LINQ to SQL Table Extensibility Methods

If I have a LINQ to SQL table that has a field called say Alias.

There is then a method stub called OnAliasChanging(string value);

What I want to do is to grab the value, check the database whether the value already exists and then set the value to the already entered value.

So I may be changing my alias from "griegs" to "slappy" and if slappy exists then I want to revert to the already existing value of "griegs".

So I have;

    partial void OnaliasChanging(string value)
    {
开发者_如何学运维        string prevValue = this.alias;
        this.Changed = true;
    }

When I check the value of prevValue it's always null.

How can I get the current value of a field?

Update

If I implement something like;

    partial void OnaliasChanging(string value)
    {
        if (this.alias != null)
            this.alias = "TEST VALUE";
    }

it goes into an infinte loop which is unhealthy.

If I include a check to see whether alias already == "TEST VALUE" the infinate loop still remains as the value is always the original value.

Is there a way to do this?

The code snippets you've posted don't lend themselves to any plausible explanation of why you'd end up with an infinite loop. I'm thinking that this.alias might be a property, as opposed to a field as the character casing would imply, but would need to see more. If it is a property, then you are invoking the OnAliasChanging method before the property is ever set; therefore, trying to set it again in the same method will always cause an infinite loop. Normally the way to design this scenario is to either implement a Cancel property in your OnXyzChanging EventArgs derivative, or save the old value in the OnXyzChanging method and subsequently perform the check/rollback in the OnXyzChanged method if you can't use the first (better) option.

Fundamentally, though, what you're trying to do is not very good design in general and goes against the principles of Linq to SQL specifically. A Linq to SQL entity is supposed to be a POCO with no awareness of sibling entities or the underlying database at all. To perform a dupe-check on every property change not only requires access to the DataContext or SqlConnection, but also causes what could technically be called a side-effect (opening up a new database connection and/or silently discarding the property change). This kind of design just screams for mysterious crashes down the road.

In fact, your particular scenario is one of the main reasons why the DataContext class was made extensible in the first place. This type of operation belongs in there. Let's say that the entity here is called User with table Users.

partial class MyDataContext
{
    public bool ChangeAlias(Guid userID, string newAlias)
    {
        User userToChange = Users.FirstOrDefault(u => u.ID == userID);
        if ((userToChange == null) || Users.Any(u => u.Alias == newAlias))
        {
            return false;
        }
        userToChange.Alias = newAlias;

        // Optional - remove if consumer will make additional changes
        SubmitChanges();

        return true;
    }
}

This encapsulates the operation you want to perform, but doesn't prevent consumers from changing the Alias property directly. If you can live with this, I would stop right there - you should still have a UNIQUE constraint in your database itself, so this method can simply be documented and used as a safe way to attempt a name-change without risking a constraint violation later on (although there is always some risk - you can still have a race condition unless you put this all into a transaction or stored procedure).

If you absolutely must limit access to the underlying property, one way to do this is to hide the original property and make a read-only wrapper. In the Linq designer, click on the Alias property, and on the property sheet, change the Access to Internal and the Name to AliasInternal (but don't touch the Source!). Finally, create a partial class for the entity (I would do this in the same file as the MyDataContext partial class) and write a read-only wrapper for the property:

partial class User
{
    public string Alias
    {
        get { return AliasInternal; }
    }
}

You'll also have to update the Alias references in our ChangeAlias method to AliasInternal.

Be aware that this may break queries that try to filter/group on the new Alias wrapper (I believe Linq will complain that it can't find a SQL mapping). The property itself will work fine as an accessor, but if you need to perform lookups on the Alias then you will likely need another GetUserByAlias helper method in MyDataContext, one which can perform the "real" query on AliasInternal.

Things start to get a little dicey when you decide you want to mess with the data-access logic of Linq in addition to the domain logic, which is why I recommend above that you just leave the Alias property alone and document its usage appropriately. Linq is designed around optimistic concurrency; typically when you need to enforce a UNIQUE constraint in your application, you wait until the changes are actually saved and then handle the constraint violation if it happens. If you want to do it immediately your task becomes harder, which is the reason for this verbosity and general kludginess.

One more time - I'm recommending against the additional step of creating the read-only wrapper; I've put up some code anyway in case your spec requires it for some reason.

Is it getting hung up because OnaliasChanging is firing during initialization, so your backing field (alias) never gets initialized so it is always null?

Without more context, that's what it sounds like to me.