listxml 0.5.0

An _unpythonic_ XML writer (and reader)!

Listxml

Listxml is an unpythonic XML wrangler!

Version: 0.5.0, released 2025 September 21.

Python provides the DOM and ElementTree interfaces for creating an XML tree, and serialising it for output. These work well, but can hardly be called lightweight; it can be wearisome to programmatically assemble an XML document, and because the document is assembled from a multitude of method calls, it's easy to lose the wood amongst the trees.

The listxml package provides a different way of creating a data structure which can be serialised into XML. That structure might represent an (X)HTML file, or something like an RSS feed. The structure's easy to generate programmatically (list comprehensions are your friend!), and, because it's compact, you can see more of it on the screen at once.

Rather than be clever about fancy syntax, Listxml aims for minimalism and a homogeneous representation (pssst: if you think this looks a bit lispy, you would not be mistaken).

There is also a simple function for searching a listxml list in an XML-like way. It implements a (very) small fragment of the XPath syntax (it's much less ambitious, but we aim not to be gratuitously incompatible with XPath).

This is still a beta release: the interface and functionality is still not 100% fixed, and comments would be welcome.

For example:

import listxml

# l is a list representing an element
l = ['div',                      # a <div> element
     ['p',                       # the element name is a string
      [['class', 'highlight']],  # attributes are a list of two-element lists
      "Hello, world", 99],       # element content, string or str()-friendly
     ["p", "& another <para>"]]  # no attributes, and escaped content

coll = listxml.Collector()
listxml.list_to_collector(l, coll)

# print the resulting byte content
for content in coll:
    print(content.decode('utf-8'), end='')

# or use it as an iterator
print(b''.join(coll))

In each case, the printed result is

<div><p class="highlight">Hello, world99</p><p>&amp; another &lt;para&gt;</p></div>

Alternatively, use PrintCollector as a collector, to send output to stdout or another stream (list_to_stream does that).

coll = listxml.PrintCollector()
listxml.list_to_collector(l, coll)

A ‘collector’, here, is anything which matches the Collector interface described below.

The intention is that, as long as you don't use the b'bytestring' escape mechanism mentioned below, it should be impossible to serialise an invalid XML file using this package.

For symmetry, the package also includes a way of turning XML into the sort of list it expects (wrapping the expat parser built in to Python):

# The named file contains "<xml>...</xml>"...
filename = "foo.xml"
l = listxml.construct(filename)

coll = listxml.Collector()
listxml.list_to_collector(l, coll)
b''.join(coll)

...should produce output which is equivalent, in XML terms, to the input file.

A fuller example

This example assembles an HTML page body, and drops it into a ‘template’.

import listxml

def wrap_body(title, body):
    """Create a standard XHTML document (ie, this is a form of templating)"""
    return ['html', [['xmlns', 'http://www.w3.org/1999/xhtml']],
            ['head',
             ['title', title],
             ['link', [['rel', 'stylesheet'],
                       ['type', 'text/css'],
                       ['href', 'http://example.org/mystyle.css']]]],
            ['body',
             ['h1', title],
             *body]]

# assemble a list of li elements
items = ["First item", "Second item"]
ul = [['li', i] for i in items]

# build up a list of body content elements
b = [['p', 'One paragraph'],
     ['p', "Another one, with ",
      ['a', [['href', 'http://example.org/home.html']],
       "a link"]],
     ['ul', *ul]] # append the ul list to make list contents

# use the PrintCollector to send this to stdout
coll = listxml.PrintCollector()
listxml.list_to_collector(wrap_body("My XHTML file", b), coll)

Classes and functions

The package defines the following classes and functions.

Function list_to_collector(lx, coll)

• lx: a list representation of an XML document
• coll: a Collector – see below

Convert the input list to XML and send it to the collector. See below for the structure of the input list. Returns the input collector.

In fact, the 'coll' object can be any object with an append() method.

Function list_to_stream(lx, stream=None)

• lx: a list representation of an XML document
• stream: a text stream, such as sys.stdout, the object returned from open(), or an io.StringIO object.

As with list_to_collector, except that the contents are 'collected' to stdout. If the stream argument is present, send the output there instead. This function returns the number of characters written to the stream.

