如何在方法中传递多个Lambda表达式

Question

Suppose I have a generic data access layer method for updating records with something like this code:

public virtual void Update<P>(Expression<Func<T, P>> excludeColumn, params T[] items)
{
        foreach (T item in items)
        {
            _entities.Entry(item).State = EntityState.Modified;
            _entities.Entry(item).Property(excludeColumn).IsModified = false;
        }

        _entities.SaveChanges();
}

Here I am taking excludeColumn param for excluding column from update, and I passed value into this parameter like this

_companyProfileRepository.Update(x => x.EmailAddress, records);

x => x.EmailAddress is an expression which I pass into the generic Update method. My problem is I want to pass multiple columns into Update method, because sometimes I need to exclude more than just one column, but my method doesn't support multiple column mechanism.

Can anyone help me figure this out?

Answer 1

For the multiple-column case, you could add an overload that accepts the names of unmodified columns. Something like:

    public virtual void Update<TEntity>(IEnumerable<string> excludeColumnNames, params TEntity[] items)
    {
        foreach (var item in items)
        {
            _entities.Entry(item).State = EntityState.Modified;
            foreach (var cn in excludeColumnNames)
            {
                _entities.Entry(item).Property(cn).IsModified = false;
            }

        }
        _entities.SaveChanges();
    }

Answer 2

Using the Repository Pattern you can wrap your EF data layer in something that actually has meaning, rather than generic/abstract code that conceals intention. There are tons of examples on the web of Repository Pattern implemented generically, but I consider this an anti-pattern, and I'm not alone .

eg a UserProfileRepository could have an interface like this:

public interface IUserProfileRepository
{
  void UpdateUserProfile(IUserProfile p);
}

This is the kind of abstraction that makes sense to anyone reading your code because it makes its intent clear. Furthermore, the only reason for the above code to change is if there is actually a change to the way you are storing IUserProfile, and if you have some special handling to do when updating UserProfile, you have somewhere to put it.

Compare your:

void Update<TEntity>(IEnumerable<string> excludeColumnNames, params TEntity[] items)

which obscures meaning when calls to this method are sprinkled around your code to achieve many different outcomes, and if there's special handling for updating a specific entity type, you have nowhere to put it, so it ends up getting pushed up a layer and duplicated every time its needed.

Also consider that getting an entity and updating it are two very different operations. You might get all properties on an entity, but only update some of them, which I believe you are referring to in your question. I consider the way EF makes read and writes to entities 'feel' almost the same to be an awful abstraction that leads to a great deal of pain and suffering. Reading and writing data are very distinct use cases, as the CQRS pattern proves.

Answer 3

If you want to pass multiple columns in, then pass them in...
Arrays or lists are the usual way.

public virtual void Update<P>(List<Expression<Func<T, P>>> excludeColumns, params T[] items)
{
     foreach (T item in items)
     {
         foreach (Expression<Func<T, P>> excludeColumn in excludeColumns)
         {
            _entities.Entry(item).State = EntityState.Modified;
            _entities.Entry(item).Property(excludeColumn).IsModified = false;       
         }
     }

    _entities.SaveChanges();
}

That'll be fine as long as each column has the same type P.

Answer 4

I have a similar method, but using linq-to-sql. It's a bit different approach, but this works very well. You may need to adapt it to use with EF.

public class Data
{
    public static void Update<T>(Func<T, bool> where, Action<T> change)
    {
        IEnumerable<T> items = ((IEnumerable<T>)myDataContext.GetTable(typeof(T))).Where(where);

        foreach (T item in items)
        {
            change(item);
        }

        myDataContext.SubmitChanges();
    }
}

I use it like this:

Data.Update<User>(u => u.Id == 3, u =>
{
    u.Name = "John";
    u.Age = 25;
    u.Email = "john@email.com";
});

While this may not be a direct answer to this question, I believe this is relevant and will help.