How to execute XPath one-liners from shell?

Question

Is there a package out there, for Ubuntu and/or CentOS, that has a command-line tool that can execute an XPath one-liner like foo //element@attribute filename.xml or foo //element@attribute < filename.xml and return the results line by line?

I'm looking for something that would allow me to just apt-get install foo or yum install foo and then just works out-of-the-box, no wrappers or other adaptation necessary.

Here are some examples of things that come close:

Nokogiri. If I write this wrapper I could call the wrapper in the way described above:

#!/usr/bin/ruby

require 'nokogiri'

Nokogiri::XML(STDIN).xpath(ARGV[0]).each do |row|
  puts row
end

XML::XPath. Would work with this wrapper:

#!/usr/bin/perl

use strict;
use warnings;
use XML::XPath;

my $root = XML::XPath->new(ioref => 'STDIN');
for my $node ($root->find($ARGV[0])->get_nodelist) {
  print($node->getData, "\n");
}

xpath from XML::XPath returns too much noise, -- NODE -- and attribute = "value".

xml_grep from XML::Twig cannot handle expressions that do not return elements, so cannot be used to extract attribute values without further processing.

EDIT:

echo cat //element/@attribute | xmllint --shell filename.xml returns noise similar to xpath.

xmllint --xpath //element/@attribute filename.xml returns attribute = "value".

xmllint --xpath 'string(//element/@attribute)' filename.xml returns what I want, but only for the first match.

For another solution almost satisfying the question, here is an XSLT that can be used to evaluate arbitrary XPath expressions (requires dyn:evaluate support in the XSLT processor):

<?xml version="1.0"?>
<xsl:stylesheet xmlns:xsl="http://www.w3.org/1999/XSL/Transform" version="1.0"
    xmlns:dyn="http://exslt.org/dynamic" extension-element-prefixes="dyn">
  <xsl:output omit-xml-declaration="yes" indent="no" method="text"/>
  <xsl:template match="/">
    <xsl:for-each select="dyn:evaluate($pattern)">
      <xsl:value-of select="dyn:evaluate($value)"/>
      <xsl:value-of select="'&#10;'"/>
    </xsl:for-each> 
  </xsl:template>
</xsl:stylesheet>

Run with xsltproc --stringparam pattern //element/@attribute --stringparam value . arbitrary-xpath.xslt filename.xml.

Answer 1

In my search to query maven pom.xml files I ran accross this question. However I had the following limitations:

• must run cross-platform.
• must exist on all major linux distributions without any additional module installation
• must handle complex xml-files such as maven pom.xml files
• simple syntax

I have tried many of the above without success:

• python lxml.etree is not part of the standard python distribution
• xml.etree is but does not handle complex maven pom.xml files well, have not digged deep enough
• python xml.etree does not handle maven pom.xml files for unknown reason
• xmllint does not work either, core dumps often on ubuntu 12.04 "xmllint: using libxml version 20708"

The solution that I have come across that is stable, short and work on many platforms and that is mature is the rexml lib builtin in ruby:

    ruby -r rexml/document -e 'include REXML; 
        puts XPath.first(Document.new($stdin), "/project/version/text()")' < pom.xml

What inspired me to find this one was the following articles:

• Ruby/XML, XSLT and XPath Tutorial
• IBM: Ruby on Rails and XML

2025 update

• IBM: Ruby on Rails and XML on wayback machine

Answer 2

A minimal wrapper for python's lxml module that will print all matching nodes by name (at any level), e.g. mysubnode or an XPath subset e.g. //intermediarynode/subnode. If the expression evaluates to text then text will be printed, if it evaluates to an element then the entire raw element will be rendered to text. It also attempts to handle XML namespaces in a way that allows using local tag names without prefixing. With extended XPath mode enabled via the -x flag the default namespace needs to be referenced with the p: prefix, e.g. //p:tagname/p:subtag

#!/usr/bin/env python3
import argparse
import os
import sys

from lxml import etree

DEFAULT_NAMESPACE_KEY = 'p'

def print_element(elem):
    if isinstance(elem, str):
        print(elem)
    elif isinstance(elem, bytes):
        print(elem.decode('utf-8'))
    else:
        print(elem.text and elem.text.strip() or etree.tostring(elem, encoding='unicode', pretty_print=True))


if __name__ == '__main__':

    parser = argparse.ArgumentParser(description='XPATH lxml wrapper',
                                     usage="""
    Print all nodes by name in XML file:                                     
    \t{0} myfile.xml somename

    Print all nodes by XPath selector (findall: reduced subset):                                     
    \t{0} myfile.xml //itermediarynode/childnode

