unicode.txt

============
Unicode data
============

Django natively supports Unicode data everywhere. Providing your database can
somehow store the data, you can safely pass around Unicode strings to
templates, models and the database.

This document tells you what you need to know if you're writing applications
that use data or templates that are encoded in something other than ASCII.

Creating the database
=====================

Make sure your database is configured to be able to store arbitrary string
data. Normally, this means giving it an encoding of UTF-8 or UTF-16. If you use
a more restrictive encoding -- for example, latin1 (iso8859-1) -- you won't be
able to store certain characters in the database, and information will be lost.

* MySQL users, refer to the `MySQL manual`_ (section 9.1.3.2 for MySQL 5.1)
  for details on how to set or alter the database character set encoding.

* PostgreSQL users, refer to the `PostgreSQL manual`_ (section 21.2.2 in
  PostgreSQL 8) for details on creating databases with the correct encoding.

* SQLite users, there is nothing you need to do. SQLite always uses UTF-8
  for internal encoding.

.. _MySQL manual: http://dev.mysql.com/doc/refman/5.1/en/charset-database.html
.. _PostgreSQL manual: http://www.postgresql.org/docs/8.2/static/multibyte.html#AEN24104

All of Django's database backends automatically convert Unicode strings into
the appropriate encoding for talking to the database. They also automatically
convert strings retrieved from the database into Python Unicode strings. You
don't even need to tell Django what encoding your database uses: that is
handled transparently.

For more, see the section "The database API" below.

General string handling
=======================

Whenever you use strings with Django -- e.g., in database lookups, template
rendering or anywhere else -- you have two choices for encoding those strings.
You can use Unicode strings, or you can use normal strings (sometimes called
"bytestrings") that are encoded using UTF-8.

.. versionchanged:: 1.5

In Python 3, the logic is reversed, that is normal strings are Unicode, and
when you want to specifically create a bytestring, you have to prefix the
string with a 'b'. As we are doing in Django code from version 1.5,
we recommend that you import ``unicode_literals`` from the __future__ library
in your code. Then, when you specifically want to create a bytestring literal,
prefix the string with 'b'.

Python 2 legacy::

    my_string = "This is a bytestring"
    my_unicode = u"This is an Unicode string"

Python 2 with unicode literals or Python 3::

    from __future__ import unicode_literals

    my_string = b"This is a bytestring"
    my_unicode = "This is an Unicode string"

See also :doc:`Python 3 compatibility </topics/python3>`.

.. admonition:: Warning

    A bytestring does not carry any information with it about its encoding.
    For that reason, we have to make an assumption, and Django assumes that all
    bytestrings are in UTF-8.

    If you pass a string to Django that has been encoded in some other format,
    things will go wrong in interesting ways. Usually, Django will raise a
    ``UnicodeDecodeError`` at some point.

If your code only uses ASCII data, it's safe to use your normal strings,
passing them around at will, because ASCII is a subset of UTF-8.

Don't be fooled into thinking that if your :setting:`DEFAULT_CHARSET` setting is set
to something other than ``'utf-8'`` you can use that other encoding in your
bytestrings! :setting:`DEFAULT_CHARSET` only applies to the strings generated as
the result of template rendering (and email). Django will always assume UTF-8
encoding for internal bytestrings. The reason for this is that the
:setting:`DEFAULT_CHARSET` setting is not actually under your control (if you are the
application developer). It's under the control of the person installing and
using your application -- and if that person chooses a different setting, your
code must still continue to work. Ergo, it cannot rely on that setting.

In most cases when Django is dealing with strings, it will convert them to
Unicode strings before doing anything else. So, as a general rule, if you pass
in a bytestring, be prepared to receive a Unicode string back in the result.

Translated strings
------------------

Aside from Unicode strings and bytestrings, there's a third type of string-like
object you may encounter when using Django. The framework's
internationalization features introduce the concept of a "lazy translation" --
a string that has been marked as translated but whose actual translation result
isn't determined until the object is used in a string. This feature is useful
in cases where the translation locale is unknown until the string is used, even
though the string might have originally been created when the code was first
imported.

Normally, you won't have to worry about lazy translations. Just be aware that
if you examine an object and it claims to be a
``django.utils.functional.__proxy__`` object, it is a lazy translation.
Calling ``unicode()`` with the lazy translation as the argument will generate a
Unicode string in the current locale.

For more details about lazy translation objects, refer to the
:doc:`internationalization </topics/i18n/index>` documentation.

Useful utility functions
------------------------

Because some string operations come up again and again, Django ships with a few
useful functions that should make working with Unicode and bytestring objects
a bit easier.

