1. Introduction

django-hstore is a niche library which integrates the hstore extension of PostgreSQL into Django.

HStore brings the power of NoSQL key/value stores into PostgreSQL, giving us the advantage of flexibility and performance without renouncing to the robustness of SQL databases.

1.1. Features

  • Postgis compatibility.

  • Python3 support.

  • Nice admin widgets.

  • Possibility to define a schema and use the standard django fields (since django_hstore 1.3.0 and django 1.6)

1.2. Django Admin widget

django-hstore ships a nice admin widget that makes the field more user-friendly.

Admin Widget

Each time a key or a value is modified, the underlying textarea is updated:

Admin Widget

1.3. Grappelli Admin widget

If you use the awesome django-grappelli there’s an even nicer looking widget for you too!

Grappeli Widget

Each time a key or a value is modified, the underlying textarea is updated:

Grappeli Widget

1.4. Limitations

  • PostgreSQL’s implementation of hstore has no concept of type; it stores a mapping of string keys to string values. Values are stored as strings in the database regarding of their original type. This limitation can be overcome by using the schema mode since version 1.3.0 of django_hstore.

  • The hstore extension is not automatically installed on use with this package: you must install it manually.

  • To run tests, hstore extension must be installed on template1 database.

  • The admin widget will work with inlines only if using StackedInline. It won’t work on TabularInline.

  • If django.middleware.transaction.TransactionMiddleware is enabled and the project is deployed through uwsgi, the first request to a view working with models featuring hstore fields will raise an exception; see Django Ticket #22297 for more details on this issue. This issue is specific to Django 1.6 and below.

Warning

Due to hstore being a postgresql extension and not a native type, its oid is different on each database, which causes strange behavior with type conversions (see more on this issue) if hstore extension is installed individually in each database.

To avoid this strange behavior you have two options:

  • Install hstore on template1 postgresql template database and recreate all databases/templates from it, which allows all database to have the same oid for the hstore type (this is the recommended way).

  • Disable global registering setting DJANGO_HSTORE_ADAPTER_REGISTRATION by setting it to connection in your settings. This can have a performance impact because it registers the hstore extension for each new connection created (if you are using django 1.6, persistent connections - or any other connection pool - will help to reduce this impact).

2. Project Maturity

django-hstore is stable, widely used library with well defined deprecation policy.

3. Deprecation policy

At any momment of time, django-hstore developers will mantain support for three versions of django.

As example: The current stable release of django is 1.6, so django-hstore official supported versions are: 1.4, 1.5 and 1.6. When django 1.7 is released, version 1.4 will become an unsupported django version.

4. User Guide

This section covers all aspectes that user want know about django-hstore.

4.1. Install

This section covers a installing django-hstore and its requirements.

4.1.1. Requirements

  • Python 2.6, 2.7 or 3.3+

  • Django 1.5, 1.6 and 1.7

  • Psycopg2 2.4.3+

  • PostgreSQL 9.0+

4.1.2. Stable version

pip install django-hstore

4.1.3. Development version

pip install -e git+git://github.com/djangonauts/django-hstore#egg=django-hstore

4.1.4. Upgrade from older versions

In version 1.2.x some internals have been changed in order to simplify usage and prevent errors.

Values are automatically converted to strings, fields constantly validate input and so on.

If you are upgrading from an older version, ensure your application code works as expected. If it doesn’t you will either have to update your code or tie your application’s requirement to the older version of django-hstore (1.1.1).

4.2. Setup

4.2.1. Basic setup

First, add django_hstore to your settings.INSTALLED_APPS:

INSTALLED_APPS = (
    ...
    "django_hstore",
    ...
)

Second, collect static files (needed for the admin widget) with:

python manage.py collectstatic

4.2.2. Multiple database setup

If for some reason you have to use django-hstore in a multi-database setup and some of the database you are using don’t have the hstore extension installed, you can skip hstore registration by setting HAS_HSTORE to False in your database config:

DATABASES = {
    'default': {
        'ENGINE': 'django.db.backends.postgresql_psycopg2',
        'NAME': 'name',
        'USER': 'user',
        'PASSWORD': 'pass',
        'HOST': 'localhost',
        'PORT': '',
    },
    'other': {
        'ENGINE': 'django.db.backends.postgresql_psycopg2',
        'NAME': 'other',
        'USER': 'user',
        'PASSWORD': 'pass',
        'HOST': 'localhost',
        'PORT': '',
        'HAS_HSTORE': False,
    }
}

If you do that, then don’t try to create DictionaryField in that database.

Be sure to check out allow_syncdb documentation.

4.2.3. Note to South users

If you keep getting errors like "There is no South database module south.db.None" for your database., add the following to settings.py:

SOUTH_DATABASE_ADAPTERS = {'default': 'south.db.postgresql_psycopg2'}

4.3. Usage

The library provides three principal classes:

  • django_hstore.hstore.DictionaryField
    An ORM field which stores a mapping of string key/value pairs in a hstore column.

