>
          <deviceCategory>com.hp.ov.nms.devices.switchrouter</deviceCategory>
          <deviceDescription>Cisco Nexus C7018 DataCenter Switch</deviceDescription>
          <deviceFamily>com.hp.ov.nms.devices.cisconexus7000seriesswitches</deviceFamily>
          <deviceModel>ciscoNexusC7018</deviceModel>
          <deviceVendor>com.hp.ov.nms.devices.cisco</deviceVendor>
          <discoveryState>DISCOVERY_COMPLETED</discoveryState>
          <systemContact>xxxxxxxxxxxxxxxxxxx</systemContact>
          <systemDescription>xxxxxxxxxxxxxxxxxxxx</systemDescription>
          <systemLocation>xxxxxxxxxxxxxxxxxx</systemLocation>
          <systemName>xxxxxxxxxxxxxxxxxxx</systemName>
          <systemObjectId>.1.3.6.1.4.1.9.12.3.1.3.777</systemObjectId>
          <uuid>c8652440-caf2-490b-8892-cb914a39d19e</uuid>
        </item>
        <item>
          <created>2013-04-22T12:49:36.750Z</created>
          <deviceCategory>com.hp.ov.nms.devices.switchrouter</deviceCategory>
          <deviceDescription>Cisco Nexus C7018 DataCenter Switch</deviceDescription>
          <deviceFamily>com.hp.ov.nms.devices.cisconexus7000seriesswitches</deviceFamily>
          <deviceModel>ciscoNexusC7018</deviceModel>
          <deviceVendor>com.hp.ov.nms.devices.cisco</deviceVendor>
          <discoveryState>DISCOVERY_COMPLETED</discoveryState>
          <systemContact>xxxxxxxxxxxxxxxxx</systemContact>
          <systemDescription>xxxxxxxxxxxxxxxxxx</systemDescription>
          <systemLocation>xxxxxxxxxxxxxx</systemLocation>
          <systemName>xxxxxxxxxxxxxxxxxx</systemName>
          <systemObjectId>.1.3.6.1.4.1.9.12.3.1.3.777</systemObjectId>
          <uuid>6f5ef089-6a51-459f-bde1-9cf18e4f8ca7</uuid>
        </item>
        <item>
          <created>2013-04-22T12:51:56.872Z</created>
          <deviceCategory>com.hp.ov.nms.devices.switchrouter</deviceCategory>
          <deviceDescription>Cisco Nexus C7018 DataCenter Switch</deviceDescription>
          <deviceFamily>com.hp.ov.nms.devices.cisconexus7000seriesswitches</deviceFamily>
          <deviceModel>ciscoNexusC7018</deviceModel>
          <deviceVendor>com.hp.ov.nms.devices.cisco</deviceVendor>
          <discoveryState>DISCOVERY_COMPLETED</discoveryState>
          <systemContact>xxxxxxxxxxxxxxxxxx</systemContact>
          <systemDescription>xxxxxxxxxxxxxxxxxxxxxxxxx</systemDescription>
          <systemLocation>xxxxxxxxxxxxxxxxxxx</systemLocation>
          <systemName>xxxxxxxxxxxxxxxxxxx</systemName>
          <systemObjectId>.1.3.6.1.4.1.9.12.3.1.3.777</systemObjectId>
          <uuid>bae02b8c-25d4-4b53-bef0-2d5b14536e0b</uuid>
        </item>
        </item>
      </return>
    </ns1:getNodesResponse>
  </env:Body>
</env:Envelope>

How could I go about iterating through each <item> and then for each item, print out different values for the item? I was thinking about just grepping for <item> s and then picking through the data between every <item> and </item> , but I was not sure if there was a better way to do this. I would be using bash/linux shell commands

Pseudo code:

for i in item
     print i.uuid,i.systemName

Answer 1

It's better to use a xml parser or xml querying language instead of regex and bash commands. If you are programming in some language see DOM , SAX , StAX etc based xml parsers. You can also use SQL like syntax for xml by using XQuery ; another language to get your data can be xpath .

http://www.w3schools.com/xsl/xpath_intro.asp
http://www.w3schools.com/xsl/xquery_intro.asp

But if you still insist using bash tools.. here is a sed 1-liner:

$ sed -n -e '/<item>/,/<\/item>/p' xml | sed -r -e 's/^\s*<uuid>(.*)<\/uuid>/\1/g' -e 's/^\s*<systemName>(.*)<\/systemName>/\1/g' -e '/^\s*</d' | sed -n 'N;s/\n/,/g;p'
xxxxxxxxxxxxxxxxxxx,c8652440-caf2-490b-8892-cb914a39d19e
xxxxxxxxxxxxxxxxxx,6f5ef089-6a51-459f-bde1-9cf18e4f8ca7
xxxxxxxxxxxxxxxxxxx,bae02b8c-25d4-4b53-bef0-2d5b14536e0b
$ 

Breakdown:

1. sed -n -e '/<item>/,/<\\/item>/p' xml
2. sed -r -e 's/^\\s*<uuid>(.*)<\\/uuid>/\\1/g' -e
's/^\\s*<systemName>(.*)<\\/systemName>/\\1/g' -e '/^\\s*</d'
3. sed -n 'N;s/\\n/,/g;p'

Expression 1: suppress default printing, and p (print) lines with a range. start line should be match regex <item> and end line must match regex </item>. This gives you all items </item>. This gives you all items ...`.

Expression 2: Now we strip off the tags <uuid> , </uuid> , <SystemName> , </SystemName> and just keep the inner portion using regex and s (substitute) command.

Expression 3: Suppressed default printing( -n ); N reads next line from input and concatenates it to previous line(already read by sed into pattern space); therefore concatenating consecutive lines separated by a newline \\n . Then we substitute the \\n char with comman and print the pattern space( p ).