DOM Nodes

xml4h provides node objects and convenience methods that make it easier to work with an in-memory XML document object model (DOM).

This section of the document covers the main features of xml4h nodes. For the full API-level documentation see DOM Nodes API.

Traversing Nodes

xml4h aims to provide a simple and intuitive API for traversing and manipulating the XML DOM. To that end it includes a number of convenience methods for performing common tasks:

  • Get the Document or root Element from any node via the document and root attributes respectively.
  • You can get the name attribute of nodes that have a name, or look up the different name components with prefix to get the namespace prefix (if any) and local_name to get the name portion without the prefix.
  • Nodes that have a value expose it via the value attribute.
  • A node’s parent attribute returns its parent, while the ancestors attribute returns a list containing its parent, grand-parent, great-grand-parent etc.
  • A node’s children attribute returns the child nodes that belong to it, while the siblings attribute returns all other nodes that belong to its parent. You can also get the siblings_before or siblings_after the current node.
  • Look up a node’s namespace URI with namespace_uri or the alias ns_uri.
  • Check what type of Node you have with Boolean attributes like is_element, is_text, is_entity etc.

“Magical” Node Traversal

To make it easy to traverse XML documents with a known structure xml4h performs some minor magic when you look up attributes or keys on Document and Element nodes. If you like, you can take advantage of magical traversal to avoid peppering your code with find and xpath searches, or with child and children node attribute lookups.

The principle is simple:

  • Child elements are available as Python attributes of the parent element class.
  • XML element attributes are available as a Python dict in the owning element.

Here is an example of retrieving information from our Monty Python films document using element names as Python attributes (MontyPythonFilms, Film, Title) and XML attribute names as Python keys (year):

>>> # Parse an example XML document about Monty Python films
>>> import xml4h
>>> doc = xml4h.parse('tests/data/monty_python_films.xml')

>>> for film in doc.MontyPythonFilms.Film:
...     print(film['year'] + ' : ' + film.Title.text)  
1971 : And Now for Something Completely Different
1974 : Monty Python and the Holy Grail

Python class attribute lookups of child elements work very well when your XML document contains only camel-case tag names LikeThisOne or LikeThat. However, if your document contains lower-case tag names there is a chance the element names will clash with existing Python attribute or method names in the xml4h classes.

To work around this potential issue you can add an underscore (_) character at the end of a magical attribute lookup to avoid the naming clash; xml4h will remove that character before looking for a child element. For example, to look up a child of the element elem1 which is named child, the code elem1.child_ will return the child element whereas elem1.child would access the child() Node method instead.


Not all XML child element tag names are accessible using magical traversal. Names with leading underscore characters will not work, and nor will names containing hyphens because they are not valid Python attribute names. If you have to deal with XML names like this use the full API methods like child() and children() instead.

All the gory details about how magical traversal works are documented at NodeAttrAndChildElementLookupsMixin. Depending on how you feel about magical behaviour this feature might feel like a great convenience, or black magic that makes you wary. The right attitude probably lies somewhere in the middle…


The behaviour of namespaced XML elements and attributes is inconsistent. You can do magical traversal of elements regardless of what namespace the elements are in, but to look up XML attributes with a namespace prefix you must include that prefix in the name e.g. prefix:attribute-name.

Searching with Find and XPath

There are two ways to search for elements within an xml4h document: find and xpath.

The find methods provided by the library are easy to use but can only perform relatively simple searches that return Element results, whereas you need to be familiar with XPath query syntax to search effectively with the xpath method but you can perform more complex searches and get results other than just elements.

Find Methods

xml4h provides three different find methods:

  • find() searches descendants of the current node for elements matching the given constraints. You can search by element name, by namespace URI, or with no constraints at all:

    >>> # Find ALL elements in the document
    >>> elems = doc.find()
    >>> [ for e in elems]  
    ['MontyPythonFilms', 'Film', 'Title', 'Description', 'Film', 'Title', 'Description',...
    >>> # Find the seven <Film> elements in the XML document
    >>> film_elems = doc.find('Film')
    >>> [e.Title.text for e in film_elems]  
    ['And Now for Something Completely Different', 'Monty Python and the Holy Grail',...

