FormStorm (v1.0)

FormStorm is a Python library that easily creates unit tests for Django forms. Given valid and invalid values for each field, the library computes every combination of the fields' values and submits them to the form. This validates all of the fields and tests for unintended interdependence between fields. In addition to testing single- and multi-field validation, FormStorm can also test single-field uniqueness constraints by double-submitting a valid submission and checking which fields become invalid on the 2nd submission.

Example:

Suppose we have a form to create a Book object. The book's name is mandatory, but the subtitle is optional. A FormTest is created that provides examples of valid and invalid values for each field:

from django.forms import ModelForm
from formstorm import FormTest, FormElement
from django.test import TestCase


class Book(models.Model):
    title = models.CharField(max_length=100, blank=False, null=False)
    subtitle = models.CharField(max_length=100, blank=True, default="")


class BookForm(ModelForm):
    class Meta:
        model = Book
        exclude = []


class BookFormTest(FormTest):
	form = BookForm
	title = FormElement(
		good = ["Moby Dick"],
		bad = [None, "", "A"*101],
	)
	subtitle = FormElement(
		good = [None, "", "or The Whale"],
		bad = ["A"*101]
	)


class BookTestCase(TestCase):
    def setUp(self):
        self.theBookFormTest = BookFormTest()

    def test_book_form(self):
        self.theBookFormTest.run()

When the FormTest runs, the form will be tested with every combination of each field's possible values. Namely, the form will be tested with these values:

title	subtitle	result
Moby Dick	""	Valid
Moby Dick	None	Valid
None	None	Invalid
""	None	Invalid
AA[...]AA	None	Invalid
None	""	Invalid
""	""	Invalid
AA[...]AA	""	Invalid
Moby Dick	or The Whale	Valid
None	or The Whale	Invalid
""	or The Whale	Invalid
AA[...]AA	or The Whale	Invalid
Moby Dick	AA[...]AA	Invalid
None	AA[...]AA	Invalid
""	AA[...]AA	Invalid
AA[...]AA	AA[...]AA	Invalid

Without something like FormStorm, you either have to tediously create test cases for each possible input value, or you have to just trust that the form behaves how you intend it to.

(A runnable implementation of the example above can be found in tests/minimalapp/.)

Advanced Example:

An example showing how to use different field types can be found in tests/fstestapp/tests.py.

Basically, all fields work as above, with the exception of ForeignKey and Many2Many fields whose values must be specified with Q() objects. Also, example values for multi-valued fields (such as Many2Many) can be created with the every_combo() function which returns every combination of the Many2Many options.

Validating multi-field constraints can be tested by specifying the values (as a dictionary) along with the expected results. For example, if the "title" and "subtitle" fields can't have a combined length greater than 150 characters, we can test this constraint like so:

additional_values = [
    ({'title': "A"*100, 'subtitle': "A"*50}, True),
    ({'title': "A"*50, 'subtitle': "A"*100}, True),
    ({'title': "A"*100, 'subtitle': "A"*51}, False),
    ({'title': "A"*51, 'subtitle': "A"*100}, False),
]

Install:

pip install formstorm

TODO:

• End-to-end testing (with Selenium): This is partially implemented, and most of the necessary FormStorm functions have been abstracted. Just need to subclass FormTest and fully implement.
• Tests for DRF Serializers. "SerializerStorm"
• Handle the obscure, "long tail" cases by implementing a framework to define tests for any type of constraint (such as multi-column uniqueness constraints). To do this, a "sub-test" would define one or more form submissions and the (boolean) result expected. Sub-tests would be combined with each other and with the standard combinatorial mix of good/bad values so that all fields are tested simultaneously.

A tentative definition of the sub-tests is below:

sub_tests = [
    { # Sub-test 1
        field_names=["field1","field2",..."fieldN"],
        submissions = [ Each sub-test consists of multilpe submissions.
            (  # Submission 1 
                value1,  # the value for field1
                value2,  # the value for field2
                ...
                valueN,  # the value for fieldN
                result,  # the expected result of the submission
            ),
            (...), # Submission 2
            ...
            (...)  # Submission N
        ]
    },
    {...}, # Sub-test 2
    ...
    {...}  # Sub-test N
]

For example, suppose a model has two fields that have a multi-column uniqueness constraint:

class SomeModel(models.Model):
    field1 = models.TextField()
    field2 = models.TextField()
    field3 = models.TextField()
    ...
    fieldN = models.TextField()

    class Meta:
        constraints = [
            UniqueConstraint(
                fields=['field1', 'field2'], name='unique_together'
            )
        ] 

The sub-test to test this constraint would be defined like this:

class SomeModelFormTest(FormTest):
	form = SomeModelForm
    sub_tests = [
        {
            field_names=["field1","field2"],
            submissions = [
                ("a","a", True),
                ("a","b", True),  # Duplicate values in one column are fine
                ("b","a", True),  # ... as are duplicates in the other column
                ("a","a", False)  # ... but the same values again should fail
            ]
        }
    ]
    field3 = FormElement(good=[...], bad=[...])
    ...
    fieldN = FormElement(good=[...], bad=[...])

The advantage of this is that we can define a test only for the fields affected by a constraint, and have values for the other fields supplied by the normal good/bad value tests.