Conversion functions
~~~~~~~~~~~~~~~~~~~~

The ``django.utils.encoding`` module contains a few functions that are handy
for converting back and forth between Unicode and bytestrings.

* ``smart_unicode(s, encoding='utf-8', strings_only=False, errors='strict')``
  converts its input to a Unicode string. The ``encoding`` parameter
  specifies the input encoding. (For example, Django uses this internally
  when processing form input data, which might not be UTF-8 encoded.) The
  ``strings_only`` parameter, if set to True, will result in Python
  numbers, booleans and ``None`` not being converted to a string (they keep
  their original types). The ``errors`` parameter takes any of the values
  that are accepted by Python's ``unicode()`` function for its error
  handling.

  If you pass ``smart_unicode()`` an object that has a ``__unicode__``
  method, it will use that method to do the conversion.

* ``force_unicode(s, encoding='utf-8', strings_only=False,
  errors='strict')`` is identical to ``smart_unicode()`` in almost all
  cases. The difference is when the first argument is a :ref:`lazy
  translation <lazy-translations>` instance. While ``smart_unicode()``
  preserves lazy translations, ``force_unicode()`` forces those objects to a
  Unicode string (causing the translation to occur). Normally, you'll want
  to use ``smart_unicode()``. However, ``force_unicode()`` is useful in
  template tags and filters that absolutely *must* have a string to work
  with, not just something that can be converted to a string.

* ``smart_str(s, encoding='utf-8', strings_only=False, errors='strict')``
  is essentially the opposite of ``smart_unicode()``. It forces the first
  argument to a bytestring. The ``strings_only`` parameter has the same
  behavior as for ``smart_unicode()`` and ``force_unicode()``. This is
  slightly different semantics from Python's builtin ``str()`` function,
  but the difference is needed in a few places within Django's internals.

Normally, you'll only need to use ``smart_unicode()``. Call it as early as
possible on any input data that might be either Unicode or a bytestring, and
from then on, you can treat the result as always being Unicode.

.. _uri-and-iri-handling:

URI and IRI handling
~~~~~~~~~~~~~~~~~~~~

Web frameworks have to deal with URLs (which are a type of IRI_). One
requirement of URLs is that they are encoded using only ASCII characters.
However, in an international environment, you might need to construct a
URL from an IRI_ -- very loosely speaking, a URI_ that can contain Unicode
characters. Quoting and converting an IRI to URI can be a little tricky, so
Django provides some assistance.

* The function ``django.utils.encoding.iri_to_uri()`` implements the
  conversion from IRI to URI as required by the specification (:rfc:`3987`).

* The functions ``django.utils.http.urlquote()`` and
  ``django.utils.http.urlquote_plus()`` are versions of Python's standard
  ``urllib.quote()`` and ``urllib.quote_plus()`` that work with non-ASCII
  characters. (The data is converted to UTF-8 prior to encoding.)

These two groups of functions have slightly different purposes, and it's
important to keep them straight. Normally, you would use ``urlquote()`` on the
individual portions of the IRI or URI path so that any reserved characters
such as '&' or '%' are correctly encoded. Then, you apply ``iri_to_uri()`` to
the full IRI and it converts any non-ASCII characters to the correct encoded
values.

.. note::
    Technically, it isn't correct to say that ``iri_to_uri()`` implements the
    full algorithm in the IRI specification. It doesn't (yet) perform the
    international domain name encoding portion of the algorithm.

The ``iri_to_uri()`` function will not change ASCII characters that are
otherwise permitted in a URL. So, for example, the character '%' is not
further encoded when passed to ``iri_to_uri()``. This means you can pass a
full URL to this function and it will not mess up the query string or anything
like that.

An example might clarify things here::

    >>> urlquote(u'Paris & Orléans')
    u'Paris%20%26%20Orl%C3%A9ans'
    >>> iri_to_uri(u'/favorites/François/%s' % urlquote('Paris & Orléans'))
    '/favorites/Fran%C3%A7ois/Paris%20%26%20Orl%C3%A9ans'

If you look carefully, you can see that the portion that was generated by
``urlquote()`` in the second example was not double-quoted when passed to
``iri_to_uri()``. This is a very important and useful feature. It means that
you can construct your IRI without worrying about whether it contains
non-ASCII characters and then, right at the end, call ``iri_to_uri()`` on the
result.

The ``iri_to_uri()`` function is also idempotent, which means the following is
always true::

    iri_to_uri(iri_to_uri(some_string)) = iri_to_uri(some_string)

So you can safely call it multiple times on the same IRI without risking
double-quoting problems.

.. _URI: http://www.ietf.org/rfc/rfc2396.txt
.. _IRI: http://www.ietf.org/rfc/rfc3987.txt

Models
======

Because all strings are returned from the database as Unicode strings, model
fields that are character based (CharField, TextField, URLField, etc) will
contain Unicode values when Django retrieves data from the database. This
is *always* the case, even if the data could fit into an ASCII bytestring.

You can pass in bytestrings when creating a model or populating a field, and
Django will convert it to Unicode when it needs to.

Choosing between ``__str__()`` and ``__unicode__()``
----------------------------------------------------

One consequence of using Unicode by default is that you have to take some care
when printing data from the model.

In particular, rather than giving your model a ``__str__()`` method, we
recommended you implement a ``__unicode__()`` method. In the ``__unicode__()``
method, you can quite safely return the values of all your fields without
having to worry about whether they fit into a bytestring or not. (The way
Python works, the result of ``__str__()`` is *always* a bytestring, even if you
accidentally try to return a Unicode object).

