Skip to content

A simple django tool for generating random/customized test data for django model descriptions!

License

Notifications You must be signed in to change notification settings

mostafa-mahmoud/djenerator

Repository files navigation

Djenerator

Djenerator is a simple tool made to consistently generate test, random or dummy data from the model descriptions of Django. Djenerator has several advantages over other alternatives:

  1. Entry level is extremely easy, just install the package and include it in your project, there's no requirement by the user to specify the random data factory (it is optional).
  2. It generates data for dependent models (based on related fields), for example, if a model A has a ForeignKey to model B, then djenerator will recognize that generating data for model A requires having some instances of model B, and will generate them when necessary.
  3. Generators can simultaneously satisfy many constraints, like the unique flag, null flag, unique_together constraints, or django field validators.
  4. Easy to extend to include your own values for some of the fields.
  5. It can generate a big dump for a database at once, not only individual models.

Installation

  1. Using pip, you can install djenerator using this command:
$ pip3 install djenerator
  1. Add 'djenerator' to your INSTALLED_APPS:
INSTALLED_APPS = [
   'django.contrib.admin',
   'django.contrib.auth',
   'django.contrib.contenttypes',
   'django.contrib.sessions',
   # ...
   'djenerator',
   # ...
]

Usage

The generation of data can be done by a command using the manage.py file in your project:

$ python3 manage.py jenerate app_name size

Equivalently, this can be done within python code

from djenerator import generate_test_data
generate_test_data(app_name, size)

To generate for specific models

$ python3 manage.py jenerate app_name size --models ModelA ModelB ...

Equivalently, this can be done within python code

from djenerator import generate_test_data
generate_test_data(app_name, size, models_cls=["ModelA", "ModelB"])

To allow some null values

By default, djenerator generates data for all fields even if null values are allowed. To allow some null values, allow the following option:

$ python3 manage.py jenerate app_name size --allow-null

Equivalently, this can be done within python code

from djenerator import generate_test_data
generate_test_data(app_name, size, allow_null=True)

Writing your custom generators

You can add a customized values generator for a some fields in some models. This will most likely be required if you are writing your own validators. Writing your customized generator, can be done by adding a class to the module {app_name}.test_data:

  1. Write a class with the same name as the model.
  2. For the fields to write a custom generator, write an attirbute with a matching name, with the generator as a value.

The generator is either a function with no any required arguments that generates random values, or can also be an iterable of all possible values.

As an example, in an app testapp, if testapp/models.py is:

from django.core.exceptions import ValidationError
from django.db import models
from django.utils.translation import gettext_lazy as _


def validate_mod91(value):
    if value % 91 != 0:
        raise ValidationError(
            _('%(value)s is not an even number'), params={'value': value},
        )


class TestModel(models.Model):
    field1 = models.CharField(max_length=256)
    field2 = models.BigIntegerField(validators=[validate_mod91])
    field3 = models.IntegerField()

    class Meta:
        unique_together = (('field1', 'field2'),)

then testapp/test_data.py will contain the following:

import random

class TestModel:
    field2 = lambda: random.randint(1, 1000) * 91
    field3 = list(range(10000))

The value generator does not necessarily need to always generate unique or valid values, however, it should generate them with high probability. In the example above, the validate_mod91 checks if a number is divisible by 91, two bad generators can be:

  1. lambda: random.randint(0, 100000) will generate valid values but with only 1% chance; however, a chance higher than 20% or even 50% would be much better.
  2. lambda: random.randint(0, 10) * 91 will generate only 11 unique valid values; however, it is recommended to return a factor higher than the total number of models to be generated (especially if there are unique or many unique_together constraints).

Running the tests

Run the tests by running the command:

$ python3 manage.py test

The following combinations are tested:

Django Python Status
1.10.8 3.5
1.11.29 3.5
1.11.29 3.6
1.11.29 3.7
2.2.28 3.5
2.2.28 3.6
2.2.28 3.7
3.2.15 3.6
3.2.15 3.7
4.0.7 3.8
4.0.7 3.9
4.1.1 3.8
4.1.1 3.9

Requirements

  1. django >= 1.10.
  2. pytz is required to be manually installed for django < 1.11, otherwise it is installed by django when it is required (it is not required for some higher versions of django).
  3. pillow if ImageFields are used, we don't require it be default, but django will.

Our setup requires only django, other packages are reported by django.

TODOs and BUGS

See: https://github.com/mostafa-mahmoud/djenerator/issues