Elements are lists

To make the access to these subelements easy and straight forward, elements mimic the behaviour of normal Python lists as closely as possible:

>>> child = root[0]
>>> print(child.tag)
child1

>>> print(len(root))
3

>>> root.index(root[1]) # lxml.etree only!
1

>>> children = list(root)

>>> for child in root:
...     print(child.tag)
child1
child2
child3

>>> root.insert(0, etree.Element("child0"))
>>> start = root[:1]
>>> end   = root[-1:]

>>> print(start[0].tag)
child0
>>> print(end[0].tag)
child3

Prior to ElementTree 1.3 and lxml 2.0, you could also check the truth value of an Element to see if it has children, i.e. if the list of children is empty:

if root:   # this no longer works!
    print("The root element has children")

This is no longer supported as people tend to expect that a "something" evaluates to True and expect Elements to be "something", may they have children or not. So, many users find it surprising that any Element would evaluate to False in an if-statement like the above. Instead, use len(element), which is both more explicit and less error prone.

>>> print(etree.iselement(root))  # test if it's some kind of Element
True
>>> if len(root):                 # test if it has children
...     print("The root element has children")
The root element has children

There is another important case where the behaviour of Elements in lxml (in 2.0 and later) deviates from that of lists and from that of the original ElementTree (prior to version 1.3 or Python 2.7/3.2):

>>> for child in root:
...     print(child.tag)
child0
child1
child2
child3
>>> root[0] = root[-1]  # this moves the element in lxml.etree!
>>> for child in root:
...     print(child.tag)
child3
child1
child2

In this example, the last element is moved to a different position, instead of being copied, i.e. it is automatically removed from its previous position when it is put in a different place. In lists, objects can appear in multiple positions at the same time, and the above assignment would just copy the item reference into the first position, so that both contain the exact same item:

>>> l = [0, 1, 2, 3]
>>> l[0] = l[-1]
>>> l
[3, 1, 2, 3]

Note that in the original ElementTree, a single Element object can sit in any number of places in any number of trees, which allows for the same copy operation as with lists. The obvious drawback is that modifications to such an Element will apply to all places where it appears in a tree, which may or may not be intended.

The upside of this difference is that an Element in lxml.etree always has exactly one parent, which can be queried through the getparent() method. This is not supported in the original ElementTree.

>>> root is root[0].getparent()  # lxml.etree only!
True

If you want to copy an element to a different position in lxml.etree, consider creating an independent deep copy using the copy module from Python's standard library:

>>> from copy import deepcopy

>>> element = etree.Element("neu")
>>> element.append( deepcopy(root[1]) )

>>> print(element[0].tag)
child1
>>> print([ c.tag for c in root ])
['child3', 'child1', 'child2']

The siblings (or neighbours) of an element are accessed as next and previous elements:

>>> root[0] is root[1].getprevious() # lxml.etree only!
True
>>> root[1] is root[0].getnext() # lxml.etree only!
True

Elements carry attributes as a dict

XML elements support attributes. You can create them directly in the Element factory:

>>> root = etree.Element("root", interesting="totally")
>>> etree.tostring(root)
b'<root interesting="totally"/>'

Attributes are just unordered name-value pairs, so a very convenient way of dealing with them is through the dictionary-like interface of Elements:

>>> print(root.get("interesting"))
totally

>>> print(root.get("hello"))
None
>>> root.set("hello", "Huhu")
>>> print(root.get("hello"))
Huhu

>>> etree.tostring(root)
b'<root interesting="totally" hello="Huhu"/>'

>>> sorted(root.keys())
['hello', 'interesting']

>>> for name, value in sorted(root.items()):
...     print('%s = %r' % (name, value))
hello = 'Huhu'
interesting = 'totally'

For the cases where you want to do item lookup or have other reasons for getting a 'real' dictionary-like object, e.g. for passing it around, you can use the attrib property:

>>> attributes = root.attrib

>>> print(attributes["interesting"])
totally
>>> print(attributes.get("no-such-attribute"))
None

>>> attributes["hello"] = "Guten Tag"
>>> print(attributes["hello"])
Guten Tag
>>> print(root.get("hello"))
Guten Tag

