With repository pattern, should I check first if the object queried exists in memory before it makes an actual database call - repository

Question

I have a repository, for example, UserRepository. It returns a user by given userId. I work on web application, so objects are loaded into memory, used, and disposed when the request ends.
So far, when I write a repository, I simply retrieved the data from the database. I don't store the retrieved User object into memory (I mean in a collection of the repository). When the repository's GetById() method is called, I don't check if the object is already in the collection. I simply query the database.
My questions are
Should I store retrieved objects in the memory, and when a repository's Get method is called, should I check if the object exists in the memory first before I make any Database call?
Or is the memory collection unnecessary, as web request is a short-lived session and all objects are disposed afterward

Answer 1

1) Should I store retrieved objects in the memory, and when a repository's Get method is called, should I check if the object exists in the memory first before I make any Database call?
Since your repository should be abstracted enough to simulate the purpose of an in-memory collection, I think it is really up to you and to your use case.
If you store your object after being retrieved from the database you will probably end-up with an implementation of the so-called IdentityMap. If you do this, it can get very complicated (well it depends on your domain).
Depending on the infrastructure layer you rely on, you may use the IdentityMap provided by your ORM if any.
But the real question is, is it worth implementing an IdentityMap?
I mean, we agree that repeating a query may be wrong for two reasons, performance and integrity, here a quote of Martin Fowler:
An old proverb says that a man with two watches never knows what time it is. If two watches are confusing, you can get in an even bigger mess with loading objects from a database.
But sometimes you need to be pragmatic and just load them every time you need it.
2) Or is the memory collection unnecessary, as web request is a short-lived session and all objects are disposed afterward
It depends™, for example, in some case you may have to play with your object in different place, in that case, it may be worth, but let's say you need to refresh your user session identity by loading your user from database, then there are cases where you only do it once within the whole request.

Answer 2

As is the usual case I don't think there is going to be a "one-size-fits-all".
There may be situations where one may implement a form of caching on a repository when the data is retrieved often, does not go stale too quickly, or simply for efficiency.
However, you could very well implement a type of generic cache decorator that can wrap a repository when you do need this.
So one should take each use case on merit.

Answer 3

When you're using an ORM like Entity Framework or NHibernate it's already taken care of - all read entities are tracked via IdentityMap mechanism, searching by keys (DbSet.Find in EF) won't even hit the database if the entity is already loaded.
If you're using direct database access or a microORM as base for your repository, you should be careful - without IdentityMap you're essentially working with value objects:
using System;
using System.Collections.Generic;
using System.Linq;
namespace test
{
internal class Program
{
static void Main()
{
Console.WriteLine("Identity map");
var artrepo1 = new ArticleIMRepository();
var o1 = new Order();
o1.OrderLines.Add(new OrderLine {Article = artrepo1.GetById(1, "a1", 100), Quantity = 50});
o1.OrderLines.Add(new OrderLine {Article = artrepo1.GetById(1, "a1", 100), Quantity = 30});
o1.OrderLines.Add(new OrderLine {Article = artrepo1.GetById(2, "a2", 100), Quantity = 20});
o1.ConfirmOrder();
o1.PrintChangedStock();
/*
Art. 1/a1, Stock: 20
Art. 2/a2, Stock: 80
*/
Console.WriteLine("Value objects");
var artrepo2 = new ArticleVORepository();
var o2 = new Order();
o2.OrderLines.Add(new OrderLine {Article = artrepo2.GetById(1, "a1", 100), Quantity = 50});
o2.OrderLines.Add(new OrderLine {Article = artrepo2.GetById(1, "a1", 100), Quantity = 30});
o2.OrderLines.Add(new OrderLine {Article = artrepo2.GetById(2, "a2", 100), Quantity = 20});
o2.ConfirmOrder();
o2.PrintChangedStock();
/*
Art. 1/a1, Stock: 50
Art. 1/a1, Stock: 70
Art. 2/a2, Stock: 80
*/
Console.ReadLine();
}
#region "Domain Model"
public class Order
{
public List<OrderLine> OrderLines = new List<OrderLine>();
public void ConfirmOrder()
{
foreach (OrderLine line in OrderLines)
{
line.Article.Stock -= line.Quantity;
}
}
public void PrintChangedStock()
{
foreach (var a in OrderLines.Select(x => x.Article).Distinct())
{
Console.WriteLine("Art. {0}/{1}, Stock: {2}", a.Id, a.Name, a.Stock);
}
}
}
public class OrderLine
{
public Article Article;
public int Quantity;
}
public class Article
{
public int Id;
public string Name;
public int Stock;
}
#endregion
#region Repositories
public class ArticleIMRepository
{
private static readonly Dictionary<int, Article> Articles = new Dictionary<int, Article>();
public Article GetById(int id, string name, int stock)
{
if (!Articles.ContainsKey(id))
Articles.Add(id, new Article {Id = id, Name = name, Stock = stock});
return Articles[id];
}
}
public class ArticleVORepository
{
public Article GetById(int id, string name, int stock)
{
return new Article {Id = id, Name = name, Stock = stock};
}
}
#endregion
}
}