I want to get the value of name and put it in a variable using XMLLint
<body>
<value name="abc"></value>
</body>
echo 'cat //body/value/@name' | xmllint --shell "test.xml"
/ > -------
name="abc"
/ >
So I want to assign the value "abc" to variable $test
You need to use fn:string(), which will return the value of its argument as xs:string. In case its argument is an attribute, it will therefore return the attribute's value as xs:string.
test=$(xmllint --xpath "string(//body/value/@name)" test.xml)
Try this, it's not beautiful but it works :)
I just erase lines containing > from stdout , cut the string to get the second part after the = , and delete "
test=$(echo 'cat //body/value/@name' | xmllint --shell "test.xml" | grep -v ">" | cut -f 2 -d "=" | tr -d \");
echo $test
An approach with a helper awk command that supports multiple attributes (a streamlined version of ego's approach):
echo 'cat //*/@name' | xmllint --shell file | awk -F\" 'NR % 2 == 0 { print $2 }'
The awk command:
splits xmllint's output lines into fields by " chars. (-F\")
xmllint normalizes quoting around attribute values to "..." on output, even if the input had '...', so it's sufficient to split by ".only processes even-numbered lines (NR %2 == 0), thereby filtering out the separator lines that cat invariably prints.
print $2 then prints only the 2nd field, which is the value of each attribute without the enclosing "...".
Assuming the following sample XML in file:
<body>
<value name="abc"></value>
<value name="def"></value>
</body>
the above yields:
abc
def
read the output lines into variables (either in a while loop or, in Bash, into an array with read -a); if you need further guidance, please ask a new question.I recently had to port my original simpler solution using --xpath to a platform lacking this feature, so had to adopt the "cat" solution too. This will handle multiple matches, tested on Ubuntu 12.04 and Solaris 11:
getxml() { # $1 = xml file, $2 = xpath expression
echo "cat $2" | xmllint --shell $1 |\
sed -n 's/[^\"]*\"\([^\"]*\)\"[^\"]*/\1/gp'
}
e.g. extracting instance names from a glassfish domain config:
$ getxml /tmp/test.xml "//server[@node-ref]/@name"
inst1
inst2
The sed post-processing just grabs all quoted values which was adequate for my needs (getting bits of glassfish config).
With the goal of matching multiple lines and extract the value of their homonymous attribute and building on previous answers this worked for me and it's shorter hehe
xmllint --xpath "//*/@value" file.xml | awk -F\" '{ print $2 }'
xmllint --xpath "//*/@value" file.xml | cut -f2 -d\"The next works for me:
xmllint --xpath '//o/@loc' file.xml | sed -n 's/[^\"]*\"\([^\"]*\)\"[^\"]*/\1\n/gp'
//o/@loc??? And why would one need this awful sed-command?