Note that attrib is a dict-like object backed by the Element itself. This means that any changes to the Element are reflected in attrib and vice versa. It also means that the XML tree stays alive in memory as long as the attrib of one of its Elements is in use. To get an independent snapshot of the attributes that does not depend on the XML tree, copy it into a dict:

>>> d = dict(root.attrib)
>>> sorted(d.items())
[('hello', 'Guten Tag'), ('interesting', 'totally')]

Elements contain text

Elements can contain text:

>>> root = etree.Element("root")
>>> root.text = "TEXT"

>>> print(root.text)
TEXT

>>> etree.tostring(root)
b'<root>TEXT</root>'

In many XML documents (data-centric documents), this is the only place where text can be found. It is encapsulated by a leaf tag at the very bottom of the tree hierarchy.

However, if XML is used for tagged text documents such as (X)HTML, text can also appear between different elements, right in the middle of the tree:

<html><body>Hello<br/>World</body></html>

Here, the <br/> tag is surrounded by text. This is often referred to as document-style or mixed-content XML. Elements support this through their tail property. It contains the text that directly follows the element, up to the next element in the XML tree:

>>> html = etree.Element("html")
>>> body = etree.SubElement(html, "body")
>>> body.text = "TEXT"

>>> etree.tostring(html)
b'<html><body>TEXT</body></html>'

>>> br = etree.SubElement(body, "br")
>>> etree.tostring(html)
b'<html><body>TEXT<br/></body></html>'

>>> br.tail = "TAIL"
>>> etree.tostring(html)
b'<html><body>TEXT<br/>TAIL</body></html>'

The two properties .text and .tail are enough to represent any text content in an XML document. This way, the ElementTree API does not require any special text nodes in addition to the Element class, that tend to get in the way fairly often (as you might know from classic DOM APIs).

However, there are cases where the tail text also gets in the way. For example, when you serialise an Element from within the tree, you do not always want its tail text in the result (although you would still want the tail text of its children). For this purpose, the tostring() function accepts the keyword argument with_tail:

>>> etree.tostring(br)
b'<br/>TAIL'
>>> etree.tostring(br, with_tail=False) # lxml.etree only!
b'<br/>'

If you want to read only the text, i.e. without any intermediate tags, you have to recursively concatenate all text and tail attributes in the correct order. Again, the tostring() function comes to the rescue, this time using the method keyword:

>>> etree.tostring(html, method="text")
b'TEXTTAIL'

Using XPath to find text

Another way to extract the text content of a tree is XPath, which also allows you to extract the separate text chunks into a list:

>>> print(html.xpath("string()")) # lxml.etree only!
TEXTTAIL
>>> print(html.xpath("//text()")) # lxml.etree only!
['TEXT', 'TAIL']

If you want to use this more often, you can wrap it in a function:

>>> build_text_list = etree.XPath("//text()") # lxml.etree only!
>>> print(build_text_list(html))
['TEXT', 'TAIL']

Note that a string result returned by XPath is a special 'smart' object that knows about its origins. You can ask it where it came from through its getparent() method, just as you would with Elements:

>>> texts = build_text_list(html)
>>> print(texts[0])
TEXT
>>> parent = texts[0].getparent()
>>> print(parent.tag)
body

>>> print(texts[1])
TAIL
>>> print(texts[1].getparent().tag)
br

You can also find out if it's normal text content or tail text:

>>> print(texts[0].is_text)
True
>>> print(texts[1].is_text)
False
>>> print(texts[1].is_tail)
True

While this works for the results of the text() function, lxml will not tell you the origin of a string value that was constructed by the XPath functions string() or concat():

>>> stringify = etree.XPath("string()")
>>> print(stringify(html))
TEXTTAIL
>>> print(stringify(html).getparent())
None

Tree iteration

For problems like the above, where you want to recursively traverse the tree and do something with its elements, tree iteration is a very convenient solution. Elements provide a tree iterator for this purpose. It yields elements in document order, i.e. in the order their tags would appear if you serialised the tree to XML:

>>> root = etree.Element("root")
>>> etree.SubElement(root, "child").text = "Child 1"
>>> etree.SubElement(root, "child").text = "Child 2"
>>> etree.SubElement(root, "another").text = "Child 3"

>>> print(etree.tostring(root, pretty_print=True))
<root>
  <child>Child 1</child>
  <child>Child 2</child>
  <another>Child 3</another>
</root>

>>> for element in root.iter():
...     print("%s - %s" % (element.tag, element.text))
root - None
child - Child 1
child - Child 2
another - Child 3

If you know you are only interested in a single tag, you can pass its name to iter() to have it filter for you. Starting with lxml 3.0, you can also pass more than one tag to intercept on multiple tags during iteration.

