Sean KilleenI was recently working with the excellent Octokit.NET library for working with GitHub, and like many developers, I needed to write a test.
I figured I’d give a quick run-down on the journey of that test, how I ultimately used the Bogus library to help wire up my fake, and then talk about some of the other great functionality of Bogus.
Mocking a Large Object
I have a function in my app called GetRepoQuotaFromOrg. Its purpose is to use the Octokit library and pull in the repository quota that the organization has. I wanted to write a test that proves my code is using the PrivateRepos field from the GitHub organization.
But, there’s a challenge here. I need my mock IOrganizationClient client to return an Organization object. But here’s the constructor for the Organization object:
That’s a lot of required fields. With it, my first passing test looked like the gnarly beast below:
[Fact]
public async Task GetsRepoLimitFromOrganizationPlan()
{
var privateRepoLimit = 123;
const int DUMMY_INT = 0;
const long DUMMY_LONG = 0;
var DUMMY_DATE = DateTimeOffset.MaxValue;
var plan = new Plan(DUMMY_LONG, null, privateRepoLimit, DUMMY_LONG, null);
var org = new Organization(null, null, null, DUMMY_INT,
null, DUMMY_DATE, DUMMY_INT, null, DUMMY_INT, DUMMY_INT,
null, null, DUMMY_INT, DUMMY_INT, null, null,
null, null, DUMMY_INT, plan, DUMMY_INT, DUMMY_INT,
DUMMY_INT, null, null);
// Reader note: This, and _sut (my "situation under test") below, are part of the constructor and not fully
// shown here because they are less relevant.
_mockOrgClient.Setup(x =>
x.Get(It.IsAny<string>()))
.Returns(Task.FromResult(org));
var result = await _sut.GetRepoQuotaForOrg();
result.PrivateRepoLimit.Should().Be(privateRepoLimit);
}
Gross. Dummy values all over the place, and tons of noise in my test.
I could extract these to a method that takes a parameter, but I knew that I’d be using this construct in other test classes as well, so I decided to utilize the builder pattern.
Introducing the Builder Pattern
I knew I would likely need default values for most items but the ability to override a select few fields. To be clear and intentional about this in my tests, I introduced an OrganizationBuilder to build an Organization object.
public class OrganizationBuilder
{
private long _privateRepoLimit = 0;
private int _numberOfOwnedPrivateRepos = 0;
public OrganizationBuilder WithPrivateRepoLimit(long limit)
{
_privateRepoLimit = limit;
return this;
}
public OrganizationBuilder WithPrivateRepoCount(int privateRepoCount)
{
_numberOfOwnedPrivateRepos = privateRepoCount;
return this;
}
public Organization Build()
{
const int DUMMY_INT = 0;
const long DUMMY_LONG = 0;
var DUMMY_DATE = DateTimeOffset.MaxValue;
var plan = new Plan(DUMMY_LONG, null, _privateRepoLimit, DUMMY_LONG, null);
var result = new Organization(null, null, null, DUMMY_INT,
null, DUMMY_DATE, DUMMY_INT, null, DUMMY_INT, DUMMY_INT,
null, null, DUMMY_INT, DUMMY_INT, null, null,
null, null, _numberOfOwnedPrivateRepos, plan, DUMMY_INT, DUMMY_INT,
DUMMY_INT, null, null);
return result;
}
}
Now my test looks much better:
[Fact]
public async Task GetsRepoLimitFromOrganizationPlan()
{
var privateRepoLimit = 123;
// Way less noise!
var org = new OrganizationBuilder().WithPrivateRepoLimit(123).Build();
// Reader note: This, and _sut (my "situation under test") below, are part of the constructor and not fully
// shown here because they are less relevant.
_mockOrgClient.Setup(x =>
x.Get(It.IsAny<string>()))
.Returns(Task.FromResult(org));
var result = await _sut.GetRepoQuotaForOrg();
result.PrivateRepoLimit.Should().Be(privateRepoLimit);
}
But there are a few things I don’t like still – namely, all those dummy values still in my builder object. I’d prefer to populate them with legitimate values each time to make things a little more realistic. Enter the Bogus library.
A Step Better: Using Bogus to Remove the Builder Pattern Cruft
The Bogus .NET library allows us to generate fake data in objects, while supplying the real data that we care about. In my builder, I can use a new Faker<Organization>(), set rules, and generate the object without needing to explicitly specify all those dummy values.
Now, my OrganizationBuilder class looks quite bit cleaner:
public class OrganizationBuilder
{
private readonly Faker<Organization> _faker = new Faker<Organization>();
private readonly Faker<Plan> _planFaker = new Faker<Plan>();
public OrganizationBuilder WithPrivateRepoLimit(long limit)
{
_planFaker.RuleFor(pl => pl.PrivateRepos, limit);
return this;
}
public OrganizationBuilder WithPrivateRepoCount(int privateRepoCount)
{
_faker.RuleFor(x => x.OwnedPrivateRepos, privateRepoCount);
return this;
}
public Organization Build()
{
// First generate the plan object
// If "WithPrivateRepoLimit" has been called, a rule will be in place; otherwise, dummy object.
var plan = _planFaker.Generate();
_faker.RuleFor(x => x.Plan, plan); // Add a rule to use the generated plan in the organization object
_faker.AssertConfigurationIsValid(); // For demo only; if strict mode is enabled, this will throw if we're missing a rule
return _faker.Generate(); // Generate the fake organization
}
}
Why Use Bogus?
Bogus certainly isn’t the only library we could use here – AutoMoq and AutoFixture both come to mind and the result so far could be achieved using any of them.
However, Bogus has some additional use cases that are worth considering, which is why I suggest using it instead. A few off the top of my head:
- It has built-in support for a huge number data types – People, Addresses, Latitude/Longitude, images, e-mail addresses, lorem ipsum, and many more. This makes generating compliant fake data a breeze.
- You can set a list of values and have Bogus pick one at random
- If you’d like, you can create side effects as the faker object is used. I’d use sparingly, but a good example is creating an ID and incrementing it for each use of the faker, so that you have a sequential ID list.
- Quickly generate random strings that require letters and numbers, e.g.
new Bogus.Randomizer().Repace("###-##-####")for an SSN.
What Data Do you Like to Fake?
I’d be interested to hear about what sorts of fake test data you’re working to generate. Sound off in the comments!
Until next time – happy faking!
