370 lines
17 KiB
Python
370 lines
17 KiB
Python
# coding: utf-8
|
|
from django.db import models
|
|
|
|
try:
|
|
sorted
|
|
except NameError:
|
|
from django.utils.itercompat import sorted # For Python 2.3
|
|
|
|
class Author(models.Model):
|
|
name = models.CharField(max_length=100)
|
|
age = models.IntegerField()
|
|
friends = models.ManyToManyField('self', blank=True)
|
|
|
|
def __unicode__(self):
|
|
return self.name
|
|
|
|
class Publisher(models.Model):
|
|
name = models.CharField(max_length=255)
|
|
num_awards = models.IntegerField()
|
|
|
|
def __unicode__(self):
|
|
return self.name
|
|
|
|
class Book(models.Model):
|
|
isbn = models.CharField(max_length=9)
|
|
name = models.CharField(max_length=255)
|
|
pages = models.IntegerField()
|
|
rating = models.FloatField()
|
|
price = models.DecimalField(decimal_places=2, max_digits=6)
|
|
authors = models.ManyToManyField(Author)
|
|
contact = models.ForeignKey(Author, related_name='book_contact_set')
|
|
publisher = models.ForeignKey(Publisher)
|
|
pubdate = models.DateField()
|
|
|
|
def __unicode__(self):
|
|
return self.name
|
|
|
|
class Store(models.Model):
|
|
name = models.CharField(max_length=255)
|
|
books = models.ManyToManyField(Book)
|
|
original_opening = models.DateTimeField()
|
|
friday_night_closing = models.TimeField()
|
|
|
|
def __unicode__(self):
|
|
return self.name
|
|
|
|
# Tests on 'aggregate'
|
|
# Different backends and numbers.
|
|
__test__ = {'API_TESTS': """
|
|
>>> from django.core import management
|
|
>>> try:
|
|
... from decimal import Decimal
|
|
... except:
|
|
... from django.utils._decimal import Decimal
|
|
>>> from datetime import date
|
|
|
|
# Reset the database representation of this app.
|
|
# This will return the database to a clean initial state.
|
|
>>> management.call_command('flush', verbosity=0, interactive=False)
|
|
|
|
# Empty Call - request nothing, get nothing.
|
|
>>> Author.objects.all().aggregate()
|
|
{}
|
|
|
|
>>> from django.db.models import Avg, Sum, Count, Max, Min
|
|
|
|
# Single model aggregation
|
|
#
|
|
|
|
# Single aggregate
|
|
# Average age of Authors
|
|
>>> Author.objects.all().aggregate(Avg('age'))
|
|
{'age__avg': 37.4...}
|
|
|
|
# Multiple aggregates
|
|
# Average and Sum of Author ages
|
|
>>> Author.objects.all().aggregate(Sum('age'), Avg('age'))
|
|
{'age__sum': 337, 'age__avg': 37.4...}
|
|
|
|
# Aggreates interact with filters, and only
|
|
# generate aggregate values for the filtered values
|
|
# Sum of the age of those older than 29 years old
|
|
>>> Author.objects.all().filter(age__gt=29).aggregate(Sum('age'))
|
|
{'age__sum': 254}
|
|
|
|
# Depth-1 Joins
|
|
#
|
|
|
|
# On Relationships with self
|
|
# Average age of the friends of each author
|
|
>>> Author.objects.all().aggregate(Avg('friends__age'))
|
|
{'friends__age__avg': 34.07...}
|
|
|
|
# On ManyToMany Relationships
|
|
#
|
|
|
|
# Forward
|
|
# Average age of the Authors of Books with a rating of less than 4.5
|
|
>>> Book.objects.all().filter(rating__lt=4.5).aggregate(Avg('authors__age'))
|
|
{'authors__age__avg': 38.2...}
|
|
|
