Month: February 2016

EF6.x Correlating Poor Performing SQL to Application Code

romiller.com

When using an O/RM, poor performing SQL statements are often not discovered until you (or your DBA) find that a particular query is slowing down your database server. At this point, it becomes hard to identify which piece of application code is causing that SQL to be executed.

An interceptor to log slow/failed SQL

In EF6.0.0 we introduced interceptors, which allow you to get into the pipeline just before, and just after, a query/command is sent to the database.

Here is an interceptor that detects queries/commands that either failed, or exceeded a chosen threshold for execution time. In addition to the query being executed, it logs the call stack. This allows us to identify which piece of our application code initiated the database operation.

This example writes to a log file, but you can log the information to wherever needed.

Registering the interceptor

The easiest way to get our…

View original post 133 more words

Entity Framework validation with partial updates

I had this issue today, basically Entity Framework always validate all the entity, so if you want to make a partial update using Attach/Entity State way, you can’t do it if you have configured validation on your model and you don’t fill all your entity with good values, but we want to do a partial update without do another query before, and also without filling up all the entity with data I don’t have, so how to do it?

The most easiest and fastest way is to completely disable the validation for your operation, but this is not the ideal way, what about validating the property I’m near to udpate?

using (var db = new NorthwindDbContext())
{
    // disable the validation for this operation
    db.Configuration.ValidateOnSaveEnabled = false;

    var customer = new Customer
    {
        CustomerId = "ALFKI"
    };
    db.Customers.Attach(customer);

    customer.PostalCode = "123456789000";

    // throw sql exception because PostalCode goes over 10 chars maximum
    // String or binary data would be truncated.
    await db.SaveChangesAsync();
}

Thanks to Shimmy I found what I think is the most elegant solution:

protected override DbEntityValidationResult ValidateEntity(DbEntityEntry entityEntry, IDictionary<object, object> items)
{
    var result = base.ValidateEntity(entityEntry, items);

    var falseErrors = result.ValidationErrors
                            .Where(error =>
                            {
                                if (entityEntry.State != EntityState.Modified) return false;
                                var member = entityEntry.Member(error.PropertyName);
                                var property = member as DbPropertyEntry;
                                if (property != null) return !property.IsModified;
                                return false;
                            });

    foreach (var error in falseErrors.ToArray())
    {
        result.ValidationErrors.Remove(error);
    }

    return result;
}

To the Shimmy sample I’ve added the condition “if (entityEntry.State != EntityState.Modified) return false;” to avoid bypass the validation when adding an entity.

Now if you test the same code as before, you will receive an entity framework exception and no round-trip will be made on sql server:

using (var db = new NorthwindDbContext())
{
    var customer = new Customer
    {
        CustomerId = "ALFKI"
    };
    db.Customers.Attach(customer);

    customer.PostalCode = "123456789000";

    // Validation failed for one or more entities. See 'EntityValidationErrors' property for more details.
    // The field PostalCode must be a string or array type with a maximum length of '10'.
    await db.SaveChangesAsync();
}