  • django_hstore.hstore.ReferencesField
    An ORM field which builds on DictionaryField to store a mapping of string keys to django object references, much like ForeignKey.

  • django_hstore.hstore.HStoreManager
    An ORM manager which provides much of the query functionality of the library.

  • django_hstore.hstore.HStoreGeoManager
    An additional ORM manager to provide Geodjango functionality as well.

4.3.1. Model setup

the DictionaryField definition is straightforward:
from django.db import models
from django_hstore import hstore

class Something(models.Model):
    name = models.CharField(max_length=32)
    data = hstore.DictionaryField()  # can pass attributes like null, blank, ecc.

    objects = hstore.HStoreManager()
    # IF YOU ARE USING POSTGIS:
    # objects = hstore.HStoreGeoManager()

Since django_hstore 1.3.0 it is possible to use the DictionaryField in schema mode in order to overcome the limit of values being only strings. Another advantage of using the schema mode is that you can recycle the standard django fields in the admin and hopefully elsewhere. This feature is available only from django 1.6 onwards.

To use the schema mode you just need to supply a schema parameter to the DictionaryField:

# models.py
from django.db import models

class SomethingWithSchema(models.Model):
    name = models.CharField(max_length=32)
    data = hstore.DictionaryField(schema=[
        {
            'name': 'number',
            'class': 'IntegerField',
            'kwargs': {
                'default': 0
            }
        },
        {
            'name': 'float',
            'class': 'FloatField',
            'kwargs': {
                'default': 1.0
            }
        },
        {
            'name': 'char',
            'class': 'CharField',
            'kwargs': {
                'default': 'test', 'blank': True, 'max_length': 10
            }
        },
        {
            'name': 'text',
            'class': 'TextField',
            'kwargs': {
                'blank': True
            }
        },
        {
            'name': 'choice',
            'class': 'CharField',
            'kwargs': {
                'blank': True,
                'max_length': 10,
                'choices': (('choice1', 'choice1'), ('choice2', 'choice2'))
            }
        }
    ])

    objects = hstore.HStoreManager()

After this declaration some additional virtual fields will be available in the model. Each virtual field will map to a key in the dictionary field, types are mantained behind the scenes by using the to_python method of the field class that has been specified for each key.

The schema parameter is a list of dictionaries, each dictionary representing a field.

Each dictionary should have the following keys:

name: indicates the name of the attribute that will be created on the model

class: the field class that will be used to create the virtual field, you can pass it a string and it will look into django.db.models, alternatively you can pass it a concrete class derived from django.db.models.Field imported from anywhere

kwargs: the keyword arguments that will be passed to the Field class. Common arguments are verbose_name, max_length, blank, choices, default.

The following standard django fields fields have been tested successfully:

  • IntegerField

  • FloatField

  • DecimalField

  • BooleanField

  • CharField

  • TextField

  • DateField

  • DateTimeField

  • EmailField

  • GenericIPAddressField

  • URLField

Other fields might work as well except for FileField, ImageField, and BinaryField which would need some additional work.

the ReferenceField definition is also straightforward:
class ReferenceContainer(models.Model):
    name = models.CharField(max_length=32)
    refs = hstore.ReferencesField()

    objects = hstore.HStoreManager()

4.4. Python API

You then treat the data field as simply a dictionary of string pairs:

instance = Something.objects.create(name='something', data={'a': '1', 'b': '2'})
assert instance.data['a'] == '1'

empty = Something.objects.create(name='empty')
assert empty.data == {}

empty.data['a'] = '3'
empty.save()
assert Something.objects.get(name='empty').data['a'] == '3'

In default mode, Booleans, integers, floats, lists, and dictionaries will be converted to strings, while lists, dictionaries, and booleans are converted into JSON formatted strings, so can be decoded if needed:

instance = Something.objects.create(name='something', data={'int': 1, 'bool': True})

instance.data['int'] == '1'
instance.data['bool'] == 'true'

import json
instance.data['dict'] = { 'list': ['a', False, 1] }
instance.data['dict'] == '{"list": ["a", false, 1]}'
json.loads(instance.data['dict']) == { 'list': ['a', False, 1] }
# => True

Since version 1.3.0 you can use the schema mode and you will be able to use virtual fields derived from standard django fields which will take care of validation, default values, type casting, choices and so on. Each virtual field will be mapped to a key of the DictionaryField:

>>> obj = SomethingWithSchema()
>>> obj.number
0
>>> obj.float
1.0
>>> obj.number = 3
>>> obj.float = 9.99
>>> obj.save()
>>> obj = SomethingWithSchema.objects.last()
>>> obj.number
3
>>> obj.data['number']
3
>>> obj.float
9.99
>>> obj.data['float']
9.99