>>> for element in root.iter("child"):
...     print("%s - %s" % (element.tag, element.text))
child - Child 1
child - Child 2

>>> for element in root.iter("another", "child"):
...     print("%s - %s" % (element.tag, element.text))
child - Child 1
child - Child 2
another - Child 3

By default, iteration yields all nodes in the tree, including ProcessingInstructions, Comments and Entity instances. If you want to make sure only Element objects are returned, you can pass the Element factory as tag parameter:

>>> root.append(etree.Entity("#234"))
>>> root.append(etree.Comment("some comment"))

>>> for element in root.iter():
...     if isinstance(element.tag, basestring):
...         print("%s - %s" % (element.tag, element.text))
...     else:
...         print("SPECIAL: %s - %s" % (element, element.text))
root - None
child - Child 1
child - Child 2
another - Child 3
SPECIAL: ê - ê
SPECIAL: <!--some comment--> - some comment

>>> for element in root.iter(tag=etree.Element):
...     print("%s - %s" % (element.tag, element.text))
root - None
child - Child 1
child - Child 2
another - Child 3

>>> for element in root.iter(tag=etree.Entity):
...     print(element.text)
&#234;

Note that passing a wildcard "*" tag name will also yield all Element nodes (and only elements).

In lxml.etree, elements provide further iterators for all directions in the tree: children, parents (or rather ancestors) and siblings.

Serialisation

Serialisation commonly uses the tostring() function that returns a string, or the ElementTree.write() method that writes to a file, a file-like object, or a URL (via FTP PUT or HTTP POST). Both calls accept the same keyword arguments like pretty_print for formatted output or encoding to select a specific output encoding other than plain ASCII:

>>> root = etree.XML('<root><a><b/></a></root>')

>>> etree.tostring(root)
b'<root><a><b/></a></root>'

>>> print(etree.tostring(root, xml_declaration=True))
<?xml version='1.0' encoding='ASCII'?>
<root><a><b/></a></root>

>>> print(etree.tostring(root, encoding='iso-8859-1'))
<?xml version='1.0' encoding='iso-8859-1'?>
<root><a><b/></a></root>

>>> print(etree.tostring(root, pretty_print=True))
<root>
  <a>
    <b/>
  </a>
</root>

Note that pretty printing appends a newline at the end.

In lxml 2.0 and later (as well as ElementTree 1.3), the serialisation functions can do more than XML serialisation. You can serialise to HTML or extract the text content by passing the method keyword:

>>> root = etree.XML(
...    '<html><head/><body><p>Hello<br/>World</p></body></html>')

>>> etree.tostring(root) # default: method = 'xml'
b'<html><head/><body><p>Hello<br/>World</p></body></html>'

>>> etree.tostring(root, method='xml') # same as above
b'<html><head/><body><p>Hello<br/>World</p></body></html>'

>>> etree.tostring(root, method='html')
b'<html><head></head><body><p>Hello<br>World</p></body></html>'

>>> print(etree.tostring(root, method='html', pretty_print=True))
<html>
<head></head>
<body><p>Hello<br>World</p></body>
</html>

>>> etree.tostring(root, method='text')
b'HelloWorld'

As for XML serialisation, the default encoding for plain text serialisation is ASCII:

>>> br = next(root.iter('br'))  # get first result of iteration
>>> br.tail = u'W\xf6rld'

>>> etree.tostring(root, method='text')  # doctest: +ELLIPSIS
Traceback (most recent call last):
  ...
UnicodeEncodeError: 'ascii' codec can't encode character u'\xf6' ...

>>> etree.tostring(root, method='text', encoding="UTF-8")
b'HelloW\xc3\xb6rld'

Here, serialising to a Python unicode string instead of a byte string might become handy. Just pass the unicode type as encoding:

>>> etree.tostring(root, encoding=unicode, method='text')
u'HelloW\xf6rld'

The W3C has a good article about the Unicode character set and character encodings.

