This is a much clearer way to specify choices for fields in models and forms. A basic example:
>>> from dj.choices import Choices >>> class Gender(Choices): ... _ = Choices.Choice ... ... male = _("Male") ... female = _("Female") ... >>> Gender() [(1, u'Male'), (2, u'Female')] >>> Gender.male <Choice: male (id: 1)> >>> Gender.female <Choice: female (id: 2)> >>> Gender.male.id 1 >>> Gender.male.desc u'Male' >>> Gender.male.raw 'Male' >>> Gender.male.name u'male' >>> Gender.from_name("male") <Choice: male (id: 1)> >>> Gender.id_from_name("male") 1 >>> Gender.raw_from_name("male") 'Male' >>> Gender.desc_from_name("male") u'Male' >>> Gender.name_from_id(2) 'female' >>> Gender.name_from_id(3) Traceback (most recent call last): ... ValueError: Nothing found for '3'. >>> Gender.from_name("perez") Traceback (most recent call last): ... ValueError: Nothing found for 'perez'.
You define a class of choices, specifying each choice as a class attribute.
Those attributes are
int subclasses, numbered automatically starting with
1. The class provides several features which support the DRY principle:
- An object instantiated from the choices class is basically a list of
(id, localized_description)pairs straight for consumption by Django.
- Each attribute defined can be retrieved directly from the class.
- Metadata (e.g. attribute name, raw and localized description, numeric ID) of each attribute is accessible.
- Choices which are suffixed by
_to avoid clashing with Python keywords have this suffix automatically removed in their
- Lookup functions are available to help getting attributes or their metadata.
_ = Choices.Choice trick makes it possible for
makemessages to find each choice description and include it in
files for localization. It masks ugettext only in the scope of the class so
the rest of the module can safely use ugettext or ugettext_lazy. Having to
_ each time is not a particularly pretty solution but it's
explicit. Suggestions for a better approach are welcome.
One of the worst problems with choices is their weak extensibility. For instance, an application defines a group of possible choices like this:
>>> class License(Choices): ... _ = Choices.Choice ... ... gpl = _("GPL") ... bsd = _("BSD") ... proprietary = _("Proprietary") ... >>> License() [(1, u'GPL'), (2, u'BSD'), (3, u'Proprietary')]
All is well until the application goes live and after a while the developer
wants to include LGPL. The natural choice would be to add it after
when we do that, the indexing would break. On the other hand, adding the new
entry at the end of the definition looks ugly and makes the resulting combo
boxes in the UI sorted in a counter-intuitive way. Grouping lets us solve this
problem by explicitly defining the structure within a class of choices:
>>> class License(Choices): ... _ = Choices.Choice ... ... COPYLEFT = Choices.Group(0) ... gpl = _("GPL") ... ... PUBLIC_DOMAIN = Choices.Group(100) ... bsd = _("BSD") ... ... OSS = Choices.Group(200) ... apache2 = _("Apache 2") ... ... COMMERCIAL = Choices.Group(300) ... proprietary = _("Proprietary") ... >>> License() [(1, u'GPL'), (101, u'BSD'), (201, u'Apache 2'), (301, u'Proprietary')]
This enables the developer to include more licenses of each group later on:
>>> class License(Choices): ... _ = Choices.Choice ... ... COPYLEFT = Choices.Group(0) ... gpl_any = _("GPL, any") ... gpl2 = _("GPL 2") ... gpl3 = _("GPL 3") ... lgpl = _("LGPL") ... agpl = _("Affero GPL") ... ... PUBLIC_DOMAIN = Choices.Group(100) ... bsd = _("BSD") ... public_domain = _("Public domain") ... ... OSS = Choices.Group(200) ... apache2 = _("Apache 2") ... mozilla = _("MPL") ... ... COMMERCIAL = Choices.Group(300) ... proprietary = _("Proprietary") ... >>> License() [(1, u'GPL, any'), (2, u'GPL 2'), (3, u'GPL 3'), (4, u'LGPL'), (5, u'Affero GPL'), (101, u'BSD'), (102, u'Public domain'), (201, u'Apache 2'), (202, u'MPL'), (301, u'Proprietary')]
Note the behaviour:
- the developer renamed the GPL choice but its meaning and ID remained stable
- BSD, Apache and proprietary choices have their IDs unchanged
- the resulting class is self-descriptive, readable and extensible
As a bonus, the explicitly specified groups can be used when needed:
>>> License.COPYLEFT <ChoiceGroup: COPYLEFT (id: 0)> >>> License.gpl2 in License.COPYLEFT.choices True >>> [(c.id, c.desc) for c in License.COPYLEFT.choices] [(1, u'GPL, any'), (2, u'GPL 2'), (3, u'GPL 3'), (4, u'LGPL'), (5, u'Affero GPL')]
Choices can be used with generic
When you do that though, some minor API deficiencies show up fairly quickly.