    Print attribute values by XPath selector 'p' maps to default namespace (xpath 1.0: extended subset):                                     
    \t{0} myfile.xml //p:itermediarynode/p:childnode/@src -x
                          
     """.format(os.path.basename(sys.argv[0])))
    parser.add_argument('xpath_file',
                        help='XPath file path')
    parser.add_argument('xpath_expression',
                        help='tag name or xpath expression')
    parser.add_argument('--force_xpath', '-x',
                        action='store_true',
                        default=False,
                        help='Use lxml.xpath (rather than findall)'
    )

    args = parser.parse_args(sys.argv[1:])
    xpath_expression = args.xpath_expression

    tree = etree.parse(args.xpath_file)

    ns = tree.getroot().nsmap

    if args.force_xpath:
        if ns.keys() and None in ns:
            ns[DEFAULT_NAMESPACE_KEY] = ns.pop(None)
        for node in tree.xpath(xpath_expression, namespaces=ns):
            print_element(node)

    elif xpath_expression.isalpha():
        for node in tree.xpath(f"//*[local-name() = '{xpath_expression}']"):
            print_element(node)
    else:
        for el in tree.findall(xpath_expression, namespaces=ns):
            print_element(el)

It uses lxml — a fast XML parser written in C which is not included in the standard python library. Install it with pip install lxml. On Linux/OSX might need prefixing with sudo.

Usage:

python3 xmlcat.py file.xml "//mynode"

lxml can also accept an URL as input:

python3 xmlcat.py http://example.com/file.xml "//mynode" 

Extract the url attribute under an enclosure node i.e. <enclosure url="http:...""..>) (-x forces an extended XPath 1.0 subset):

python3 xmlcat.py xmlcat.py file.xml "//enclosure/@url" -x

Xpath in Google Chrome

As an unrelated side note: If by chance you want to run an XPath expression against the markup of a web page then you can do it straight from the Chrome devtools: right-click the page in Chrome > select Inspect, and then in the DevTools console paste your XPath expression as $x("//spam/eggs").

Example: get all authors on this page:

$x("//*[@class='user-details']/a/text()")

Answer 3

For GitHub action with Ubuntu:

To print pom version and updating:

  - name: "test xmlstarlet"
    id: xmlstarlet
    run: |
      sudo apt-get install xmlstarlet
      xmlstarlet el pom.xml
      xmlstarlet sel -t -v '/_:project/_:version' pom.xml
      xmlstarlet edit  -L --update "/_:project/_:version" --value '1.0.0' pom.xml  

_ is the default namespace:

How to declare XPath namespaces in xmlstarlet?

Answer 4

A solution that works even when namespace declarations exist on top:

Most of the commands proposed in the answers do not work out of the box if the xml has a namespace declared on top. Consider this:

input xml:

<elem1 xmlns="urn:x" xmlns:prefix="urn:y">
    <elem2 attr1="false" attr2="value2">
        elem2 value
    </elem2>
    <elem2 attr1="true" attr2="value2.1">
        elem2.1 value
    </elem2>    
    <prefix:elem3>
        elem3 value
    </prefix:elem3>        
</elem1>

Does not work:

xmlstarlet sel -t -v "/elem1" input.xml
# nothing printed
xmllint -xpath "/elem1" input.xml
# XPath set is empty

Solution:

# Requires >=java11 to run like below (but the code requires >=java17 for case syntax to be recognized)

# Prints the whole document
java ExtractXpath.java "/" example-inputs/input.xml

# Prints the contents and self of "elem1"
java ExtractXpath.java "/elem1" input.xml

# Prints the contents and self of "elem2" whose attr2 value is: 'value2'
java ExtractXpath.java "//elem2[@attr2='value2']" input.xml

# Prints the value of the attribute 'attr2': "value2", "value2.1"
java ExtractXpath.java "/elem1/elem2/@attr2" input.xml

# Prints the text inside elem3: "elem3 value"
java ExtractXpath.java "/elem1/elem3/text()" input.xml

# Prints the name of the matched element: "prefix:elem3"
java ExtractXpath.java "name(/elem1/elem3)" input.xml
# Same as above: "prefix:elem3"
java ExtractXpath.java "name(*/elem3)" input.xml