Class Collector

Collect strings or bytes, and return them as a iterator of bytestrings. The Collector object is given to the list_to_collector function to accumulate the results of the conversion of the list. The Collector object may subsequently be treated as an iterator, returning a sequence of bytestrings. This may therefore be printed as:

content = ['div', ['p', "Hello, world"], ["p", "Another paragraph"]]

coll = listxml.Collector()
listxml.list_to_collector(content, coll)
for bs in coll:
    print(bs.decode('utf-8'), end='')

or write it out as a single bytestring:

with open('output.xml', 'wb') as f:
    f.write(b''.join(coll))

We can also ‘append’ strings to the collector in an unsusprising way:

coll = listxml.Collector()
coll.append('<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">\n')
listxml.list_to_collector(content, coll)
...

Methods:

• append(s): append something to the collector, which can be a string, a bytestring, or anything str() can work with. Returns self.
• get_length(): return the length of the contents, in bytes.

Class PrintCollector

A Collector-like object which 'collects’ its output and sends it to a stream. The output is sent to sys.stdout, unless an alternative is set with the set_stream() method.

The ‘stream’ must be a text file, such as sys.stdout, the stream returned by the open() function, or an in-memory object such as io.StringIO.

This is not in fact a subclass of Collector, though it has the same interface.

coll = listxml.PrintCollector()
listxml.list_to_collector(content, coll)

Keyword arguments:

• stream : if the object is constructed with stream=foo, then that stream is installed as the default stream to which the collector writes, instead of sys.stdout.
• file : the default stream is created as an output file pointing to this file; when called in this way, the object can be used as a Context Manager, or the object's close() function can close the collector and stream later.

Methods: as with Collector, with some adjustments

• append(s): append something to the collector, as with Collector. Returns self.
• get_length(): returns the number of characters written to the stream.
• set_stream(s): set the stream that is written to.
• close(): close the output stream, if it is a file (but not if it is a stream, since the caller may wish to write more to the stream)).

The PrintCollector object can also be used as a context manager (in which case it will usually make sense to include the file parameter).

with listxml.PrintCollector(file='myoutput.xml') as coll:
    listxml.list_to_collector(content, coll)

Or alternatively just use the list_to_stream function:

with open('output.xml', 'w') as f:
    listxml.list_to_stream(content, f)

To print to a string:

s = io.StringIO()
with PrintCollector(stream=s) as coll:
    list_to_collector(['p', "hello"], coll)
print(s.getvalue())

Function construct(filename_or_stream, keywords...)

For symmetry, there is also a function to turn an XML source into a list. Given a (string) filename or a text stream containing XML, this constructs a list representation of the XML, and returns it.

Keyword arguments:

• attributes_as_dict: If attributes_as_dict is False (default) then attributes are [['name','value'], ...]; if it is True, then attributes are a dict {'name': 'value', ...}.
• omit_empty_attlist: If omit_empty_attlist is False (default) then there is always an attribute element, even when the attribute list is empty (ie, [] or {}); if it is True, then empty attribute lists are suppressed.

This reading function is, in this version, not XML Namespace-aware. Adding that isn't hard, but it's currently unclear how best to represent namespaces in a convenient way, when generating the input list for writing. Thus, at present, xmlns attributes in the input XML are not interpreted in any special way.

If the argument is not a file name, it is an input stream, which can come from a file open() or via io.StringIO, to read from a string containing XML:

lx = construct("/path/to/file")

with open("/path/to/file") as f:
    lx = construct(f)

lx = construct(io.StringIO("<p>hello</p>"))

Function search_for_path(path, lx, with_element=False)

A simple path query. The syntax is a (very small!) subset of the XPath syntax.

The path argument is

path      ::= node-spec ( '/' node-spec)*
node-spec ::=   element-name
              | element-name '[' predicate ']'
              | '@' attribute-name
              | <number>
predicate ::= '@' attribute-name '=' attribute-value
element-name, attribute-name ::= [a-zA-Z0-9_-]+
attribute-value ::= '"' [^"]* '"' | "'" [^']* "'"

The function selects elements and attributes within the list which have parent-child relationships which match the path, and returns a list of all of the matching elements (rather than merely the first one).

If the path ends with a @attname then the function instead returns a list of attribute values.

If an element is qualified by a predicate, [@att="value"], then it matches only if it has an element @att with the given value. The attribute-value can be enclosed in single- or double-quotes.

If it has predicate [<n>], then it matches only the n-th element in the list (where the first element is no.1).

If the node-spec doesn't match the above syntax, then a ValueError is raised.