|
# Backward
|
|
# Average rating of the Books whose Author's name contains the letter 'a'
|
|
>>> Author.objects.all().filter(name__contains='a').aggregate(Avg('book__rating'))
|
|
{'book__rating__avg': 4.0}
|
|
|
|
# On OneToMany Relationships
|
|
#
|
|
|
|
# Forward
|
|
# Sum of the number of awards of each Book's Publisher
|
|
>>> Book.objects.all().aggregate(Sum('publisher__num_awards'))
|
|
{'publisher__num_awards__sum': 30}
|
|
|
|
# Backward
|
|
# Sum of the price of every Book that has a Publisher
|
|
>>> Publisher.objects.all().aggregate(Sum('book__price'))
|
|
{'book__price__sum': Decimal("270.27")}
|
|
|
|
# Multiple Joins
|
|
#
|
|
|
|
# Forward
|
|
>>> Store.objects.all().aggregate(Max('books__authors__age'))
|
|
{'books__authors__age__max': 57}
|
|
|
|
# Backward
|
|
# Note that the very long default alias may be truncated
|
|
>>> Author.objects.all().aggregate(Min('book__publisher__num_awards'))
|
|
{'book__publisher__num_award...': 1}
|
|
|
|
# Aggregate outputs can also be aliased.
|
|
|
|
# Average amazon.com Book rating
|
|
>>> Store.objects.filter(name='Amazon.com').aggregate(amazon_mean=Avg('books__rating'))
|
|
{'amazon_mean': 4.08...}
|
|
|
|
# Tests on annotate()
|
|
|
|
# An empty annotate call does nothing but return the same QuerySet
|
|
>>> Book.objects.all().annotate().order_by('pk')
|
|
[<Book: The Definitive Guide to Django: Web Development Done Right>, <Book: Sams Teach Yourself Django in 24 Hours>, <Book: Practical Django Projects>, <Book: Python Web Development with Django>, <Book: Artificial Intelligence: A Modern Approach>, <Book: Paradigms of Artificial Intelligence Programming: Case Studies in Common Lisp>]
|
|
|
|
# Annotate inserts the alias into the model object with the aggregated result
|
|
>>> books = Book.objects.all().annotate(mean_age=Avg('authors__age'))
|
|
>>> books.get(pk=1).name
|
|
u'The Definitive Guide to Django: Web Development Done Right'
|
|
|
|
>>> books.get(pk=1).mean_age
|
|
34.5
|
|
|
|
# On ManyToMany Relationships
|
|
|
|
# Forward
|
|
# Average age of the Authors of each book with a rating less than 4.5
|
|
>>> books = Book.objects.all().filter(rating__lt=4.5).annotate(Avg('authors__age'))
|
|
>>> sorted([(b.name, b.authors__age__avg) for b in books])
|
|
[(u'Artificial Intelligence: A Modern Approach', 51.5), (u'Practical Django Projects', 29.0), (u'Python Web Development with Django', 30.3...), (u'Sams Teach Yourself Django in 24 Hours', 45.0)]
|
|
|
|
# Count the number of authors of each book
|
|
>>> books = Book.objects.annotate(num_authors=Count('authors'))
|
|
>>> sorted([(b.name, b.num_authors) for b in books])
|
|
[(u'Artificial Intelligence: A Modern Approach', 2), (u'Paradigms of Artificial Intelligence Programming: Case Studies in Common Lisp', 1), (u'Practical Django Projects', 1), (u'Python Web Development with Django', 3), (u'Sams Teach Yourself Django in 24 Hours', 1), (u'The Definitive Guide to Django: Web Development Done Right', 2)]
|
|
|
|
# Backward
|
|
# Average rating of the Books whose Author's names contains the letter 'a'
|
|