    Note that the find() method only finds descendants of the node you run it on:

    >>> # Find <Title> elements in a single <Film> element; there's only one
    >>> film_elem = doc.find('Film', first_only=True)
    >>> film_elem.find('Title')
    [<xml4h.nodes.Element: "Title">]
  • find_first() searches descendants of the current node but only returns the first result element, not a list. If there are no matching element results this method returns None:

    >>> # Find the first <Film> element in the document
    >>> doc.find_first('Film')
    <xml4h.nodes.Element: "Film">
    >>> # Search for an element that does not exist
    >>> print(doc.find_first('OopsWrongName'))

    If you were paying attention you may have noticed in the example above that you can make the find() method do exactly same thing as find_first() by passing the keyword argument first_only=True.

  • find_doc() is a convenience method that searches the entire document no matter which node you run it on:

    >>> # Normal find only searches descendants of the current node
    >>> len(film_elem.find('Title'))
    >>> # find_doc searches the entire document
    >>> len(film_elem.find_doc('Title'))

    This method is exactly like calling xml4h_node.document.find(), which is actually what happens behind the scenes.

XPath Querying

xml4h provides a single XPath search method which is available on Document and Element nodes:

xpath() takes an XPath query string and returns the result which may be a list of elements, a list of attributes, a list of values, or a single value. The result depends entirely on the kind of query you perform.


XPath querying is currently only available if you use the lxml or ElementTree implementation libraries. You can check whether the XPath feature is available with has_feature().


Although ElementTree supports XPath queries, this support is very limited and most of the example XPath queries below will not work. If you want to use XPath, you should install lxml for better support.

XPath queries are powerful and complex so we cannot describe them in detail here, but we can at least present some useful examples. Here are queries that perform the same work as the find queries we saw above:

>>> # Query for ALL elements in the document
>>> elems = doc.xpath('//*')  
>>> [ for e in elems]  
['MontyPythonFilms', 'Film', 'Title', 'Description', 'Film', 'Title', 'Description',...

>>> # Query for the seven <Film> elements in the XML document
>>> film_elems = doc.xpath('//Film')
>>> [e.Title.text for e in film_elems]  
['And Now for Something Completely Different', 'Monty Python and the Holy Grail',...

>>> # Query for the first <Film> element in the document (returns list)
>>> doc.xpath('//Film[1]')
[<xml4h.nodes.Element: "Film">]

>>> # Query for <Title> elements in a single <Film> element; there's only one
>>> film_elem = doc.xpath('Film[1]')[0]
>>> film_elem.xpath('Title')
[<xml4h.nodes.Element: "Title">]

You can also do things with XPath queries that you simply cannot with the find method, such as find all the attributes of a certain name or apply rich constraints to the query:

>>> # Query for all year attributes
>>> doc.xpath('//@year')
['1971', '1974', '1979', '1982', '1983', '2009', '2012']

>>> # Query for the title of the film released in 1982
>>> doc.xpath('//Film[@year="1982"]/Title/text()')
['Monty Python Live at the Hollywood Bowl']

Namespaces and XPath

Finally, let’s discuss how you can run XPath queries on documents with namespaces, because unfortunately this is not a simple subject.

First, you need to understand that if you are working with a namespaced document your XPath queries must refer to those namespaces or they will not find anything:

>>> # Parse a namespaced version of the Monty Python Films doc
>>> ns_doc = xml4h.parse('tests/data/monty_python_films.ns.xml')
>>> print(ns_doc.xml())  
<?xml version="1.0" encoding="utf-8"?>
<MontyPythonFilms source="" xmlns="uri:monty-python" xmlns:work="uri:artistic-work">
    <work:Film year="1971">
        <Title>And Now for Something Completely Different</Title>

>>> # XPath queries without prefixes won't find namespaced elements
>>> ns_doc.xpath('//Film')

To refer to namespaced nodes in your query the namespace must have a prefix alias assigned to it. You can specify prefixes when you call the xpath method by providing a namespaces keyword argument with a dictionary of alias-to-URI mappings:

