==============================
The public C-API of lxml.etree
==============================

As of version 1.1, lxml.etree provides a public C-API.  This allows external
C extensions to efficiently access public functions and classes of lxml,
without going through the Python API.

The API is described in the file `etreepublic.pxd`_, which is directly
c-importable by extension modules implemented in Cython_.

.. _`etreepublic.pxd`: https://github.com/lxml/lxml/blob/master/src/lxml/includes/etreepublic.pxd
.. _Cython: https://cython.org

.. contents::
..
   1  Passing generated trees through Python
   2  Writing external modules in Cython
   3  Writing external modules in C


Passing generated trees through Python
--------------------------------------

This is the most simple way to integrate with lxml.  It does not require
any C-level integration but uses a Python function to wrap an externally
generated libxml2 document in lxml.

The external module that creates the libxml2 tree must pack the document
pointer into a `PyCapsule <https://docs.python.org/3/c-api/capsule.html>`_
object.  This can then be passed into lxml with the function
``lxml.etree.adopt_external_document()``.  It also takes an optional lxml
parser instance to associate with the document, in order to configure the
Element class lookup, relative URL lookups, etc.

See the `API reference <api/lxml.etree-module.html#adopt_external_document>`_
for further details.

The same functionality is available as part of the public C-API in form
of the C function ``adoptExternalDocument()``.


Writing external modules in Cython
----------------------------------

This is the easiest way of extending lxml at the C level.  A Cython_
module should start like this::

    # My Cython extension

    # directive pointing compiler to lxml header files;
    # use ``aliases={"LXML_PACKAGE_DIR": lxml.__path__}``
    # argument to cythonize in setup.py to dynamically
    # determine dir at compile time
    # distutils: include_dirs = LXML_PACKAGE_DIR

    # import the public functions and classes of lxml.etree
    cimport lxml.includes.etreepublic as cetree

    # import the lxml.etree module in Python
    cdef object etree
    from lxml import etree

    # initialize the access to the C-API of lxml.etree
    cetree.import_lxml__etree()

From this line on, you can access all public functions of lxml.etree
from the ``cetree`` namespace like this::

    # build a tag name from namespace and element name
    py_tag = cetree.namespacedNameFromNsName("http://some/url", "myelement")

Public lxml classes are easily subclassed.  For example, to implement
and set a new default element class, you can write Cython code like
the following::

    from lxml.includes.etreepublic cimport ElementBase
    cdef class NewElementClass(ElementBase):
         def set_value(self, myval):
             self.set("my_attribute", myval)

    etree.set_element_class_lookup(
         etree.ElementDefaultClassLookup(element=NewElementClass))


Writing external modules in C
-----------------------------

If you really feel like it, you can also interface with lxml.etree straight
from C code.  All you have to do is include the header file for the public
API, import the ``lxml.etree`` module and then call the import function:

.. sourcecode:: c

    /* My C extension */

    /* common includes */
    #include "Python.h"
    #include "stdio.h"
    #include "string.h"
    #include "stdarg.h"
    #include "libxml/xmlversion.h"
    #include "libxml/encoding.h"
    #include "libxml/hash.h"
    #include "libxml/tree.h"
    #include "libxml/xmlIO.h"
    #include "libxml/xmlsave.h"
    #include "libxml/globals.h"
    #include "libxml/xmlstring.h"

    /* lxml.etree specific includes */
    #include "lxml-version.h"
    #include "etree_defs.h"
    #include "etree.h"

    /* setup code */
    import_lxml__etree()

Note that including ``etree.h`` does not automatically include the
header files it requires.  Note also that the above list of common
includes may not be sufficient.