EF , automatic history of table and T4 files (TT files)

Usually the data of the tables should be tracking for who modified it.

Think about inserting/updating/deleting an employee :   you must know who did those actions and when. So you create another table, identically as structure, and you add another 3 fields , such as [ModifiedDate](when), [ModifiedBy](who), [ModifiedType] (what : insert, update, delete).

There are several methods to do it :

  1. from database :
  2. from programming code  – every time you modify an object, you remember to modify the history object with appropiate data.

The drawback with the database approach is that you can not retrieve who done the modifications ( usually the applications connect under a single account and have a roles table)

The drawback with the programming approach is that the programmer must REMEMBER doing so…If he does not(and does not wrote tests for history), you are stuck…

In the following I propose an automatically history – that maps convention over configuration in my template, but it is easy for you to modify.

The solution works with Entity Framework 4.0 and, for more easily spearation of concerns , with POCO generators.

Let’s say you have the following tables :

database diagram

As you see we have a Employee and a employee_history, an Department and Department_history

The conventions are:

the history table name = “object” table name  +  “_history” suffix

the history table fields = “object” table name  fields +[ModifiedDate], [ModifiedBy], [ModifiedType]

(if you change those conventions , please change the modelhistory.tt file)

If you want to see in action , please  download code history and do the following
1. create database tests
2. run history.sql
3. run project
4. if necessay, re-create the model1.edmx with the same name and replace the console application app.config with the new connection string

After works, please add any fields to department table  and to department_history table(same field names/type) .  Re-compile the application and modify the new field in department. You will see the modifications in the department_history table.

Ok,now how we do the magic :

We create two new tt file that points to the model.edmx .

The first one ModelHistory.tt , takes care of  creating the constructor for history entities by taking a parameter from the original entity :

public Department_History(Department original):this()
{
this.IDDepartment=original.IDDepartment;
this.Name=original.Name;
}

How it do this magic ? Simple : the ModelHistory.tt recognize the model and history in the name of tables:

</pre>
string inputFile = @"Model1.edmx";
string History = "_History";
<pre>

then it generate code for constructor :

	#>
		public <#=code.Escape(entity)#>():base()
		{
		}
		public <#=code.Escape(entity)#>(<#=NameEntityOriginal #> original):this()
		{
		<#
	foreach (EdmProperty edmProperty in entityOriginal.Properties.Where(p => p.TypeUsage.EdmType is PrimitiveType && p.DeclaringType == entityOriginal))
	{
		#>
				this.<#= code.Escape(edmProperty.Name) #>=original.<#= code.Escape(edmProperty.Name) #>;
		<#

	}
	#>
		}
	<#
</pre>

Ok, and then how to create the history entity ? I wish that the POCO template has had an event “Database saving” – but the only thing I can have is SaveChanges from the ObjectContext – so I create a new ObjectContext , derived from the default one that comes with the project, and creates a new history object :


public override int SaveChanges(SaveOptions options)
{
this.DetectChanges();
DateTime dtModified=DateTime.Now;
string UserModified=clsUser.UserName;
foreach (ObjectStateEntry ose in this.ObjectStateManager.GetObjectStateEntries(EntityState.Added | EntityState.Deleted | EntityState.Modified))
{

//could do this way too
//if (ose.Entity != null && ose.Entity.GetType() == typeof(...))
//{
//}
if (ose.Entity != null)
{
string NameType=ose.EntitySet.ElementType.Name;

switch(NameType)
{

case "Department":
var itemDepartment_History = new Department_History(ose.Entity as Department);
//if compile error here, that means you keep tracking
//of which modified with another properties
//please modify the tt accordingly
itemDepartment_History.ModifiedType= ose.State.ToString();
itemDepartment_History.ModifiedDate= dtModified;
itemDepartment_History.ModifiedBy= UserModified;
base.Department_History.AddObject(itemDepartment_History);
break;

case "Employee":
var itemEmployee_History = new Employee_History(ose.Entity as Employee);
//if compile error here, that means you keep tracking
//of which modified with another properties
//please modify the tt accordingly
itemEmployee_History.ModifiedType= ose.State.ToString();
itemEmployee_History.ModifiedDate= dtModified;
itemEmployee_History.ModifiedBy= UserModified;
base.Employee_History.AddObject(itemEmployee_History);
break;