# Prints the count of the matched elements: 2.0
java ExtractXpath.java "count(/elem2)" input.xml


# known issue: while "//elem2" works. "//elem3" does not (it works only with: '*/elem3' )

ExtractXpath.java:

import java.io.File;
import java.io.FileInputStream;
import java.io.StringWriter;
import java.util.ArrayList;
import java.util.Iterator;
import java.util.List;
import java.util.stream.Collectors;

import javax.xml.XMLConstants;
import javax.xml.namespace.NamespaceContext;
import javax.xml.parsers.DocumentBuilder;
import javax.xml.parsers.DocumentBuilderFactory;
import javax.xml.transform.OutputKeys;
import javax.xml.transform.Transformer;
import javax.xml.transform.TransformerConfigurationException;
import javax.xml.transform.TransformerException;
import javax.xml.transform.TransformerFactory;
import javax.xml.transform.dom.DOMSource;
import javax.xml.transform.stream.StreamResult;
import javax.xml.xpath.XPath;
import javax.xml.xpath.XPathConstants;
import javax.xml.xpath.XPathEvaluationResult;
import javax.xml.xpath.XPathExpression;
import javax.xml.xpath.XPathExpressionException;
import javax.xml.xpath.XPathFactory;

import org.w3c.dom.Document;
import org.w3c.dom.Node;
import org.w3c.dom.NodeList;

public class ExtractXpath {

    public static void main(String[] args) throws Exception {
        assertThat(args.length==2, "Wrong number of args");
        String xpath = args[0];
        File file = new File(args[1]);
             
        assertThat(file.isFile(), file.getAbsolutePath()+" is not a file.");
        FileInputStream fileIS = new FileInputStream(file);
        DocumentBuilderFactory builderFactory = DocumentBuilderFactory.newInstance();
        DocumentBuilder builder = builderFactory.newDocumentBuilder();
        Document xmlDocument = builder.parse(fileIS);
        XPath xPath = XPathFactory.newInstance().newXPath();
        String expression = xpath;
        XPathExpression xpathExpression =  xPath.compile(expression);
        
        XPathEvaluationResult xpathEvalResult =  xpathExpression.evaluateExpression(xmlDocument);
        System.out.println(applyXpathExpression(xmlDocument, xpathExpression, xpathEvalResult.type().name()));
    }

    private static String applyXpathExpression(Document xmlDocument, XPathExpression expr, String xpathTypeName) throws TransformerConfigurationException, TransformerException, XPathExpressionException {

        // see: https://www.w3.org/TR/1999/REC-xpath-19991116/#corelib
        List<String> retVal = new ArrayList();
        if(xpathTypeName.equals(XPathConstants.NODESET.getLocalPart())){ //e.g. xpath: /elem1/*
            NodeList nodeList = (NodeList)expr.evaluate(xmlDocument, XPathConstants.NODESET);
            for (int i = 0; i < nodeList.getLength(); i++) {
                retVal.add(convertNodeToString(nodeList.item(i)));
            }
        }else if(xpathTypeName.equals(XPathConstants.STRING.getLocalPart())){ //e.g. xpath: name(/elem1/*)
            retVal.add((String)expr.evaluate(xmlDocument, XPathConstants.STRING));
        }else if(xpathTypeName.equals(XPathConstants.NUMBER.getLocalPart())){ //e.g. xpath: count(/elem1/*)
            retVal.add(((Number)expr.evaluate(xmlDocument, XPathConstants.NUMBER)).toString());
        }else if(xpathTypeName.equals(XPathConstants.BOOLEAN.getLocalPart())){ //e.g. xpath: contains(elem1, 'sth')
            retVal.add(((Boolean)expr.evaluate(xmlDocument, XPathConstants.BOOLEAN)).toString());
        }else if(xpathTypeName.equals(XPathConstants.NODE.getLocalPart())){ //e.g. xpath: fixme: find one
            System.err.println("WARNING found xpathTypeName=NODE");
            retVal.add(convertNodeToString((Node)expr.evaluate(xmlDocument, XPathConstants.NODE)));
        }else{
            throw new RuntimeException("Unexpected xpath type name: "+xpathTypeName+". This should normally not happen");
        }
        return retVal.stream().map(str->"==MATCH_START==\n"+str+"\n==MATCH_END==").collect(Collectors.joining ("\n"));
        
    }
    
    private static String convertNodeToString(Node node) throws TransformerConfigurationException, TransformerException {
            short nType = node.getNodeType();
        switch (nType) {
            case Node.ATTRIBUTE_NODE , Node.TEXT_NODE -> {
                return node.getNodeValue();
            }
            case Node.ELEMENT_NODE, Node.DOCUMENT_NODE -> {
                StringWriter writer = new StringWriter();
                Transformer trans = TransformerFactory.newInstance().newTransformer();
                trans.setOutputProperty(OutputKeys.OMIT_XML_DECLARATION, "yes");
                trans.setOutputProperty(OutputKeys.INDENT, "yes");
                trans.transform(new DOMSource(node), new StreamResult(writer));
                return writer.toString();
            }
            default -> {
                System.err.println("WARNING: FIXME: Node type:"+nType+" could possibly be handled in a better way.");
                return node.getNodeValue();
            }
                