>>> authors = Author.objects.all().filter(name__contains='a').annotate(Avg('book__rating'))
|
|
>>> sorted([(a.name, a.book__rating__avg) for a in authors])
|
|
[(u'Adrian Holovaty', 4.5), (u'Brad Dayley', 3.0), (u'Jacob Kaplan-Moss', 4.5), (u'James Bennett', 4.0), (u'Paul Bissex', 4.0), (u'Stuart Russell', 4.0)]
|
|
|
|
# Count the number of books written by each author
|
|
>>> authors = Author.objects.annotate(num_books=Count('book'))
|
|
>>> sorted([(a.name, a.num_books) for a in authors])
|
|
[(u'Adrian Holovaty', 1), (u'Brad Dayley', 1), (u'Jacob Kaplan-Moss', 1), (u'James Bennett', 1), (u'Jeffrey Forcier', 1), (u'Paul Bissex', 1), (u'Peter Norvig', 2), (u'Stuart Russell', 1), (u'Wesley J. Chun', 1)]
|
|
|
|
# On OneToMany Relationships
|
|
|
|
# Forward
|
|
# Annotate each book with the number of awards of each Book's Publisher
|
|
>>> books = Book.objects.all().annotate(Sum('publisher__num_awards'))
|
|
>>> sorted([(b.name, b.publisher__num_awards__sum) for b in books])
|
|
[(u'Artificial Intelligence: A Modern Approach', 7), (u'Paradigms of Artificial Intelligence Programming: Case Studies in Common Lisp', 9), (u'Practical Django Projects', 3), (u'Python Web Development with Django', 7), (u'Sams Teach Yourself Django in 24 Hours', 1), (u'The Definitive Guide to Django: Web Development Done Right', 3)]
|
|
|
|
# Backward
|
|
# Annotate each publisher with the sum of the price of all books sold
|
|
>>> publishers = Publisher.objects.all().annotate(Sum('book__price'))
|
|
>>> sorted([(p.name, p.book__price__sum) for p in publishers])
|
|
[(u'Apress', Decimal("59.69")), (u"Jonno's House of Books", None), (u'Morgan Kaufmann', Decimal("75.00")), (u'Prentice Hall', Decimal("112.49")), (u'Sams', Decimal("23.09"))]
|
|
|
|
# Calls to values() are not commutative over annotate().
|
|
|
|
# Calling values on a queryset that has annotations returns the output
|
|
# as a dictionary
|
|
>>> Book.objects.filter(pk=1).annotate(mean_age=Avg('authors__age')).values()
|
|
[{'rating': 4.5, 'isbn': u'159059725', 'name': u'The Definitive Guide to Django: Web Development Done Right', 'pubdate': datetime.date(2007, 12, 6), 'price': Decimal("30..."), 'contact_id': 1, 'id': 1, 'publisher_id': 1, 'pages': 447, 'mean_age': 34.5}]
|
|
|
|
>>> Book.objects.filter(pk=1).annotate(mean_age=Avg('authors__age')).values('pk', 'isbn', 'mean_age')
|
|
[{'pk': 1, 'isbn': u'159059725', 'mean_age': 34.5}]
|
|
|
|
# Calling values() with parameters reduces the output
|
|
>>> Book.objects.filter(pk=1).annotate(mean_age=Avg('authors__age')).values('name')
|
|
[{'name': u'The Definitive Guide to Django: Web Development Done Right'}]
|
|
|
|
# An empty values() call before annotating has the same effect as an
|
|
# empty values() call after annotating
|
|
>>> Book.objects.filter(pk=1).values().annotate(mean_age=Avg('authors__age'))
|
|
[{'rating': 4.5, 'isbn': u'159059725', 'name': u'The Definitive Guide to Django: Web Development Done Right', 'pubdate': datetime.date(2007, 12, 6), 'price': Decimal("30..."), 'contact_id': 1, 'id': 1, 'publisher_id': 1, 'pages': 447, 'mean_age': 34.5}]