You can issue indexed queries against hstore fields:

# equivalence
Something.objects.filter(data={'a': '1', 'b': '2'})

# comparison (greater than, less than or equal to, ecc)
Something.objects.filter(data__gt={'a': '1'})
Something.objects.filter(data__gte={'a': '1'})
Something.objects.filter(data__lt={'a': '2'})
Something.objects.filter(data__lte={'a': '2'})

# more than one key can be supplied, the result will include the objects which satisfy the
# condition (greater than, less than or equal to, ecc) on all supplied keys
Something.objects.filter(data__gt={'a': '1','b': '2'})
Something.objects.filter(data__gte={'a': '1','b': '2'})
Something.objects.filter(data__lt={'a': '2', 'b': '3'})
Something.objects.filter(data__lte={'a': '2', 'b: '3'})

# subset by key/value mapping
Something.objects.filter(data__contains={'a': '1'})

# subset by list of some key values
Something.objects.filter(data__contains={'a': ['1', '2']})

# subset by list of keys
Something.objects.filter(data__contains=['a', 'b'])

# subset by single key
Something.objects.filter(data__contains=['a'])

You can still do classic django "contains" lookups as you would normally do for normal text fields if you were looking for a particular string. In this case, the HSTORE field will be converted to text and the lookup will be performed on all the keys and all the values:

Something.objects.create(data={ 'some_key': 'some crazy Value' })

# classic text lookup (look up for occurence of string in all the keys)
Something.objects.filter(data__contains='crazy')
Something.objects.filter(data__contains='some_key')
# classic case insensitive text looup
Something.objects.filter(data__icontains='value')
Something.objects.filter(data__icontains='SOME_KEY')

4.5. HSTORE manager

You can also take advantage of some db-side functionality by using the manager:

# identify the keys present in an hstore field
>>> Something.objects.hkeys(id=instance.id, attr='data')
['a', 'b']

# peek at a a named value within an hstore field
>>> Something.objects.hpeek(id=instance.id, attr='data', key='a')
'1'

# do the same, after filter
>>> Something.objects.filter(id=instance.id).hpeek(attr='data', key='a')
'1'

# remove a key/value pair from an hstore field
>>> Something.objects.filter(name='something').hremove('data', 'b')

The hstore methods on manager pass all keyword arguments aside from `attr` and
`key` to `.filter()`.

4.6. ReferenceField Usage

ReferenceField is a field that allows to reference other database objects without using a classic ManyToMany relationship.

Here’s an example with the ReferenceContainer model defined in the Model fields section:

r = ReferenceContainer(name='test')
r.refs['another_object'] = AnotherModel.objects.get(slug='another-object')
r.refs['some_object'] = AnotherModel.objects.get(slug='some-object')
r.save()

r = ReferenceContainer.objects.get(name='test')
r.refs['another_object']
'<AnotherModel: AnotherModel object>'
r.refs['some_object']
'<AnotherModel: AnotherModel some_object>'

The database is queried only when references are accessed directly. Once references have been retrieved they will be stored for any eventual subsequent access:

r = ReferenceContainer.objects.get(name='test')
# this won't query the database
r.refs
{ u'another_object': u'myapp.models.AnotherModel:1',
  u'some_object': u'myapp.models.AnotherModel:2' }

# this will query the database
r.refs['another_object']
'<AnotherModel: AnotherModel object>'

# retrieved reference is now visible also when calling the HStoreDict object:
r.refs
{ u'another_object': <AnotherModel: AnotherModel object>,
  u'some_object': u'myapp.models.AnotherModel:2' }

5. Developers Guide

This section covers everything that django-hstore developer / contributor want know.

5.1. Running tests

Assuming one has the dependencies installed, and a PostgreSQL 9.0+ server up and running:

python runtests.py

By default the tests run with the postgis backend.

If you want to run the tests with psycopg2 backend you can do:

python runtests.py --settings=settings_psycopg

You might need to tweak the DB settings according to your DB configuration.

If you need to do so you can copy the file local_settings.py.example to local_settings.py and add your database tweaks on it. local_settings.py will be automatically imported in settings.py. The same applies for local_settings_psycopg.py.example, which will be imported in local_settings_psycopg.py.

If after running this command you get an error saying:

type "hstore" does not exist

Try this:

psql template1 -c 'create extension hstore;'

5.2. How to contribute

  1. Join the mailing List: link: django-hstore mailing list and announce your intentions.

  2. Follow PEP8, Style Guide for Python Code

  3. Fork this repo

  4. Write code

  5. Write tests for your code

  6. Ensure all tests pass

  7. Ensure test coverage is not under 90%

  8. Document your changes

  9. Send pull request

6. License

Copyright (C) 2013-2014 Federico Capoano

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.


Original Author
===============
Copyright (C) 2011 Jordan McCoy

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.