        }
    }

    
    private static void assertThat(boolean b, String msg) {
        if(!b){
            System.err.println(msg+"\n\nUSAGE: program xpath xmlFile");
            System.exit(-1);
        }
    }
}

@SuppressWarnings("unchecked")
class NamespaceResolver implements NamespaceContext {
    //Store the source document to search the namespaces
    private final Document sourceDocument;
    public NamespaceResolver(Document document) {
        sourceDocument = document;
    }

    //The lookup for the namespace uris is delegated to the stored document.
    @Override
    public String getNamespaceURI(String prefix) {
        if (prefix.equals(XMLConstants.DEFAULT_NS_PREFIX)) {
            return sourceDocument.lookupNamespaceURI(null);
        } else {
            return sourceDocument.lookupNamespaceURI(prefix);
        }
    }

    @Override
    public String getPrefix(String namespaceURI) {
        return sourceDocument.lookupPrefix(namespaceURI);
    }

    @SuppressWarnings("rawtypes")
    @Override
    public Iterator getPrefixes(String namespaceURI) {
        return null;
    }
}

and for simplicity:

xpath-extract command:

#!/bin/bash
java ExtractXpath.java "$1" "$2"

Answer 5

I wasn't happy with Python one-liners for HTML XPath queries, so I wrote my own. Assumes that you installed python-lxml package or ran pip install --user lxml:

function htmlxpath() { python -c 'for x in __import__("lxml.html").html.fromstring(__import__("sys").stdin.read()).xpath(__import__("sys").argv[1]): print(x)' $1 }

Once you have it, you can use it like in this example:

> curl -s https://slashdot.org | htmlxpath '//title/text()'
Slashdot: News for nerds, stuff that matters

Answer 6

Here's one xmlstarlet use case to extract data from nested elements elem1, elem2 to one line of text from this type of XML (also showing how to handle namespaces):

<?xml version="1.0" encoding="UTF-8" standalone="yes" ?>
<mydoctype xmlns="http://xml-namespace-uri" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://xml-namespace-uri http://xsd-uri" format="20171221A" date="2018-05-15">

  <elem1 time="0.586" length="10.586">
      <elem2 value="cue-in" type="outro" />
  </elem1>

</mydoctype>

The output will be

0.586 10.586 cue-in outro

In this snippet, -m matches the nested elem2, -v outputs attribute values (with expressions and relative addressing), -o literal text, -n adds a newline:

xml sel -N ns="http://xml-namespace-uri" -t -m '//ns:elem1/ns:elem2' \
 -v ../@time -o " " -v '../@time + ../@length' -o " " -v @value -o " " -v @type -n file.xml

If more attributes are needed from elem1, one can do it like this (also showing the concat() function):

xml sel -N ns="http://xml-namespace-uri" -t -m '//ns:elem1/ns:elem2/..' \
 -v 'concat(@time, " ", @time + @length, " ", ns:elem2/@value, " ", ns:elem2/@type)' -n file.xml

Note the (IMO unnecessary) complication with namespaces (ns, declared with -N), that had me almost giving up on xpath and xmlstarlet, and writing a quick ad-hoc converter.

Answer 7

Similar to Mike's and clacke's answers, here is the python one-liner (using python >= 2.5) to get the build version from a pom.xml file that gets around the fact that pom.xml files don't normally have a dtd or default namespace, so don't appear well-formed to libxml:

python -c "import xml.etree.ElementTree as ET; \
  print(ET.parse(open('pom.xml')).getroot().find('\
  {http://maven.apache.org/POM/4.0.0}version').text)"

Tested on Mac and Linux, and doesn't require any extra packages to be installed.

Answer 8

clacke’s answer is great but I think only works if your source is well-formed XML, not normal HTML.