You can still create a ``__str__()`` method on your models if you want, of
course, but you shouldn't need to do this unless you have a good reason.
Django's ``Model`` base class automatically provides a ``__str__()``
implementation that calls ``__unicode__()`` and encodes the result into UTF-8.
This means you'll normally only need to implement a ``__unicode__()`` method
and let Django handle the coercion to a bytestring when required.

Taking care in ``get_absolute_url()``
-------------------------------------

URLs can only contain ASCII characters. If you're constructing a URL from
pieces of data that might be non-ASCII, be careful to encode the results in a
way that is suitable for a URL. The ``django.db.models.permalink()`` decorator
handles this for you automatically.

If you're constructing a URL manually (i.e., *not* using the ``permalink()``
decorator), you'll need to take care of the encoding yourself. In this case,
use the ``iri_to_uri()`` and ``urlquote()`` functions that were documented
above_. For example::

    from django.utils.encoding import iri_to_uri
    from django.utils.http import urlquote

    def get_absolute_url(self):
        url = u'/person/%s/?x=0&y=0' % urlquote(self.location)
        return iri_to_uri(url)

This function returns a correctly encoded URL even if ``self.location`` is
something like "Jack visited Paris & Orléans". (In fact, the ``iri_to_uri()``
call isn't strictly necessary in the above example, because all the
non-ASCII characters would have been removed in quoting in the first line.)

.. _above: `URI and IRI handling`_

The database API
================

You can pass either Unicode strings or UTF-8 bytestrings as arguments to
``filter()`` methods and the like in the database API. The following two
querysets are identical::

    from __future__ import unicode_literals

    qs = People.objects.filter(name__contains='Å')
    qs = People.objects.filter(name__contains=b'\xc3\x85') # UTF-8 encoding of Å

Templates
=========

You can use either Unicode or bytestrings when creating templates manually::

    from __future__ import unicode_literals
    from django.template import Template
    t1 = Template(b'This is a bytestring template.')
    t2 = Template('This is a Unicode template.')

But the common case is to read templates from the filesystem, and this creates
a slight complication: not all filesystems store their data encoded as UTF-8.
If your template files are not stored with a UTF-8 encoding, set the :setting:`FILE_CHARSET`
setting to the encoding of the files on disk. When Django reads in a template
file, it will convert the data from this encoding to Unicode. (:setting:`FILE_CHARSET`
is set to ``'utf-8'`` by default.)

The :setting:`DEFAULT_CHARSET` setting controls the encoding of rendered templates.
This is set to UTF-8 by default.

Template tags and filters
-------------------------

A couple of tips to remember when writing your own template tags and filters:

* Always return Unicode strings from a template tag's ``render()`` method
  and from template filters.

* Use ``force_unicode()`` in preference to ``smart_unicode()`` in these
  places. Tag rendering and filter calls occur as the template is being
  rendered, so there is no advantage to postponing the conversion of lazy
  translation objects into strings. It's easier to work solely with Unicode
  strings at that point.

Email
=====

Django's email framework (in ``django.core.mail``) supports Unicode
transparently. You can use Unicode data in the message bodies and any headers.
However, you're still obligated to respect the requirements of the email
specifications, so, for example, email addresses should use only ASCII
characters.

The following code example demonstrates that everything except email addresses
can be non-ASCII::

    from __future__ import unicode_literals
    from django.core.mail import EmailMessage

    subject = 'My visit to Sør-Trøndelag'
    sender = 'Arnbjörg Ráðormsdóttir <arnbjorg@example.com>'
    recipients = ['Fred <fred@example.com']
    body = '...'
    msg = EmailMessage(subject, body, sender, recipients)
    msg.attach("Une pièce jointe.pdf", "%PDF-1.4.%...", mimetype="application/pdf")
    msg.send()

Form submission
===============

HTML form submission is a tricky area. There's no guarantee that the
submission will include encoding information, which means the framework might
have to guess at the encoding of submitted data.

Django adopts a "lazy" approach to decoding form data. The data in an
``HttpRequest`` object is only decoded when you access it. In fact, most of
the data is not decoded at all. Only the ``HttpRequest.GET`` and
``HttpRequest.POST`` data structures have any decoding applied to them. Those
two fields will return their members as Unicode data. All other attributes and
methods of ``HttpRequest`` return data exactly as it was submitted by the
client.

By default, the :setting:`DEFAULT_CHARSET` setting is used as the assumed encoding
for form data. If you need to change this for a particular form, you can set
the ``encoding`` attribute on an ``HttpRequest`` instance. For example::

    def some_view(request):
        # We know that the data must be encoded as KOI8-R (for some reason).
        request.encoding = 'koi8-r'
        ...

You can even change the encoding after having accessed ``request.GET`` or
``request.POST``, and all subsequent accesses will use the new encoding.

Most developers won't need to worry about changing form encoding, but this is
a useful feature for applications that talk to legacy systems whose encoding
you cannot control.

Django does not decode the data of file uploads, because that data is normally
treated as collections of bytes, rather than strings. Any automatic decoding
there would alter the meaning of the stream of bytes.