>>> # Specify explicit prefix alias mappings
>>> films = ns_doc.xpath('//x:Film', namespaces={'x': 'uri:artistic-work'})
>>> len(films)

Or, preferably, if your document node already has prefix mappings you can use them directly:

>>> # Our root node already has a 'work' prefix defined...
>>> ns_doc.root['xmlns:work']

>>> # we can use this prefix directly
>>> films = ns_doc.root.xpath('//work:Film')
>>> len(films)

Another gotcha is when a document has a default namespace. The default namespace applies to every descendent node without its own namespace, but XPath doesn’t have a good way of dealing with this since there is no such thing as a “default namespace” prefix alias.

xml4h helps out by providing just such an alias: the underscore (_):

>>> # Our document root has a default namespace
>>> ns_doc.root.ns_uri

>>> # You need a prefix alias that refers to the default namespace
>>> ns_doc.xpath('//Title')

>>> # You could specify it explicitly...
>>> titles = ns_doc.xpath('//x:Title',
...                       namespaces={'x': ns_doc.root.ns_uri})
>>> len(titles)

>>> # ...or use xml4h's special default namespace prefix: _
>>> titles = ns_doc.xpath('//_:Title')
>>> len(titles)

Filtering Node Lists

Many xml4h node attributes return a list of nodes as a NodeList object which confers some special filtering powers. You get this special node list object from attributes like children, ancestors, and siblings, and from the find search method if it has element results.

Here are some examples of how you can easily filter a NodeList to get just the nodes you need:

  • Get the first child node using the filter method:

    >>> # Filter to get just the first child
    >>> doc.root.children.filter(first_only=True)
    <xml4h.nodes.Element: "Film">
    >>> # The document has 7 <Film> element children of the root
    >>> len(doc.root.children)
  • Get the first child node by treating children as a callable:

    >>> doc.root.children(first_only=True)
    <xml4h.nodes.Element: "Film">

    When you treat the node list as a callable it calls the filter method behind the scenes, but since doing it the callable way is quicker and clearer in code we will use that approach from now on.

  • Get the first child node with the child filtering method, which accepts the same constraints as the filter method:

    >>> doc.root.child()
    <xml4h.nodes.Element: "Film">
    >>> # Apply filtering with child
    >>> print(doc.root.child('WrongName'))
  • Get the first of a set of children with the first attribute:

    >>> doc.root.children.first
    <xml4h.nodes.Element: "Film">
  • Filter the node list by name:

    >>> for n in doc.root.children('Film'):
    ...     print(n.Title.text)
    And Now for Something Completely Different
    Monty Python and the Holy Grail
    Monty Python's Life of Brian
    Monty Python Live at the Hollywood Bowl
    Monty Python's The Meaning of Life
    Monty Python: Almost the Truth (The Lawyer's Cut)
    A Liar's Autobiography: Volume IV
    >>> len(doc.root.children('WrongName'))


    Passing a node name as the first argument will match the local name of a node. You can match the full node name, which might include a prefix for example, with a call like: .children(name='SomeName').

  • Filter with a custom function:

    >>> # Filter to films released in the year 1979
    >>> for n in doc.root.children('Film',
    ...         filter_fn=lambda node: node.attributes['year'] == '1979'):
    ...     print(n.Title.text)
    Monty Python's Life of Brian

Manipulating Nodes and Elements

xml4h provides simple methods to manipulate the structure and content of an XML DOM. The methods available depend on the kind of node you are interacting with, and by far the majority are for working with Element nodes.

Delete a Node

Any node can be removes from its owner document with delete():

>>> # Before deleting a Film element there are 7 films
>>> len(doc.MontyPythonFilms.Film)

>>> doc.MontyPythonFilms.children('Film')[-1].delete()
>>> len(doc.MontyPythonFilms.Film)


By default deleting a node also destroys it, but it can optionally be left intact after removal from the document by including the destroy=False option.

Name and Value Attributes

Many nodes have low-level name and value properties that can be read from and written to. Nodes with names and values include Text, CDATA, Comment, ProcessingInstruction, Attribute, and Element nodes.

