LINQ to SQL Covariance - is this the right way to do this?

Question

I have several database - Elephants, Giraffes, Gorillas, etc - each of which has an Inputs and Results table named ElephantInputs, ElephantResults, GiraffeInputs, GiraffeResults, GorillaInputs, GorillaResults, respectively. I can't control the table naming.

I'm using LINQ to SQL to automatically generate the ElephantInputs, ElephantResults, GiraffeInputs, GiraffeResults, etc classes.

I want to write a single class, Processor(), that can read input from any of these databases and process it. I have a Factory method that instantiates a Processor based on the user's choices.

The first thing I did is create an interface for Input and Result objects and a partial class for each database so that its input and result objects implement that interface. That gave me a common interface for each animal. I also created a repository class for each database that has some methods to return items from the database.

public interface IBaseInput
{
    int ID { get; set; }
    string InputA { get; set; }
    int InputB { get; set; }
    double InputC { get; set; }
}

public interface IBaseResult
{
    int ID { get; set; }
    string ResultA { get; set; }
    int ResultB { get; set; }
    double ResultC { get; set; }
}

public interface IRepository<I, R>
    where I : IBaseInput
    where R : IBaseResult
{
    IQueryable<I> GetInputs();
    IQueryable<R> GetResults();
}

public partial class GorillaInput : IBaseInput
{
}

public partial class GorillaResult : IBaseResult
{
}

    // A concrete repository for Gorillas
    public class GorillaRepository : IRepository<GorillaInput, GorillaResult>
    {

    GorillaDataContext db = new GorillaDataContext(GetConnectionString("Gorillas"));

    public IQueryable<GorillaInput> GetInputs()
    {
        return from p in db.GorillaInputs select p;
    }

    public IQueryable<GorillaResult> GetResults()
    {
        return from r in db.GorillaResults select r;
    }
}

Next, I created an interface for my Processor

public interface IProcessor
{
    void Process();
}

Here's a concrete Processor, and the factory that creates it.

public class Processor<T, I, R> : IProcessor
    where T : class, IRepository<I, R>, new()
    where P : IBaseInput
    where R : IBaseResult
{

    public void Process()
    {
        // Do some stuff ...

        T repository = new T();  // Get results from the rater tables
        IEnumerable<I> inputs = repository.GetInputs();
        IEnumerable<R> results = repository.GetResults();

        // Do some stuff with inputs and outputs...
    }
}

public static class ProcessorFactory
{
    public static IProcessor GetProcessor(string animal)
    {
        switch (animal) {
            case "Elephant":
                return new Processor<ElephantRepository, ElephantInput, ElephantResult>();
            case "Giraffe":
                return new Processor<GiraffeRepository, GiraffeInput, GiraffeResult>();
            case "Gorilla":
                return new Processor<GorillaRepository, GorillaInput, GorillaResult>();
            default:
                return null;
        }
    }
}

Finally, here's the program that calls the processor:

class Program
{
    public static void Main()
    {
        IProcessor processor = ProcessorFactory.GetProcessor("Gorilla");
        processor.Process();
    }
}

Am I doing this right? Is there a less complicated way? Thanks

Answer 1

With all of your concrete types implementing interfaces, why bother with generics at all?

public class Processor
{

    public void Process(IRepository repository)
    {
        // Do some stuff ...

        IEnumerable<IBaseInput> inputs = repository.GetInputs();
        IEnumerable<IBaseResult> results = repository.GetResults();

        // Do some stuff with inputs and outputs...
    }
}