For a given xml file called configurations.xml I would like to extract the value of each conf element, and store it in a variable for later use.
<configurations>
<conf name="bob"/>
<conf name="alice"/>
<conf name="ted"/>
<conf name="carol"/>
</configurations>
The expected output is:
bob
ailce
ted
carol
I have xpath and xmllint available. A xpath of //conf/@name gets the nodes, but outputs as name="bob", which is what I'm trying to avoid.
xmlstarlet sel -t -m '//configurations/conf' -v '@name' -n a.xml
worked since xmllint does not seem capable. Good intro here.
Tested on: xmlstarlet version 1.5.0, Ubuntu 14.04.
It fails however on large files: ulimit -Sv 500000 (limit it to 500Mb) dies on a 1.2Gb XML, and jams my computer without the memory limit. See also:
I don't know how to achieve what you're trying to achieve with xmllint only.
Since you have xpath installed, you have Perl's XML::XPath too. So a little bit of Perl:
#!/usr/bin/perl
use XML::Path;
my $xp=XML::XPath->new(filename => 'configurations.xml');
my $nodeset=$xp->find('//conf/@name');
foreach my $node ($nodeset->get_nodelist) {
print $node->getNodeValue,"\0";
}
will output what you want, separated with a nil character.
In a one-liner style:
perl -mXML::XPath -e 'foreach $n (XML::XPath->new(filename => "configurations.xml")->find("//conf/\@name")->get_nodelist) { print $n->getNodeValue,"\0"; }'
To retrieve them in, e.g., a Bash array:
#!/bin/bash
names=()
while IFS= read -r -d '' n; do
names+=( "$n" )
done < <(
perl -mXML::XPath -e 'foreach $n (XML::XPath->new(filename => "configurations.xml")->find("//conf/\@name")->get_nodelist) { print $n->getNodeValue,"\0" }'
)
# See what's in your array:
display -p names
Note that at this point you have the option of turning to Perl and drop Bash completely to solve your problem.
I searched everywhere for this seemingly simple answer. It appears that it is not possible for xmllint to print attribute values from multiple nodes. You can use string(//conf/@name), but that will only print a single value even if there are multiple nodes that match.
If you are stuck with xmllint, the only way is to use additional text processing. Here's a generic way that will parse out the attribute value. It assumes the values do not contain = or " characters.
xmllint --xpath //conf/@name |
tr ' ' '\n' | awk -F= '{print $2}' | sed 's/"//g'
The first pipe converts spaces to newlines.
The second pipe prints what's after the =
The last pipe removes all "
You can use awk command to make it done.
[root@myserver tmp]# cat /tmp/test.xml
<configurations>
<conf name="bob"/>
<conf name="alice"/>
<conf name="ted"/>
<conf name="carol"/>
</configurations>
[root@myserver tmp]# awk -F \" '{print $2}' /tmp/test.xml |grep -v '^$'
bob
alice
ted
carol
[root@myserver tmp]#
If you really want use xpath and to display only the attribute values without the "name=" part, then here's what worked for me:
xpath configurations.xml 'string(//conf/@name)' 2>/dev/null
In plain English, wrap your XPath query in string(), and also suppress the verbose ouput of xpath by adding 2>/dev/null at the end.