So to do the same for normal Web content—HTML docs that aren’t necessarily well-formed XML:

echo "<p>foo<div>bar</div><p>baz" | python -c "from sys import stdin; \
from lxml import html; \
print '\n'.join(html.tostring(node) for node in html.parse(stdin).xpath('//p'))"

And to instead use html5lib (to ensure you get the same parsing behavior as Web browsers—because like browser parsers, html5lib conforms to the parsing requirements in the HTML spec).

echo "<p>foo<div>bar</div><p>baz" | python -c "from sys import stdin; \
import html5lib; from lxml import html; \
doc = html5lib.parse(stdin, treebuilder='lxml', namespaceHTMLElements=False); \
print '\n'.join(html.tostring(node) for node in doc.xpath('//p'))"

Answer 9

One package that is very likely to be installed on a system already is python-lxml. If so, this is possible without installing any extra package:

python -c "from lxml.etree import parse; from sys import stdin; print('\n'.join(parse(stdin).xpath('//element/@attribute')))"

Answer 10

You should try these tools :

• xidel (xidel): xpath3
• xmlstarlet (xmlstarlet page) : can edit, select, transform... Not installed by default, xpath1
• xmllint (man xmllint): often installed by default with libxml2-utils, xpath1 (check my wrapper to have --xpath switch on very old releases and newlines delimited output (v < 2.9.9)). Can be used as interactive shell with the --shell switch.
• xpath : installed via perl's module XML::Xpath, xpath1
• xml_grep : installed via perl's module XML::Twig, xpath1 (limited xpath usage)
• saxon-lint (saxon-lint): my own project, wrapper over @Michael Kay's Saxon-HE Java library, xpath3: using SaxonHE 9.6 ,XPath 3.x (+retro compatibility)

Examples:

xmllint --xpath '//element/@attribute' file.xml
xmlstarlet sel -t -v "//element/@attribute" file.xml
xpath -q -e '//element/@attribute' file.xml
xidel -se '//element/@attribute' file.xml
saxon-lint --xpath '//element/@attribute' file.xml

Answer 11

Sorry to be yet another voice in the fray. I tried all the tools in this thread and found none of them to be satisfactory for my needs, so I wrote my own. You can find it here: https://github.com/charmparticle/xpe

It's been uploaded to pypi, so you can easily install it with pip3 like so:

sudo pip3 install xpe

Once installed, you can use it to run xpath expressions against various kinds of input with the same level of flexibility you would get from using xpaths in selenium or javascript. Yeah, you can use xpaths against HTML with this.

Answer 12

Install the BaseX database, then use it's "standalone command-line mode" like this:

basex -i - //element@attribute < filename.xml

or

basex -i filename.xml //element@attribute

The query language is actually XQuery (3.0), not XPath, but since XQuery is a superset of XPath, you can use XPath queries without ever noticing.

Answer 13

Saxon will do this not only for XPath 2.0, but also for XQuery 1.0 and (in the commercial version) 3.0. It doesn't come as a Linux package, but as a jar file. Syntax (which you can easily wrap in a simple script) is

java net.sf.saxon.Query -s:source.xml -qs://element/attribute

2020 UPDATE

Saxon 10.0 includes the Gizmo tool, which can be used interactively or in batch from the command line. For example

java net.sf.saxon.Gizmo -s:source.xml
/>show //element/@attribute
/>quit

Answer 14

My Python script xgrep.py does exactly this. In order to search for all attributes attribute of elements element in files filename.xml ..., you would run it as follows:

xgrep.py "//element/@attribute" filename.xml ...

There are various switches for controlling the output, such as -c for counting matches, -i for indenting the matching parts, and -l for outputting filenames only.

The script is not available as a Debian or Ubuntu package, but all of its dependencies are.

Answer 15

Since this project is apparently fairly new, check out https://github.com/jeffbr13/xq , seems to be a wrapper around lxml, but that is all you really need (and posted ad hoc solutions using lxml in other answers as well)

Answer 16

You can also try my Xidel. It is not in a package in the repository, but you can just download it from the webpage (it has no dependencies).

It has simple syntax for this task:

xidel filename.xml -e '//element/@attribute' 

And it is one of the rare of these tools that supports XPath 2.

Answer 17

It bears mentioning that nokogiri itself ships with a command line tool, which should be installed with gem install nokogiri.

You might find this blog post useful.

Answer 18

You might also be interested in xsh. It features an interactive mode where you can do whatever you like with the document:

open 1.xml ;
ls //element/@id ;
for //p[@class="first"] echo text() ;

Answer 19

In addition to XML::XSH and XML::XSH2 there are some grep-like utilities suck as App::xml_grep2 and XML::Twig (which includes xml_grep rather than xml_grep2). These can be quite useful when working on a large or numerous XML files for quick oneliners or Makefile targets. XML::Twig is especially nice to work with for a perl scripting approach when you want to a a bit more processing than your $SHELL and xmllint xstlproc offer.

The numbering scheme in the application names indicates that the "2" versions are newer/later version of essentially the same tool which may require later versions of other modules (or of perl itself).