Slide 1
Slide 1 text
@cmaiacd
factory_boy
DjangoCon US - Oct 2022
Camila Maia
testing like a pro
Slide 2
Slide 2 text
@cmaiacd
You can ο¬nd this presentation at:
speakerdeck.com/cmaiacd
Slide 3
Slide 3 text
@cmaiacd
Who am I?
Slide 4
Slide 4 text
@cmaiacd
Brazilian living in Berlin
Bachelor of Computer Information System
Coding since 2010 π΅
Python π and Ruby π
Open Source and Community β€
Conferences
Backend Developer @ SoundCloud
Slide 5
Slide 5 text
@cmaiacd
Creator of ScanAPI
Slide 6
Slide 6 text
@cmaiacd
First GitHub proο¬le to be accepted for
the GitHub Sponsors program in Brazil
σ°
Quem sou eu?
Slide 7
Slide 7 text
@cmaiacd
factory_boy: what is it?
Slide 8
Slide 8 text
@cmaiacd
β It is a ο¬xtures replacement
β Based on factory_bot (Thoughtbot)
β First version: Django only
β Nowadays: framework-independent
β Unittest, Pytest...
Slide 9
Slide 9 text
@cmaiacd
For complex objects:
β Fixtures: static, hard to maintain
β
Factories: easy-to-use
Slide 10
Slide 10 text
@cmaiacd
β Fixtures:
exhaustive test setup with every possible combination
of corner cases
β
Factories:
objects customized for the current test, while only
declaring the test-speciο¬c ο¬elds
Slide 11
Slide 11 text
@cmaiacd
It offers many tools that help with the creation of the tests:
β Sequence
β Faker
β Fuzzy attributes
β LazyFunction
β LazyAttribute
β Inheritance
β Params
β Traits
β Strategies
β RelatedFactory/SubFactory
Slide 12
Slide 12 text
@cmaiacd
Demo App
Slide 13
Slide 13 text
@cmaiacd
Polls
Slide 14
Slide 14 text
@cmaiacd
Poll
Slide 15
Slide 15 text
@cmaiacd
Results
Slide 16
Slide 16 text
@cmaiacd
Model - Poll
Poll Question Choice
n:1
pub_date : DateTimeField
premium : BooleanField
author : CharField
question_text : CharField
language : CharField
choice_text : CharField
votes : IntegerField
1:1
Slide 17
Slide 17 text
@cmaiacd
Create Question - Admin
Slide 18
Slide 18 text
@cmaiacd
Create Question - Admin
Slide 19
Slide 19 text
@cmaiacd
Questions - Admin
Slide 20
Slide 20 text
@cmaiacd
Model - Poll
Slide 21
Slide 21 text
@cmaiacd
Model - Poll
Slide 22
Slide 22 text
@cmaiacd
Model - Question
Slide 23
Slide 23 text
@cmaiacd
Model - Choice
Slide 24
Slide 24 text
@cmaiacd
Best Practices
Slide 25
Slide 25 text
@cmaiacd
β +3 years
β Django Monolith
β +230 tables
β +2200 relevant ο¬les
β +75k relevant lines
My experience
Slide 26
Slide 26 text
@cmaiacd
Bad factories are like viruses
β Factories can get too tied
β Implicit errors might happen
β Factory usage involves a lot of copy/paste with small changes
β A poorly designed factory might affect many tests
β The tests are created in a way to fit the factory (factory-oriented
testing)
β Developers bump into the same issue again and again
β The same hack has to be done several times
Slide 27
Slide 27 text
@cmaiacd
Bad factories are like viruses
Slide 28
Slide 28 text
@cmaiacd
1. FACTORIES SHOULD REPRESENT THEIR
MODELS
Slide 29
Slide 29 text
@cmaiacd
BAD β
Slide 30
Slide 30 text
@cmaiacd
GOOD β
Slide 31
Slide 31 text
@cmaiacd
2. DO NOT RELY ON DEFAULTS FROM
FACTORIES
Slide 32
Slide 32 text
@cmaiacd
β If a default value is changed, all tests that
depend on it will break
β The setup of a test should contain all the logic
to ensure it will always pass
β Explicit better than implicit
Slide 33
Slide 33 text
@cmaiacd
BAD β
Slide 34
Slide 34 text
@cmaiacd
GOOD β
Slide 35
Slide 35 text
@cmaiacd
3. FACTORIES SHOULD CONTAIN ONLY THE
REQUIRED DATA
Slide 36
Slide 36 text
@cmaiacd
If the ο¬eld is nullable (null=True) the
attribute should be under a trait and not as a
default value
Slide 37
Slide 37 text
@cmaiacd
BAD β
Slide 38
Slide 38 text
@cmaiacd
GOOD β
Slide 39
Slide 39 text
@cmaiacd
β If we want to have an author, we can use
PollFactory(with_author=True) now
β When are we going to remember to test the case
PollFactory(author=None)?
β We should not assume there is an author when
DB actually allows to not have it.
Slide 40
Slide 40 text
@cmaiacd
4. BUILD OVER CREATE
Slide 41
Slide 41 text
@cmaiacd
β MyFactory.build()
creates a local object (memory)
β MyFactory.create()
creates a local object + stores it in the DB
Slide 42
Slide 42 text
@cmaiacd
BAD β
Slide 43
Slide 43 text
@cmaiacd
GOOD β
Slide 44
Slide 44 text
@cmaiacd
BUILD STRATEGY
======================== 14 passed in 1.76 seconds =========================
CREATE STRATEGY
======================== 14 passed in 3.26 seconds =========================
Slide 45
Slide 45 text
@cmaiacd
5.
IF FK IS IN THE TABLE:
SUBFACTOR
IF FK IS IN THE OTHER TABLE:
RELATEDFACTORY + TRAIT
Slide 46
Slide 46 text
@cmaiacd
β SubFactory: builds/creates the
SubFactory during the process of
creation of the main factory
β RelatedFactory: builds/creates the
RelatedFactory after creating the main
factory
Slide 47
Slide 47 text
@cmaiacd
Good β
Slide 48
Slide 48 text
@cmaiacd
6. USE FIXTURES TO WRAP FACTORIES TO
AVOID DUPLICATION
Slide 49
Slide 49 text
@cmaiacd
BAD β
Slide 50
Slide 50 text
@cmaiacd
GOOD β
Slide 51
Slide 51 text
@cmaiacd
7. AVOID SHARING FACTORIES OR
FIXTURES AMONG DIFFERENT FILES
Slide 52
Slide 52 text
@cmaiacd
β Many tests depending on the same
factory/fixture
β Tends to inflate the factory/fixture
β Hard to maintain
β Change a factory/fixture => tons of tests
breaking
β Fixture/Factory-oriented testing
Slide 53
Slide 53 text
@cmaiacd
CΓ³digo
Code
Slide 54
Slide 54 text
@cmaiacd
1. Factories should represent their models
2. Do not rely on defaults from factories
3. Factories should contain only the required data
4. Build over create
5. If FK is in the table: SubFactor
If FK is in the other table: RelatedFactory + trait
6. Avoid sharing factories or ο¬xtures among different ο¬les
7. Use ο¬xtures to wrap factories to avoid duplication
Review
Slide 55
Slide 55 text
@cmaiacd
Documentation
https://factoryboy.readthedocs.io/en/stable/
Common recipes
https://factoryboy.readthedocs.io/en/stable/recipes.html
Code
https://github.com/FactoryBoy/factory_boy
Links
Slide 56
Slide 56 text
@cmaiacd
THANK YOU!
@cmaiacd camilamaia
σ°
cmaiacd.com