|
|
|
|
# Calling annotate() on a ValuesQuerySet annotates over the groups of
|
|
# fields to be selected by the ValuesQuerySet.
|
|
|
|
# Note that an extra parameter is added to each dictionary. This
|
|
# parameter is a queryset representing the objects that have been
|
|
# grouped to generate the annotation
|
|
|
|
>>> Book.objects.all().values('rating').annotate(n_authors=Count('authors__id'), mean_age=Avg('authors__age')).order_by('rating')
|
|
[{'rating': 3.0, 'n_authors': 1, 'mean_age': 45.0}, {'rating': 4.0, 'n_authors': 6, 'mean_age': 37.1...}, {'rating': 4.5, 'n_authors': 2, 'mean_age': 34.5}, {'rating': 5.0, 'n_authors': 1, 'mean_age': 57.0}]
|
|
|
|
# If a join doesn't match any objects, an aggregate returns None
|
|
>>> authors = Author.objects.all().annotate(Avg('friends__age')).order_by('id')
|
|
>>> len(authors)
|
|
9
|
|
>>> sorted([(a.name, a.friends__age__avg) for a in authors])
|
|
[(u'Adrian Holovaty', 32.0), (u'Brad Dayley', None), (u'Jacob Kaplan-Moss', 29.5), (u'James Bennett', 34.0), (u'Jeffrey Forcier', 27.0), (u'Paul Bissex', 31.0), (u'Peter Norvig', 46.0), (u'Stuart Russell', 57.0), (u'Wesley J. Chun', 33.6...)]
|
|
|
|
|
|
# The Count aggregation function allows an extra parameter: distinct.
|
|
# This restricts the count results to unique items
|
|
>>> Book.objects.all().aggregate(Count('rating'))
|
|
{'rating__count': 6}
|
|
|
|
>>> Book.objects.all().aggregate(Count('rating', distinct=True))
|
|
{'rating__count': 4}
|
|
|
|
# Retreiving the grouped objects
|
|
|
|
# When using Count you can also omit the primary key and refer only to
|
|
# the related field name if you want to count all the related objects
|
|
# and not a specific column
|
|
>>> explicit = list(Author.objects.annotate(Count('book__id')))
|
|
>>> implicit = list(Author.objects.annotate(Count('book')))
|
|
>>> explicit == implicit
|
|
True
|
|
|
|
# Ordering is allowed on aggregates
|
|
>>> Book.objects.values('rating').annotate(oldest=Max('authors__age')).order_by('oldest', 'rating')
|
|
[{'rating': 4.5, 'oldest': 35}, {'rating': 3.0, 'oldest': 45}, {'rating': 4.0, 'oldest': 57}, {'rating': 5.0, 'oldest': 57}]
|
|
|
|
>>> Book.objects.values('rating').annotate(oldest=Max('authors__age')).order_by('-oldest', '-rating')
|
|
[{'rating': 5.0, 'oldest': 57}, {'rating': 4.0, 'oldest': 57}, {'rating': 3.0, 'oldest': 45}, {'rating': 4.5, 'oldest': 35}]
|
|
|
|
# It is possible to aggregate over anotated values
|
|
>>> Book.objects.all().annotate(num_authors=Count('authors__id')).aggregate(Avg('num_authors'))
