A small python / django model designed to decouple creation of models for testing from the creation of models in production to make updating tests less painful.
Create generic models without defining fields
class User(AbstractBaseUser):
username = models.CharField()
class UserBuilder(ModelBuilder):
model = User
def get_default_fields(self):
return {'username': 'test_username'}
user = UserBuilder().build()
print(user.username)
>>> test_usernameOverride defaults when required
user = UserBuilder().with_username('test').build()
>>> user.username
>>> testCreate multiple models with the same values
builder = UserBuilder().with_username('test')
user_1 = builder.build()
user_2 = builder.build()
user_1.username == user_2.username
>>> True
user_1 == user_2
>>> FalseUpdate models without updating tests
class User(AbstractBaseUser):
username = models.CharField()
dob = models.DateField()
class UserBuilder(ModelBuilder):
model = User
def get_default_fields(self):
return {
'username': random_string,
'dob': date(1990, 1, 1),
}
user = UserBuilder().build()
user.dob
>>> date(1990, 1, 1)
user = (
UserBuilder()
.with_dob(date(2000, 1, 1))
.build()
)
user.dob
>>> date(2000, 1, 1)Setting defaults
The get_default_fields returns a dictionary used to populate any unset
model fields when the model is created. These can be values or callables if you
need to delay the creation of models until it is needed or want to generate
random data for each instance to avoid breaking database constraints.
class UserBuilder(ModelBuilder):
model = User
def get_default_fields():
return {
# Callable, each user will have a random username.
'username': random_string,
# Value, each user will have the same date of birth.
'dob': date(1990, 1, 1),
# Called with uninitiated build() call so duplicate model isn't
# generated until comparison with any custom `with_` setter
# functions, this field will be thrown away
# if custom setter is present. You can also use a
# lambda to achieve the same thing.
'user': UserBuilder().build
}Providing custom values using the "with_" prefix
with_ functions are dynamically generated, these are used to override
defaults.
class UserBuilder(ModelBuilder):
model = User
def get_default_fields():
return {
'username': random_string,
'dob': date(1990, 1, 1),
}
user = UserBuilder().with_dob(date(2019, 10, 10)).build()
user.dob
>>> date(2019, 10, 10)All these functions do is set the passed in value as the function name in an internal dictionary. This pattern can be used to create more readable tests.
Any function prefixed with with_ is automatically wrapped with a function
that returns a copy of the builder for side-effect-free chaining.
You can also explicitly define these with_<> on the ModelBuilder subclass
to add your own implementation.
from datetime import timedelta
class UserBuilder(ModelBuilder):
model = User
def get_default_fields():
return {
'username': random_string,
'dob': date(1990, 1, 1)
}
def with_under_18():
self.data['dob'] = date.today() - timedelta(years=17)
UserBuilder().under_18().build()Finally the with_ prefix is adjustable in case you have a blocking field that
you want use. For example you can change this to use the prefix set_ by going
class CustomAuthorBuilder(AuthorBuilder):
dynamic_field_setter_prefix = 'set_'
author = (
CustomAuthorBuilder()
.set_publishing_name('Billy Fakeington')
.build()
)
author.publishing_name
>>> 'Billy Fakeington'Calling .build()
Building the model is broken into four steps.
- Prepare the data dictionary.
- Perform pre processing.
- Create the instance.
- Perform post possessing.
There is also a save_to_db kwarg that can be set to optionally persist the
built model to memory only for use in more complicated tests.
Perform pre processing
By default this method changes models to their their _id suffix. This can be
extended to perform additional preprocessing of fields.
from datetime import timedelta
class UserBuilder(ModelBuilder):
model = User
def get_default_fields():
return {
'username': random_string,
'dob': date(1990, 1, 1),
}
def pre(self):
self['dob'] += timedelta(days=1)
UserBuilder().build().dob
# date(1990, 1, 2)If you wanting to add non field values for accession by the pre/post hooks
you can override the get_builder_context call to load any extra fields
which will be made available to the self.data dict after the initial model
fields have been set, for instance:
class AuthorBuilder(ModelBuilder):
def get_default_fields():
return {
'username': random_string,
'dob': date(1990, 1, 1)
}
def get_builder_context(self):
return {
'email_address': fake_email
}
def post(self):
print(self.dict)
AuthorBuilder().build()
>>> {
>>> 'username': random_string,
>>> 'dob': date(1990, 1, 1),
>>> 'email_address': fake_email
>>> }Create the instance
By default instances are created by calling model.objects.create
with the models fields from the data dictionary. This behavior can be changed
by overriding the builders .create method, this method must set the builders
instance attribute`self.instance = ...`.
class UserBuilder(ModelBuilder):
model = User
def get_default_fields():
return {
'username': random_string,
}
def create(self):
model = self.get_model()
try:
instance = self.model.objects.get(
username=self.data['username']
)
except model.objects.DoesNotExist:
super(UserBuilder, self).create()
builder = UserBuilder().with_username('test')
user_1 = builder.build()
user_2 = builder.build()
user_1 == user_2
>>> TruePreform post processing
Post processing is carried out after the instance has been created. By default it does nothing, but provides a useful place to do things like add related models.
class UserBuilder(ModelBuilder):
model = User
def get_default_fields():
return {
'username': random_string,
}
def with_emails(*args):
self.data['emails'] = args
def post(self):
for email in self.data.get('emails', []):
(
EmailBuilder()
.with_address(email)
.with_user(self.instance)
.build()
)
user = (
UserBuilder()
.with_emails(random_email(), random_email())
.build()
)
user.email_set.count()
>>> 2