For example, given the document

doc = ['div',
       ['foo', "p1"],
       ['bar', [['a1', 'v1x']], "p1x"],
       ['foo', ['p', ['bar', 'content']]],
       ['foo', ['bar', "p2"]],
       ['foo', ['bar', ['baz', "p3"]]],
       ['foo', ['bar', [['a1', 'v1']], "text", ['p', "p4"]]],
       ['foo', ['bar', [['a1', 'v2'], ['a3', 'v3']], "p5"]]]

the following path arguments produce the given results

path	search_for_path(path, doc)
foo/bar	[['p2'], [['baz', 'p3']], ['text', ['p', 'p4']], ['p5']]
foo/bar/@a1	['v1', 'v2']
foo/bar[@a1="v1"]	[['text', ['p', 'p4']]]
foo/bar[@a1="v1"]/p	[['p4']]
foo/bar[@a1="v2"]/@a3	['v3']

If with_element is True, then include the matching element, with attributes, rather than only the content. For example, with the above document, search_for_path('foo/bar[@a1="v1"]', doc, with_element=True) produces [['bar', {'a1': 'v1'}, 'p4']] rather than only the content [['p4']].

Note: the syntax of this function's path argument changed in v0.4.0.

Function is_listxml_p(lx)

Return true if the argument is a valid listxml representation of an element. See below for the definition.

Common techniques

Assemble a list:

trs = [['tr', ['td', 'foo']],
       ['tr', ['td', 'bar']]]
table = ['table', *trs]   # wrap an array of elements in a parent element

doc = ['body',
       ['p',
        [['class', 'highlight']],
        "Here is table no.", 1],
       table,
       ['p', "that was ", ['em', "easy"]]]

with open('t.xhtml', 'w') as f:
    listxml.list_to_stream(doc, stream=f)

Part of the point of this library is that in some circumstances it's convenient to generate list content:

elements = ['one', 'two']
trs = [['tr', ['td', e]] for e in elements]

In this context, note the difference between

table1 = ['table', trs]

and

table2 = ['table', *trs]

The first produces

['table', [['tr', ['td', 'one']], ['tr', ['td', 'two']]]]

which is not the structure desired, because this appears to be an attribute tr, with value ['td', 'one'] (this won't produce an error, since the package will (successfully) call str() on the attribute value). In contrast the second version produces

['table', ['tr', ['td', 'one']], ['tr', ['td', 'two']]]

which is correct, and which turns into

<table><tr><td>one</td></tr><tr><td>two</td></tr></table>

Another possibility would be table = ['table']; table.extend(trs).

Input syntax

The input list consists of a single element representing an XML document, where

element: [STRING, optional-attributes?, item ...]
optional-attributes: [] | [[STRING, stringable], ...] | DICT
item: element | stringable | BYTESTRING

where STRING and BYTESTRING are the Python types, DICT is a (STRING -> stringable) Python dictionary, and stringable is either a string, or something (other than an optional-attributes) which str() can turn into a string.

Thus:

['el', 'foo', 'b&r', ...]                         -- an element <el>foob&amp;r...</el>
['el', [['k1', 'v1'], ['k2', 'v2'], ...]], ...]   -- an element <el k1="v1" k2="v2"...>...</el>
['el', {'k1': 'v1', 'k2': 'v2', ...}, ...]        -- ditto

and the ... may include other such elements. Items which are ‘stringable’ are escaped when being printed. Items which are bytestrings are not; thus it's possible to have b'<div>content</div>' as an item and this will be emitted as-is, even if doing so would produce invalid XML.

Release notes

Release 0.5.0 (2025 September 21):

• search_for_path: the predicate was unduly strict in its syntax. It now accepts any string as an attribute-value.
• search_for_path: added predicate p[<n>], to match the n-th p element.

Release 0.4.2 (2024 September 31):

• Fixed error in README.
• Check argument types.
• Minor refactoring.
• Fixed: PrintCollector(stream=s) shouldn't close the stream.

Release 0.4.1 (2023 March 22):

• Fixed .../@att search path in search_for_path.

Release 0.4.0 (2023 March 5):

• The path argument to search_for_path changed to be a (small!) subset of the XPath syntax. This is incompatible with the 0.3.0 syntax.

Release 0.3.0 (2022 September 25):

• PrintCollector can now be used as a context manager.
• The order of the arguments to list_to_collector has been swapped.