Here is an example of accessing the low-level name and value properties of a Text node:

>>> text_node = doc.MontyPythonFilms.child('Film').child('Title').child()
>>> text_node.is_text

>>> text_node.value
'And Now for Something Completely Different'

And here is the same for an Attribute node:

>>> # Access the name/value properties of an Attribute node
>>> year_attr = doc.MontyPythonFilms.child('Film').attribute_node('year')
>>> year_attr.is_attribute

>>> year_attr.value

The name attribute of a node is not necessarily a plain string, in the case of nodes within a defined namespaced the name attribute may comprise two components: a prefix that represents the namespace, and a local_name which is the plain name of the node ignoring the namespace. For more information on namespaces see Namespaces.

Import a Node and its Descendants

In addition to manipulating nodes in a single XML document directly, you can also import a node (and all its descendant) from another document using a node clone or transplant operation.

There are two ways to import a node and its descendants:

  • Use the clone_node() Node method or clone() Builder method to copy a node into your document without removing it from its original document.
  • Use the transplant_node() Node method or transplant() Builder method to transplant a node into your document and remove it from its original document.

Here is an example of transplanting a node into a document (which also happens to undo the damage we did to our example DOM in the delete() example above):

>>> # Build a new document containing a Film element
>>> film_builder = ('DeletedFilm')
...     .element('Film').attrs(year='1971')
...         .element('Title')
...             .text('And Now for Something Completely Different').up()
...         .element('Description').text(
...             "A collection of sketches from the first and second TV"
...             " series of Monty Python's Flying Circus purposely"
...             " re-enacted and shot for film.")
...     )

>>> # Transplant the Film element from the new document
>>> node_to_transplant = film_builder.root.child('Film')
>>> doc.MontyPythonFilms.transplant_node(node_to_transplant)
>>> len(doc.MontyPythonFilms.Film)

When you transplant a node from another document it is removed from that document:

>>> # After transplanting the Film node it is no longer in the original doc
>>> len(film_builder.root.find('Film'))

If you need to leave the original document unchanged when importing a node use the clone methods instead.

Working with Elements

Element nodes have the most methods to access and manipulate their content, which is fitting since this is the most useful type of node and you will deal with elements regularly.

The leaf elements in XML documents often have one or more Text node children that contain the element’s data content. While you could iterate over such text nodes as child nodes, xml4h provides the more convenient text accessors you would expect:

>>> title_elem = doc.MontyPythonFilms.Film[0].Title
>>> orig_title = title_elem.text
>>> orig_title
'And Now for Something Completely Different'

>>> title_elem.text = 'A new, and wrong, title'
>>> title_elem.text
'A new, and wrong, title'

>>> # Let's put it back the way it was...
>>> title_elem.text = orig_title

Elements also have attributes that can be manipulated in a number of ways.

Look up an element’s attributes with:

  • the attributes() attribute (or aliases attrib and attrs) that return an ordered dictionary of attribute names and values:

    >>> film_elem = doc.MontyPythonFilms.Film[0]
    >>> film_elem.attributes
    <xml4h.nodes.AttributeDict: [('year', '1971')]>
  • or by obtaining an element’s attributes as Attribute nodes, though that is only likely to be useful in unusual circumstances:

    >>> film_elem.attribute_nodes
    [<xml4h.nodes.Attribute: "year">]
    >>> # Get a specific attribute node by name or namespace URI
    >>> film_elem.attribute_node('year')
    <xml4h.nodes.Attribute: "year">
  • and there’s also the “magical” keyword lookup technique discussed in “Magical” Node Traversal for quickly grabbing attribute values.