|
|
{'num_authors__avg': 1.66...}
|
|
|
|
# You can filter the results based on the aggregation alias.
|
|
|
|
# Lets add a publisher to test the different possibilities for filtering
|
|
>>> p = Publisher(name='Expensive Publisher', num_awards=0)
|
|
>>> p.save()
|
|
>>> Book(name='ExpensiveBook1', pages=1, isbn='111', rating=3.5, price=Decimal("1000"), publisher=p, contact_id=1, pubdate=date(2008,12,1)).save()
|
|
>>> Book(name='ExpensiveBook2', pages=1, isbn='222', rating=4.0, price=Decimal("1000"), publisher=p, contact_id=1, pubdate=date(2008,12,2)).save()
|
|
>>> Book(name='ExpensiveBook3', pages=1, isbn='333', rating=4.5, price=Decimal("35"), publisher=p, contact_id=1, pubdate=date(2008,12,3)).save()
|
|
|
|
# Publishers that have:
|
|
|
|
# (i) more than one book
|
|
>>> Publisher.objects.annotate(num_books=Count('book__id')).filter(num_books__gt=1).order_by('pk')
|
|
[<Publisher: Apress>, <Publisher: Prentice Hall>, <Publisher: Expensive Publisher>]
|
|
|
|
# (ii) a book that cost less than 40
|
|
>>> Publisher.objects.filter(book__price__lt=Decimal("40.0")).order_by('pk')
|
|
[<Publisher: Apress>, <Publisher: Apress>, <Publisher: Sams>, <Publisher: Prentice Hall>, <Publisher: Expensive Publisher>]
|
|
|
|
# (iii) more than one book and (at least) a book that cost less than 40
|
|
>>> Publisher.objects.annotate(num_books=Count('book__id')).filter(num_books__gt=1, book__price__lt=Decimal("40.0")).order_by('pk')
|
|
[<Publisher: Apress>, <Publisher: Prentice Hall>, <Publisher: Expensive Publisher>]
|
|
|
|
# (iv) more than one book that costs less than $40
|
|
>>> Publisher.objects.filter(book__price__lt=Decimal("40.0")).annotate(num_books=Count('book__id')).filter(num_books__gt=1).order_by('pk')
|
|
[<Publisher: Apress>]
|
|
|
|
# Now a bit of testing on the different lookup types
|
|
#
|
|
|
|
>>> Publisher.objects.annotate(num_books=Count('book')).filter(num_books__range=[1, 3]).order_by('pk')
|
|
[<Publisher: Apress>, <Publisher: Sams>, <Publisher: Prentice Hall>, <Publisher: Morgan Kaufmann>, <Publisher: Expensive Publisher>]
|
|
|
|
>>> Publisher.objects.annotate(num_books=Count('book')).filter(num_books__range=[1, 2]).order_by('pk')
|
|
[<Publisher: Apress>, <Publisher: Sams>, <Publisher: Prentice Hall>, <Publisher: Morgan Kaufmann>]
|
|
|
|
>>> Publisher.objects.annotate(num_books=Count('book')).filter(num_books__in=[1, 3]).order_by('pk')
|
|
[<Publisher: Sams>, <Publisher: Morgan Kaufmann>, <Publisher: Expensive Publisher>]
|
|
|
|
>>> Publisher.objects.annotate(num_books=Count('book')).filter(num_books__isnull=True)
|
|
[]
|
|
|
|
>>> p.delete()
|
|
|
|
# Does Author X have any friends? (or better, how many friends does author X have)
|
|
>> Author.objects.filter(pk=1).aggregate(Count('friends__id'))
|
|
{'friends__id__count': 2.0}
|
|
|
|
# Give me a list of all Books with more than 1 authors
|
|
>>> Book.objects.all().annotate(num_authors=Count('authors__name')).filter(num_authors__ge=2).order_by('pk')
|
|
[<Book: The Definitive Guide to Django: Web Development Done Right>, <Book: Artificial Intelligence: A Modern Approach>]
|
|
|
|
# Give me a list of all Authors that have no friends
|
|
>>> Author.objects.all().annotate(num_friends=Count('friends__id', distinct=True)).filter(num_friends=0).order_by('pk')
|
|
[<Author: Brad Dayley>]
|
|
|
|
# Give me a list of all publishers that have published more than 1 books
|
|
>>> Publisher.objects.all().annotate(num_books=Count('book__id')).filter(num_books__gt=1).order_by('pk')
|
|
[<Publisher: Apress>, <Publisher: Prentice Hall>]
|
|
|
|
# Give me a list of all publishers that have published more than 1 books that cost less than 40
|
|
>>> Publisher.objects.all().filter(book__price__lt=Decimal("40.0")).annotate(num_books=Count('book__id')).filter(num_books__gt=1)
|
|
[<Publisher: Apress>]