The fromstring() function

The fromstring() function is the easiest way to parse a string:

>>> some_xml_data = "<root>data</root>"

>>> root = etree.fromstring(some_xml_data)
>>> print(root.tag)
root
>>> etree.tostring(root)
b'<root>data</root>'

The XML() function

The XML() function behaves like the fromstring() function, but is commonly used to write XML literals right into the source:

>>> root = etree.XML("<root>data</root>")
>>> print(root.tag)
root
>>> etree.tostring(root)
b'<root>data</root>'

There is also a corresponding function HTML() for HTML literals.

The parse() function

The parse() function is used to parse from files and file-like objects.

As an example of such a file-like object, the following code uses the BytesIO class for reading from a string instead of an external file. That class comes from the io module in Python 2.6 and later. In older Python versions, you will have to use the StringIO class from the StringIO module. However, in real life, you would obviously avoid doing this all together and use the string parsing functions above.

>>> some_file_like_object = BytesIO("<root>data</root>")

>>> tree = etree.parse(some_file_like_object)

>>> etree.tostring(tree)
b'<root>data</root>'

Note that parse() returns an ElementTree object, not an Element object as the string parser functions:

>>> root = tree.getroot()
>>> print(root.tag)
root
>>> etree.tostring(root)
b'<root>data</root>'

The reasoning behind this difference is that parse() returns a complete document from a file, while the string parsing functions are commonly used to parse XML fragments.

The parse() function supports any of the following sources:

• an open file object (make sure to open it in binary mode)
• a file-like object that has a .read(byte_count) method returning a byte string on each call
• a filename string
• an HTTP or FTP URL string

Note that passing a filename or URL is usually faster than passing an open file or file-like object. However, the HTTP/FTP client in libxml2 is rather simple, so things like HTTP authentication require a dedicated URL request library, e.g. urllib2 or request. These libraries usually provide a file-like object for the result that you can parse from while the response is streaming in.

Parser objects

By default, lxml.etree uses a standard parser with a default setup. If you want to configure the parser, you can create a you instance:

>>> parser = etree.XMLParser(remove_blank_text=True) # lxml.etree only!

This creates a parser that removes empty text between tags while parsing, which can reduce the size of the tree and avoid dangling tail text if you know that whitespace-only content is not meaningful for your data. An example:

>>> root = etree.XML("<root>  <a/>   <b>  </b>     </root>", parser)

>>> etree.tostring(root)
b'<root><a/><b>  </b></root>'

Note that the whitespace content inside the <b> tag was not removed, as content at leaf elements tends to be data content (even if blank). You can easily remove it in an additional step by traversing the tree:

>>> for element in root.iter("*"):
...     if element.text is not None and not element.text.strip():
...         element.text = None

>>> etree.tostring(root)
b'<root><a/><b/></root>'

See help(etree.XMLParser) to find out about the available parser options.

Incremental parsing

lxml.etree provides two ways for incremental step-by-step parsing. One is through file-like objects, where it calls the read() method repeatedly. This is best used where the data arrives from a source like urllib or any other file-like object that can provide data on request. Note that the parser will block and wait until data becomes available in this case:

>>> class DataSource:
...     data = [ b"<roo", b"t><", b"a/", b"><", b"/root>" ]
...     def read(self, requested_size):
...         try:
...             return self.data.pop(0)
...         except IndexError:
...             return b''

>>> tree = etree.parse(DataSource())

>>> etree.tostring(tree)
b'<root><a/></root>'

The second way is through a feed parser interface, given by the feed(data) and close() methods:

>>> parser = etree.XMLParser()

>>> parser.feed("<roo")
>>> parser.feed("t><")
>>> parser.feed("a/")
>>> parser.feed("><")
>>> parser.feed("/root>")

>>> root = parser.close()

>>> etree.tostring(root)
b'<root><a/></root>'

Here, you can interrupt the parsing process at any time and continue it later on with another call to the feed() method. This comes in handy if you want to avoid blocking calls to the parser, e.g. in frameworks like Twisted, or whenever data comes in slowly or in chunks and you want to do other things while waiting for the next chunk.