}
}
}

return base.SaveChanges(options);
}

Now all is ready and I made a console application for testing manually (ok, should make a NUnit / MSTest / xUnit )

 using (var ctx = new testsEntitiesHistory())
            {
                var dep = new Department();
                dep.Name = "IT";
                ctx.Departments.AddObject(dep);
                ctx.SaveChanges();
                id = dep.IDDepartment;
            }
            using (var ctx = new testsEntitiesHistory())
            {
                var dep = ctx.Departments.Where(depart => depart.IDDepartment == id).FirstOrDefault();
                dep.Name = "Information tehnology";
                ctx.SaveChanges();
                //
            }
            using (var ctx = new testsEntitiesHistory())
            {
                var dep = ctx.Departments.Where(depart => depart.IDDepartment == id).FirstOrDefault();
                ctx.Departments.DeleteObject(dep);
                ctx.SaveChanges();

            }
            using (var ctx = new testsEntitiesHistory())
            {
                foreach (var dephist in ctx.Department_History)
                {
                    Console.WriteLine("Found {0} with state {1}", dephist.Name,dephist.ModifiedType);
                }
            }

And the output is :

history saving
automatically saving history department

Now you can add more tables to the edmx or change the fields – all is done automatically when compiling

If you want to see in action , please download code history

Update : for another way to do it( generating trigger and tables ) please see : http://msprogrammer.serviciipeweb.ro/2010/09/27/generating-history-trigger-with-ef-edmx-and-tt-files/

For all programmers

For all programmer, just remember :
http://timstall.dotnetdevelopersjournal.com/coding_is_just_the_tip_of_the_iceberg.htm
And he says nothing about clients, marketing, sales, promotion, SEO, SMM, and other usual stuff…

Monitor SqlServer Stored Proc execution time

I have had the task to made something to monitor what stored procedure, in Sql Server,in most used. The problem that I have had is that the host_name is easily changed by putting, in the connection string WSID (see http://msdn.microsoft.com/en-us/library/ms178598.aspx )
The code is for sql server 2005 and I have not realized without the help of http://sqlserver.ro/forums/thread/8332.aspx
So there are the three easy steps to have the monitor proceudre in place:
1. Define an audit table :

CREATE TABLE [Audit](
	[ID] [int] IDENTITY(1,1) NOT NULL,
	[ProcName] [varchar](128)
	[IdentifierAccess] [varchar](max)
	[DateAccess] [datetime] NULL DEFAULT (getdate())
) ON [PRIMARY]

2. Stored proc to insert data :

alter proc dbo.proc_Log(@procname varchar(100))

as
begin
declare @userid varchar(100)
BEGIN TRY
select @userid=client_net_address from sys.[dm_exec_connections] with(nolock)
where session_id = @@SPID
End TRY
BEGIN CATCH
--GRANT SELECT ON sys.dm_exec_connections TO the user or
--grant view server state to the user ! http://msdn.microsoft.com/en-us/library/ms186717.aspx
set @userid = SUSER_NAME()
End CATCH

INSERT INTO [Audit]
(
[ProcName]
,[IdentifierAccess]
)
VALUES
(@procname,
@userid
)
end

3. In each stored procedure put this at beginning

DECLARE @ProcName nvarchar(128)

    SET @ProcName = OBJECT_NAME(@@PROCID)

    exec proc_Log  @ProcName

And this at the end of stored proc :

    exec proc_Log  @ProcName

4. See what the procedure calling are and the time (basic)

select a1.*,'--',a2.* , datediff(ss,a2.DateAccess,a1.DateAccess) 
from [Audit] a1  inner join [Audit] a2 on
a2.ID =
( select max(ID) from [Audit] a2 where a1.ProcName= a2.ProcName and a1.IdentifierAcces=a2.IdentifierAcces and a1.ID > a2.ID )
order by 1

From there is a simple way to make a group by to have the sum * count or other relevant data ( which proceudres are most used in what times of the day, and so on)

Work from home :
1. From the previous select(put it into a Common table Expression ), see what procedure use most
a) number of uses
b) number of time
2. What is missing from Audit table definition ? (Hint :easy clustered index definition)

file helpers

More than one time you need to let users / application import bulk data in your system .

