xmllint is a helpful tool for anazing xml files. one common usage is its xpath query, which can extract useful information from a xml file.

for example, let this xml file be our input input.xml:

<?xml version="1.0" encoding="UTF-8"?>
<config>
    <point x="10" y="20"/>
    <point x="11" y="21"/>
    <point x="12" y="22"/>
    <point x="13" y="23"/>
    <point x="14" y="24"/>
</config>

we can extract all x coordinates using xpath:

xmllint --xpath "//point/@x" input.xml

output:

 x="10" x="11" x="12" x="13" x="14"

the result is correct, but ugly. this example has only 5 results, but in other use cases the number can be quite large and it won’t be obvious how many results there are. plus, many text-based tools work well with lines. if the result needs further processing, it should be separated by newlines, instead of spaces.

hack

there is a hack to force each output displayed in a new line:

xmllint --shell <<< "cat //point/@x" input.xml

output:

/ >  -------
 x="10"
 -------
 x="11"
 -------
 x="12"
 -------
 x="13"
 -------
 x="14"
/ >

while the result does spread across separate lines, they are separated by ugly markers. these markers would make downstream work more difficult.

solution

so let me give the correct solution. we should separate result nodes by a user-defined separator. this requires a bit research into a 3000-line-ish text. but here is a patch that works:

with this patch, we can write a command like this:

xmllint --xpath "//point/@x" --xpath-node-separator "\n" input.xml

output:

 x="10"
 x="11"
 x="12"
 x="13"
 x="14"

update (20180922)

this patch has been deprecated. for most recent changes, please visit the git repo for the patched program.