Edgewall Software

Markup Streams

A stream is the common representation of markup as a stream of events.

1   Basics

A stream can be attained in a number of ways. It can be:

  • the result of parsing XML or HTML text, or
  • the result of selecting a subset of another stream using XPath, or
  • programmatically generated.

For example, the functions XML() and HTML() can be used to convert literal XML or HTML text to a markup stream:

>>> from genshi import XML
>>> stream = XML('<p class="intro">Some text and '
...              '<a href="http://example.org/">a link</a>.'
...              '<br/></p>')
>>> stream
<genshi.core.Stream object at ...>

The stream is the result of parsing the text into events. Each event is a tuple of the form (kind, data, pos), where:

  • kind defines what kind of event it is (such as the start of an element, text, a comment, etc).
  • data is the actual data associated with the event. How this looks depends on the event kind (see event kinds)
  • pos is a (filename, lineno, column) tuple that describes where the event “comes from”.
>>> for kind, data, pos in stream:
...     print('%s %r %r' % (kind, data, pos))
START (QName('p'), Attrs([(QName('class'), u'intro')])) (None, 1, 0)
TEXT u'Some text and ' (None, 1, 17)
START (QName('a'), Attrs([(QName('href'), u'http://example.org/')])) (None, 1, 31)
TEXT u'a link' (None, 1, 61)
END QName('a') (None, 1, 67)
TEXT u'.' (None, 1, 71)
START (QName('br'), Attrs()) (None, 1, 72)
END QName('br') (None, 1, 77)
END QName('p') (None, 1, 77)

2   Filtering

One important feature of markup streams is that you can apply filters to the stream, either filters that come with Genshi, or your own custom filters.

A filter is simply a callable that accepts the stream as parameter, and returns the filtered stream:

def noop(stream):
    """A filter that doesn't actually do anything with the stream."""
    for kind, data, pos in stream:
        yield kind, data, pos

Filters can be applied in a number of ways. The simplest is to just call the filter directly:

stream = noop(stream)

The Stream class also provides a filter() method, which takes an arbitrary number of filter callables and applies them all:

stream = stream.filter(noop)

Finally, filters can also be applied using the bitwise or operator (|), which allows a syntax similar to pipes on Unix shells:

stream = stream | noop

One example of a filter included with Genshi is the HTMLSanitizer in genshi.filters. It processes a stream of HTML markup, and strips out any potentially dangerous constructs, such as Javascript event handlers. HTMLSanitizer is not a function, but rather a class that implements __call__, which means instances of the class are callable:

stream = stream | HTMLSanitizer()

Both the filter() method and the pipe operator allow easy chaining of filters:

from genshi.filters import HTMLSanitizer
stream = stream.filter(noop, HTMLSanitizer())

That is equivalent to:

stream = stream | noop | HTMLSanitizer()

For more information about the built-in filters, see Stream Filters.

3   Serialization

Serialization means producing some kind of textual output from a stream of events, which you'll need when you want to transmit or store the results of generating or otherwise processing markup.

The Stream class provides two methods for serialization: serialize() and render(). The former is a generator that yields chunks of Markup objects (which are basically unicode strings that are considered safe for output on the web). The latter returns a single string, by default UTF-8 encoded.

Here's the output from serialize():

>>> for output in stream.serialize():
...     print(repr(output))
<Markup u'<p class="intro">'>
<Markup u'Some text and '>
<Markup u'<a href="http://example.org/">'>
<Markup u'a link'>
<Markup u'</a>'>
<Markup u'.'>
<Markup u'<br/>'>
<Markup u'</p>'>

And here's the output from render():

>>> print(stream.render())
<p class="intro">Some text and <a href="http://example.org/">a link</a>.<br/></p>

Both methods can be passed a method parameter that determines how exactly the events are serialized to text. This parameter can be either a string or a custom serializer class:

>>> print(stream.render('html'))
<p class="intro">Some text and <a href="http://example.org/">a link</a>.<br></p>

Note how the <br> element isn't closed, which is the right thing to do for HTML. See serialization methods for more details.

In addition, the render() method takes an encoding parameter, which defaults to “UTF-8”. If set to None, the result will be a unicode string.

The different serializer classes in genshi.output can also be used directly:

>>> from genshi.filters import HTMLSanitizer
>>> from genshi.output import TextSerializer
>>> print(''.join(TextSerializer()(HTMLSanitizer()(stream))))
Some text and a link.

The pipe operator allows a nicer syntax:

>>> print(stream | HTMLSanitizer() | TextSerializer())
Some text and a link.

3.1   Serialization Methods

Genshi supports the use of different serialization methods to use for creating a text representation of a markup stream.

The XMLSerializer is the default serialization method and results in proper XML output including namespace support, the XML declaration, CDATA sections, and so on. It is not generally not suitable for serving HTML or XHTML web pages (unless you want to use true XHTML 1.1), for which the xhtml and html serializers described below should be preferred.

The XHTMLSerializer is a specialization of the generic XMLSerializer that understands the pecularities of producing XML-compliant output that can also be parsed without problems by the HTML parsers found in modern web browsers. Thus, the output by this serializer should be usable whether sent as "text/html" or "application/xhtml+html" (although there are a lot of subtle issues to pay attention to when switching between the two, in particular with respect to differences in the DOM and CSS).

For example, instead of rendering a script tag as <script/> (which confuses the HTML parser in many browsers), it will produce <script></script>. Also, it will normalize any boolean attributes values that are minimized in HTML, so that for example <hr noshade="1"/> becomes <hr noshade="noshade" />.

