django-pg-returning

A small library implementing PostgreSQL ability to return rows in DML statements for Django.
Link to PostgreSQL docs

Requirements

• Python 2.7 or Python 3.5+
• django >= 1.8
  Previous versions may also work, but haven't been tested.
  bulk_create_returning method doesn't support .only() and .defer() filters for django before 1.10.
• pytz
• six
• typing
• psycopg2
• PostgreSQL 9.4+
  Previous versions may also work, but haven't been tested.

Installation

Install via pip:
pip install django-pg-returning
or via setup.py:
python setup.py install

Usage

Integration

The easiest way to integrate, is to inherit your model from UpdateReturningModel instead of django.db.models.Model. It already has redeclared Manager, supporting returning operations.

from django.db import models
from django_pg_returning import UpdateReturningModel

class MyModel(UpdateReturningModel):   
    field = models.IntegerField()

If you already have custom manager, you can implement get_queryset() method in it:

from django.db import models
from django_pg_returning import UpdateReturningQuerySet, UpdateReturningModel

class MyManager(models.Manager):
    def get_queryset(self):
        return UpdateReturningQuerySet(using=self.db, model=self.model)

class MyModel(UpdateReturningModel):
    objects = MyManager()

    field = models.IntegerField()

And if you have custom manager you can use a mixin:

from django.db import models
from django_pg_returning import UpdateReturningMixin, UpdateReturningModel

class MyQuerySet(models.QuerySet, UpdateReturningMixin):
    pass

class MyManager(models.Manager):
    def get_queryset(self):
        return MyQuerySet(using=self.db, model=self.model)

class MyModel(UpdateReturningModel):
    objects = MyManager()

    field = models.IntegerField()

Methods

QuerySet methods

After QuerySet mixin is integrated with your model, your QuerySet-s will have 3 additional methods:

# Any django queryset you like
qs = MyModel.objects.all()

# Update and return a ReturningQuerySet, described below
result = qs.update_returning(field=1)

# Delete data and return a ReturningQuerySet, described below
result = qs.delete_returning()

# Acts like django's QuerySet.bulk_create() method, but updates all model fields stored in database
# Can be used to retrieve values, saved by database default/triggers etc.
result = MyModel.objects.bulk_create_returning([MyModel(field=1)])

By default methods get all fields, fetched by the model. To limit fields returned, you can use standard QuerySet.only() and QuerySet.defer() methods. bulk_create_returning doesn't support this methods for django before 1.10.

Model methods

If model instance is created, basic save() method is called.
If model is updated, database record is updated, and saved fields are refreshed with database values. This may be useful, if you update fields with F() expressions. By default all fields are saved and refreshed. Use update_fields to specify concrete fields to save and refresh.

instance = MyModel.objects.create(pk=1, field=1)
instance.field = F('field') + 1

# Basic save method will not change field and you don't know, what value is in database
instance.save()
print(instance.field)
# Output: F('field') + Value(1)

# Library method gives ability to fetch updated result 
instance.save_returning()
print(instance.field)
# Output: 2

Important notes:

1. If you don't fetch field, and then try to get it, library acts as django does - makes extra database query to fetch attribute deferred.
2. These queries are not lazy, as well as basic QuerySet.update() and QuerySet.delete() methods.
3. Primary key field is fetched not looking at limiting methods, as django needs it to form a QuerySet

ReturningQuerySet

The result of returning functions is django_pg_returning.ReturningQuerySet. It is based on django's RawQuerySet, but adds some extra methods to be used easier. The main difference is that ReturningQuerySet caches query results, while RawQuerySet executes query each time it is iterated. All ReturningQuerySet methods are not executed on database side, they are executed in python on cached result. The only way, ReturningQuerySet makes extra database query - is deferred field loading, described above. Implemented methods:

# UPDATE ... RETURNING query is executed here once. The result is cached.
result = MyModel.objects.all().update_returning(field=1)

# Get number of values fetched
print(result.count(), len(result))
# Output: 1, 1

# Index and slicing. Note that the order of result is not guaranteed by the database.
print(result[1], result[0:2])
# Output: MyModel(...), [MyModel(...), MyModel(...), MyModel(...)]

# Sintax sugar for indexing
print(result.first(), result.last())
# Output: MyModel(...), MyModel(...)

# Fetching values and values_list. Both methods use cache and return lists, not ValuesQuerySet like django does.
# values() method cakked without fields will return all fields, fetched in returning method.
# values_list() method called without fields will raise exception, as order or fields in result tuple is not obvious.

print(result.values())
# Output: [{'id': 1, 'field': 1}, {'id': 2, 'field': 2}]

print(result.values('field'))
# Output: [{'field': 1}, {'field': 2}]

print(result.values_list('field', flat=True))
# Output: [1, 2]

print(result.values_list('field', 'id', named=True))
# Output: [Row(field=1, id=1), Row(field=2, id=2)]