Set attribute values with:

  • the set_attributes() method, which allows you to add attributes without replacing existing ones. This method also supports defining XML attributes as a dictionary, list of name/value pairs, or keyword arguments:

    >>> # Set/add attributes as a dictionary
    >>> film_elem.set_attributes({'a1': 'v1'})
    >>> # Set/add attributes as a list of name/value pairs
    >>> film_elem.set_attributes([('a2', 'v2')])
    >>> # Set/add attributes as keyword arguments
    >>> film_elem.set_attributes(a3='v3', a4=4)
    >>> film_elem.attributes
    <xml4h.nodes.AttributeDict: [('a1', 'v1'), ('a2', 'v2'), ('a3', 'v3'), ('a4', '4'), ('year', '1971')]>
  • the setter version of the attributes attribute, which replaces any existing attributes with the new set:

    >>> film_elem.attributes = {'year': '1971', 'note': 'funny'}
    >>> film_elem.attributes
    <xml4h.nodes.AttributeDict: [('note', 'funny'), ('year', '1971')]>

Delete attributes from an element by:

  • using Python’s delete-in-dict technique:

    >>> del(film_elem.attributes['note'])
    >>> film_elem.attributes
    <xml4h.nodes.AttributeDict: [('year', '1971')]>
  • or by calling the delete() method on an Attribute node.

Finally, the Element class provides a number of methods for programmatically adding child nodes, for cases where you would rather work directly with nodes instead of using a Builder.

The most complex of these methods is add_element() which allows you to add a named child element, and to optionally to set the new element’s namespace, text content, and attributes all at the same time. Let’s try an example:

>>> # Add a Film element with an attribute
>>> new_film_elem = doc.MontyPythonFilms.add_element(
...     'Film', attributes={'year': 'never'})

>>> # Add a Description element with text content
>>> desc_elem = new_film_elem.add_element(
...     'Description', text='Just testing...')

>>> # Add a Title element with text *before* the description element
>>> title_elem = desc_elem.add_element(
...     'Title', text='The Film that Never Was', before_this_element=True)

>>> print(doc.MontyPythonFilms.Film[-1].xml())
<Film year="never">
    <Title>The Film that Never Was</Title>
    <Description>Just testing...</Description>

There are similar methods for handling simpler cases like adding text nodes, comments etc. Here is an example of adding text nodes:

>>> # Add a text node
>>> title_elem = doc.MontyPythonFilms.Film[-1].Title
>>> title_elem.add_text(', and Never Will Be')

>>> title_elem.text
'The Film that Never Was, and Never Will Be'

Refer to the Element documentation for more information about the other methods for adding nodes.

Wrapping and Unwrapping xml4h Nodes

You can easily convert to or from xml4h’s wrapped version of an implementation node. For example, if you prefer the lxml library’s ElementMaker document builder approach to the xml4h Builder, you can create a document in lxml

>>> from lxml.builder import ElementMaker
>>> E = ElementMaker()
>>> lxml_doc = E.DocRoot(
...     E.Item(
...         E.Name('Item 1'),
...         E.Value('Value 1')
...     ),
...     E.Item(
...         E.Name('Item 2'),
...         E.Value('Value 2')
...     )
... )
>>> lxml_doc  
<Element DocRoot at ...

…and then convert (or, more accurately, wrap) the lxml nodes with the appropriate adapter to make them xml4h versions:

>>> # Convert lxml Document to xml4h version
>>> xml4h_doc = xml4h.LXMLAdapter.wrap_document(lxml_doc)
>>> xml4h_doc.children
[<xml4h.nodes.Element: "Item">, <xml4h.nodes.Element: "Item">]

>>> # Get an element within the lxml document
>>> lxml_elem = list(lxml_doc)[0]
>>> lxml_elem  
<Element Item at ...

>>> # Convert lxml Element to xml4h version
>>> xml4h_elem = xml4h.LXMLAdapter.wrap_node(lxml_elem, lxml_doc)
>>> xml4h_elem  
<xml4h.nodes.Element: "Item">

You can reach the underlying XML implementation document or node at any time from an xml4h node:

>>> # Get an xml4h node's underlying implementation node
>>> xml4h_elem.impl_node  
<Element Item at ...
>>> xml4h_elem.impl_node == lxml_elem

>>> # Get the underlying implementatation document from any node
>>> xml4h_elem.impl_document  
<Element DocRoot at ...
>>> xml4h_elem.impl_document == lxml_doc