|
|
|
|
# Give me a list of all Books that were written by X and one other author.
|
|
>>> Book.objects.all().annotate(num_authors=Count('authors__id')).filter(authors__name__contains='Norvig', num_authors__gt=1)
|
|
[<Book: Artificial Intelligence: A Modern Approach>]
|
|
|
|
# Give me the average rating of all Books that were written by X and one other author.
|
|
#(Aggregate over objects discovered using membership of the m2m set)
|
|
|
|
# Adding an existing author to another book to test it the right way
|
|
>>> a = Author.objects.get(name__contains='Norvig')
|
|
>>> b = Book.objects.get(name__contains='Done Right')
|
|
>>> b.authors.add(a)
|
|
>>> b.save()
|
|
|
|
# This should do it
|
|
>>> Book.objects.all().annotate(num_authors=Count('authors__id')).filter(authors__name__contains='Norvig', num_authors__gt=1).aggregate(Avg('rating'))
|
|
{'rating__avg': 4.25}
|
|
>>> b.authors.remove(a)
|
|
|
|
# Give me a list of all Authors that have published a book with at least one other person
|
|
# (Filters over a count generated on a related object)
|
|
#
|
|
# Cheating: [a for a in Author.objects.all().annotate(num_coleagues=Count('book__authors__id'), num_books=Count('book__id', distinct=True)) if a.num_coleagues - a.num_books > 0]
|
|
# F-Syntax is required. Will be fixed after F objects are available
|
|
|
|
# Aggregates also work on dates, times and datetimes
|
|
>>> Publisher.objects.annotate(earliest_book=Min('book__pubdate')).exclude(earliest_book=None).order_by('earliest_book').values()
|
|
[{'earliest_book': datetime.date(1991, 10, 15), 'num_awards': 9, 'id': 4, 'name': u'Morgan Kaufmann'}, {'earliest_book': datetime.date(1995, 1, 15), 'num_awards': 7, 'id': 3, 'name': u'Prentice Hall'}, {'earliest_book': datetime.date(2007, 12, 6), 'num_awards': 3, 'id': 1, 'name': u'Apress'}, {'earliest_book': datetime.date(2008, 3, 3), 'num_awards': 1, 'id': 2, 'name': u'Sams'}]
|
|
|
|
>>> Store.objects.aggregate(Max('friday_night_closing'), Min("original_opening"))
|
|
{'friday_night_closing__max': datetime.time(23, 59, 59), 'original_opening__min': datetime.datetime(1945, 4, 25, 16, 24, 14)}
|
|
|
|
# values_list() can also be used
|
|
|
|
>>> Book.objects.filter(pk=1).annotate(mean_age=Avg('authors__age')).values_list('pk', 'isbn', 'mean_age')
|
|
[(1, u'159059725', 34.5)]
|
|
|
|
>>> Book.objects.filter(pk=1).annotate(mean_age=Avg('authors__age')).values_list('isbn')
|
|
[(u'159059725',)]
|
|
|
|
>>> Book.objects.filter(pk=1).annotate(mean_age=Avg('authors__age')).values_list('mean_age')
|
|
[(34.5,)]
|
|
|
|
>>> Book.objects.filter(pk=1).annotate(mean_age=Avg('authors__age')).values_list('mean_age', flat=True)
|
|
[34.5]
|
|
|
|
>>> qs = Book.objects.values_list('price').annotate(count=Count('price')).order_by('-count', 'price')
|
|
>>> list(qs) == [(Decimal('29.69'), 2), (Decimal('23.09'), 1), (Decimal('30'), 1), (Decimal('75'), 1), (Decimal('82.8'), 1)]
|
|
True
|
|
|
|
"""}
|