After calling the close() method (or when an exception was raised by the parser), you can reuse the parser by calling its feed() method again:

>>> parser.feed("<root/>")
>>> root = parser.close()
>>> etree.tostring(root)
b'<root/>'

Event-driven parsing

Sometimes, all you need from a document is a small fraction somewhere deep inside the tree, so parsing the whole tree into memory, traversing it and dropping it can be too much overhead. lxml.etree supports this use case with two event-driven parser interfaces, one that generates parser events while building the tree (iterparse), and one that does not build the tree at all, and instead calls feedback methods on a target object in a SAX-like fashion.

Here is a simple iterparse() example:

>>> some_file_like = BytesIO("<root><a>data</a></root>")

>>> for event, element in etree.iterparse(some_file_like):
...     print("%s, %4s, %s" % (event, element.tag, element.text))
end,    a, data
end, root, None

By default, iterparse() only generates events when it is done parsing an element, but you can control this through the events keyword argument:

>>> some_file_like = BytesIO("<root><a>data</a></root>")

>>> for event, element in etree.iterparse(some_file_like,
...                                       events=("start", "end")):
...     print("%5s, %4s, %s" % (event, element.tag, element.text))
start, root, None
start,    a, data
  end,    a, data
  end, root, None

Note that the text, tail and children of an Element are not necessarily there yet when receiving the start event. Only the end event guarantees that the Element has been parsed completely.

It also allows to .clear() or modify the content of an Element to save memory. So if you parse a large tree and you want to keep memory usage small, you should clean up parts of the tree that you no longer need:

>>> some_file_like = BytesIO(
...     "<root><a><b>data</b></a><a><b/></a></root>")

>>> for event, element in etree.iterparse(some_file_like):
...     if element.tag == 'b':
...         print(element.text)
...     elif element.tag == 'a':
...         print("** cleaning up the subtree")
...         element.clear()
data
** cleaning up the subtree
None
** cleaning up the subtree

A very important use cases for iterparse() is parsing large generated XML files, e.g. database dumps. Most often, these XML formats only have one main data item element that hangs directly below the root node and that is repeated thousands of times. In this case, it is best practice to let lxml.etree do the tree building and to only intercept exactly on this one Element, using the normal tree API for data extraction.

>>> xml_file = BytesIO('''\
... <root>
...   <a><b>ABC</b><c>abc</c></a>
...   <a><b>MORE DATA</b><c>more data</c></a>
...   <a><b>XYZ</b><c>xyz</c></a>
... </root>''')

>>> for _, element in etree.iterparse(xml_file, tag='a'):
...     print('%s -- %s' % (element.findtext('b'), element[1].text))
...     element.clear()
ABC -- abc
MORE DATA -- more data
XYZ -- xyz

If, for some reason, building the tree is not desired at all, the target parser interface of lxml.etree can be used. It creates SAX-like events by calling the methods of a target object. By implementing some or all of these methods, you can control which events are generated:

>>> class ParserTarget:
...     events = []
...     close_count = 0
...     def start(self, tag, attrib):
...         self.events.append(("start", tag, attrib))
...     def close(self):
...         events, self.events = self.events, []
...         self.close_count += 1
...         return events

>>> parser_target = ParserTarget()

>>> parser = etree.XMLParser(target=parser_target)
>>> events = etree.fromstring('<root test="true"/>', parser)

>>> print(parser_target.close_count)
1

>>> for event in events:
...     print('event: %s - tag: %s' % (event[0], event[1]))
...     for attr, value in event[2].items():
...         print(' * %s = %s' % (attr, value))
event: start - tag: root
 * test = true

You can reuse the parser and its target as often as you like, so you should take care that the .close() methods really resets the target to a usable state (also in the case of an error!).

>>> events = etree.fromstring('<root test="true"/>', parser)
>>> print(parser_target.close_count)
2
>>> events = etree.fromstring('<root test="true"/>', parser)
>>> print(parser_target.close_count)
3
>>> events = etree.fromstring('<root test="true"/>', parser)
>>> print(parser_target.close_count)
4

>>> for event in events:
...     print('event: %s - tag: %s' % (event[0], event[1]))
...     for attr, value in event[2].items():
...         print(' * %s = %s' % (attr, value))
event: start - tag: root
 * test = true