First, when you define the field, you have to instanciate the choices class and
the default value has to be converted to the proper type explicitly:
color = models.IntegerField(choices=Color(), default=Color.green.id)
Second, when getting the attribute back from a model, it has to be converted to a Choice instance to do anything interesting with it:
>>> obj = Model.objects.get(pk=3) >>> obj.color 3 >>> Color.from_id(obj.color) <Choice: Blue (id: 3, name: blue)>
To overcome those problems a
ChoiceField is available in the
dj.choices.fields package. It is based on integers on the database level but
the API exposes
Choice instances. This helps both on the definition side:
color = ChoiceField(choices=Color, default=Color.green)
and on the access side:
>>> obj = Model.objects.get(pk=3) >>> obj.color <Choice: Blue (id: 3, name: blue)> >>> obj.color = Color.green >>> obj.save() >>> Model.objects.get(pk=3).color <Choice: Green (id: 2, name: green)>
For rendering forms, the field coerces to integer values. That also means that
Choice instances are accepted, integers are also fine.
The developer can specify all possible choices for future use and then filter out only the currently applicable values on choices creation:
>>> class Language(Choices): ... _ = Choices.Choice ... ... de = _("German") ... en = _("English") ... fr = _("French") ... pl = _("Polish") ... >>> Language() [(1, u'German'), (2, u'English'), (3, u'French'), (4, u'Polish')] >>> Language(filter=("en", "pl")) [(2, u'English'), (4, u'Polish')]
This has the great advantage of keeping the IDs and sorting intact.
Custom item format
One can also change how the pairs are constructed by providing a factory
function. For instance, to use the class of choices defined above for the
LANGUAGES setting in
settings.py, one could specify:
>>> Language(item=lambda choice: (choice.name, choice.raw)) [(u'de', 'German'), (u'en', 'English'), (u'fr', 'French'), (u'pl', 'Polish')]
Extra attributes on choices
Each choice can receive extra arguments using the
>>> class Python(Choices): ... _ = Choices.Choice ... ... cpython = _("CPython").extra(language='C') ... pypy = _("PyPy").extra(language='Python') ... jython = _("Jython").extra(language='Java') ... iron_python = _("IronPython").extra(language='C#')
This adds a
language attribute to each choice so you can get it back like
>>> Python.jython.language 'Java'
This enables polymorphic attribute access later on when using models or forms. For instance, suppose you have a simple model like:
>>> class Library(models.Model): ... name = models.CharField(max_length=100) ... python_kind = models.IntegerField(choices=Python(), default=Python.cpython.id)
In that case to get the implementation language back you'd do:
>>> library = Library.objects.get(name='dj.choices') >>> Python.from_id(library.python_kind).language 'C'
That frees your user code of any conditionals or dictionaries that depend on the state of the choices class. If you would add another choice to it, no user code needs to be changed to support it. This also supports the DRY principle because the choices class becomes the single place where configuration of that kind is held.
Extra attributes on choice groups
Unsurprisingly, choice groups can have extra attributes as well. They are then inherited by choices in such a group and can be overriden if necessary. For instance:
>>> class ProfileChange(Choices): ... _ = Choices.Choice ... ... USER = Choices.Group(0).extra(icon='bookkeeping.png', is_public=True) ... email = _("e-mail").extra(is_public=False) ... first_name = _("first name") ... last_name = _("last name") ... ... BASIC_INFO = Choices.Group(10).extra(icon='bookkeeping.png', is_public=True) ... birth_date = _("birth date").extra(icon='calendar.png') ... gender = _("gender").extra(icon='male_female.png') ... country = _("country") ... city = _("city") ... ... CONTACT_INFO = Choices.Group(20).extra(icon='contactbook.png', is_public=False) ... skype = _("Skype ID") ... icq = _("ICQ number") ... msn = _("MSN login") ... xfire = _("X-Fire login") ... irc = _("IRC info").extra(is_public=True)
In that case proper inheritance takes place:
>>> ProfileChange.first_name.is_public True >>> ProfileChange.email.is_public False >>> ProfileChange.country.icon 'bookkeeping.png' >>> ProfileChange.birth_date.icon 'calendar.png'
There are several classes of choices which are very common in web applications
so they are provided already:
How do I run the tests?
The easiest way would be to run:
$ DJANGO_SETTINGS_MODULE="dj._choicestestproject.settings" django-admin.py test
- Drop compatibility for Django < 1.4 and Python 2.6
- Add support for Django 1.9 and 1.10 (use
- Add support for Python 3.5
- Works and tests run on Django 1.6 and Django 1.7; now tested with Python 3.4 as well
- Python 2.6 is now supported as well thanks to Carl van Tonder
- Long overdue Python 3 support (considered experimental)
- Choices are now
intsubclasses so you can use a choice directly instead of
int(choice)is always safe
unicode(choice)is now equivalent to
get_FIELD_display()on models with
- Values outside of defined choices for a
ChoiceFieldnow correctly display validation errors instead of throwing exceptions; fixes issue #2
- Fixed regression from 0.8.5 where
__lt(e)couldn't be used on
- Minor refinements
ChoiceFieldis now correctly South-migrable
- Models with
ChoiceFieldscan now use
__isnulllookups on them; fixes issue #1
- proper ChoiceField support if the underlying
longinstead of an
__unicode__corrections for byte strings
MANIFEST.inwas previously missing which made the source distribution hard to install
- extra attribute injection API is now public and documented
- old accessors temporarily restored for backward compatibility (undocumented and to be removed in 1.0)
- minor documentation fixes
- code separated from
- PEP8-fied the accessor APIs