Skip to content

Latest commit

 

History

History
52 lines (33 loc) · 2.28 KB

README.md

File metadata and controls

52 lines (33 loc) · 2.28 KB
Raw

Pomade - making (async) Dapper code testable

Build status

About

Dapper is a great "mini ORM" library, but it's implemented almost entirely with extension methods which makes calling code very difficult to test. Pomade helps by forwarding calls to Dapper through virtual methods.

Also it helps by only implementing the async methods. *cough* 😇

Explanation

Dapper extends the IDbConnection interface with methods like QueryAsync:

IDbConnection _connection = new SqlConnection();

public Task<Dog> LookupDog(Guid guid)
    => _connection.QueryAsync<Dog>("select Age = @Age, Id = @Id", new { Age = (int?)null, Id = guid });

Unfortunately, this makes LookupDog() difficult to test as _connection is not easily replaced by most mocking frameworks.

Pomade helps get around this limitation by instead making those extension methods virtual members, and thus, mockable. To use, pass your IDbConnection reference to the Pomade class constructor and use that as your data access layer:

Pomade _connection = new Pomade(new SqlConnection());

public Task<Dog> LookupDog(Guid guid)
    => _connection.QueryAsync<Dog>("select Age = @Age, Id = @Id", new { Age = (int?)null, Id = guid });

Other than the type of the _connection field, the important stuff is exactly the same. The real difference comes in your tests. Since Pomade.QueryAsync<T> is a virtual method, you can inject Pomade instances into your tests and once again enjoy the ability to easily mock the database layer away.

Test example using NSubstitute:

var db = Substitute.ForPartsOf<Pomade>();
db.QueryAsync<Dog>(Arg.Any<string>(), Arg.Any<object>())
    .Returns(new Dog("Fido"));

Performance

"Dapper is stupid fast, aren't you just slowing it down by making all the calls virtual?"

Nope.

Method Mean Error StdDev Scaled
Dapper 216.1 us 0.7570 us 0.5910 us 1.00
Pomade 216.6 us 1.8440 us 1.5398 us 1.00