This serializer supports the use of namespaces for compound documents, for example to use inline SVG inside an XHTML document.

The HTMLSerializer produces proper HTML markup. The main differences compared to xhtml serialization are that boolean attributes are minimized, empty tags are not self-closing (so it's <br> instead of <br />), and that the contents of <script> and <style> elements are not escaped.
The TextSerializer produces plain text from markup streams. This is useful primarily for text templates, but can also be used to produce plain text output from markup templates or other sources.

3.2   Serialization Options

Both serialize() and render() support additional keyword arguments that are passed through to the initializer of the serializer class. The following options are supported by the built-in serializers:


Whether the serializer should remove trailing spaces and empty lines. Defaults to True.

(This option is not available for serialization to plain text.)


A (name, pubid, sysid) tuple defining the name, publid identifier, and system identifier of a DOCTYPE declaration to prepend to the generated output. If provided, this declaration will override any DOCTYPE declaration in the stream.

The parameter can also be specified as a string to refer to commonly used doctypes:

Shorthand DOCTYPE
html or html-strict HTML 4.01 Strict
html-transitional HTML 4.01 Transitional
html-frameset HTML 4.01 Frameset
html5 DOCTYPE proposed for the work-in-progress HTML5 standard
xhtml or xhtml-strict XHTML 1.0 Strict
xhtml-transitional XHTML 1.0 Transitional
xhtml-frameset XHTML 1.0 Frameset
xhtml11 XHTML 1.1
svg or svg-full SVG 1.1
svg-basic SVG 1.1 Basic
svg-tiny SVG 1.1 Tiny

(This option is not available for serialization to plain text.)


The namespace prefixes to use for namespace that are not bound to a prefix in the stream itself.

(This option is not available for serialization to HTML or plain text.)


Whether to remove the XML declaration (the <?xml ?> part at the beginning of a document) when serializing. This defaults to True as an XML declaration throws some older browsers into "Quirks" rendering mode.

(This option is only available for serialization to XHTML.)


Whether the text serializer should detect and remove any tags or entity encoded characters in the text.

(This option is only available for serialization to plain text.)

4   Using XPath

XPath can be used to extract a specific subset of the stream via the select() method:

>>> substream = stream.select('a')
>>> substream
<genshi.core.Stream object at ...>
>>> print(substream)
<a href="http://example.org/">a link</a>

Often, streams cannot be reused: in the above example, the sub-stream is based on a generator. Once it has been serialized, it will have been fully consumed, and cannot be rendered again. To work around this, you can wrap such a stream in a list:

>>> from genshi import Stream
>>> substream = Stream(list(stream.select('a')))
>>> substream
<genshi.core.Stream object at ...>
>>> print(substream)
<a href="http://example.org/">a link</a>
>>> print(substream.select('@href'))
>>> print(substream.select('text()'))
a link

See Using XPath in Genshi for more information about the XPath support in Genshi.

5   Event Kinds

Every event in a stream is of one of several kinds, which also determines what the data item of the event tuple looks like. The different kinds of events are documented below.


The data item is generally immutable. If the data is to be modified when processing a stream, it must be replaced by a new tuple. Effectively, this means the entire event tuple is immutable.

5.1   START

The opening tag of an element.

For this kind of event, the data item is a tuple of the form (tagname, attrs), where tagname is a QName instance describing the qualified name of the tag, and attrs is an Attrs instance containing the attribute names and values associated with the tag (excluding namespace declarations):

START, (QName('p'), Attrs([(QName('class'), u'intro')])), pos

5.2   END

The closing tag of an element.

The data item of end events consists of just a QName instance describing the qualified name of the tag:

END, QName('p'), pos

5.3   TEXT

Character data outside of elements and comments.

For text events, the data item should be a unicode object:

TEXT, u'Hello, world!', pos

5.4   START_NS

The start of a namespace mapping, binding a namespace prefix to a URI.

The data item of this kind of event is a tuple of the form (prefix, uri), where prefix is the namespace prefix and uri is the full URI to which the prefix is bound. Both should be unicode objects. If the namespace is not bound to any prefix, the prefix item is an empty string:

START_NS, (u'svg', u'http://www.w3.org/2000/svg'), pos

5.5   END_NS

The end of a namespace mapping.

The data item of such events consists of only the namespace prefix (a unicode object):

END_NS, u'svg', pos


A document type declaration.

For this type of event, the data item is a tuple of the form (name, pubid, sysid), where name is the name of the root element, pubid is the public identifier of the DTD (or None), and sysid is the system identifier of the DTD (or None):

DOCTYPE, (u'html', u'-//W3C//DTD XHTML 1.0 Transitional//EN', \
          u'http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd'), pos


A comment.

For such events, the data item is a unicode object containing all character data between the comment delimiters:

COMMENT, u'Commented out', pos

5.8   PI

A processing instruction.

The data item is a tuple of the form (target, data) for processing instructions, where target is the target of the PI (used to identify the application by which the instruction should be processed), and data is text following the target (excluding the terminating question mark):

PI, (u'php', u'echo "Yo" '), pos


Marks the beginning of a CDATA section.

The data item for such events is always None:

START_CDATA, None, pos

5.10   END_CDATA

Marks the end of a CDATA section.

The data item for such events is always None:

END_CDATA, None, pos

See also: genshi.core, Documentation

Last modified 9 years ago Last modified on Dec 14, 2015, 6:22:32 AM