I have had many cases  – data is old data from old software, that is exported in some text files, usually csv.

In order to import fast the data, I used FileHelpers from http://www.filehelpers.com/

Let’s say you have the following data for an car evidence :

Date, StartMiles, Destination, EndMiles

And the data comes from an Excel  that has those columns. The requirements is the user can copy/paste from Excel data

When they copy the data comes in this form

01/01/2010     1000 Washington 1550

02/01/2010     1550 Dallas 2550

and so on.

It is clear that you :

  1. have a class with Date, StartMiles, Destination, EndMiles
  2. accomodate for space – but how to perform various data separator ( we suppose that first is the day, then comes the month).

Now the code in logical steps :

1. Accomodate for dates :

internal class ConvertDate : ConverterBase
{

        /// <summary>
        /// different forms for date separator : . or / or space
        /// </summary>
        /// <param name="from">the string format of date - first the day</param>
        /// <returns></returns>

        public override object StringToField(string from)
        {
            DateTime dt;

            if (DateTime.TryParseExact(from, "dd.MM.yyyy", null, DateTimeStyles.None, out dt))
                return dt;

            if (DateTime.TryParseExact(from, "dd/MM/yyyy", null, DateTimeStyles.None, out dt))
                return dt;

            if (DateTime.TryParseExact(from, "dd MM yyyy", null, DateTimeStyles.None, out dt))
                return dt;

            throw new ArgumentException("can not make a date from " + from, "from");

        }
}

2. Create the class that will hold one record:

    [IgnoreEmptyLines(true)]
    [DelimitedRecord(",")]
    internal class DestinationReader
    {
        //[FieldConverter(ConverterKind.Date,"dd.MM.yyyy")]
        [FieldConverter(typeof(ConvertDate))]
        public DateTime Date;
        [FieldConverter(ConverterKind.Int32)]
        public int StartMiles;

        [FieldQuoted(QuoteMode.OptionalForBoth)]
        public string Destination;

        [FieldConverter(ConverterKind.Int32)]
        public int  EndMiles;
    }

3. Now read the entire string:

            string Text = text that comes from the user
            string TextDelim = Text.Substring(10, 1);// the date has 10 chars - so the eleven is the separator
            while (Text.IndexOf(TextDelim + TextDelim) > 0)//consolidate for 2 delimiters
            {
                Text = Text.Replace(TextDelim + TextDelim, TextDelim);
            }
            DelimitedFileEngine<DestinationReader> flh=new DelimitedFileEngine<DestinationReader>();
            flh.Options.Delimiter = TextDelim;

            var data =flh.ReadString(Text);

In data you have a list of DestinationReader
So for any structured import of data use FileHelpers from http://www.filehelpers.com/

Selenium and testing WebInterfaces

On some cases you need to test the whole web interface. Why ? Suppose you have some Ajax call. You can test the call on server, but how do you ensure that user experience is OK ?

There are several testing projects for Web – for example selenium and Watin

I will show how to test with Selenium + NUNIT

  1. Download selenium from http://seleniumhq.org/ and download Firefox add-on from http://seleniumhq.org/projects/ide/
  2. Start selenium with java -jar selenium-server.jar
  3. Record the test with FF.
  4. Create a usual NUnit test project and add a reference to ThoughtWorks.Selenium.Core.dll
  5. The code can be like this :
  6. using System;
    using System.Text;
    using System.Text.RegularExpressions;
    using System.Threading;
    using NUnit.Framework;
    using Selenium;
    
    namespace InvoiceTest
    {
    [TestFixture]
    public class TestWeb
    {
    private ISelenium selenium;
    private StringBuilder verificationErrors;
    
    [SetUp]
    public void SetupTest()
    {
    //java -jar selenium-server.jar
    //selenium = new DefaultSelenium("localhost", 4444, @"*firefox d:\Program Files\Mozilla Firefox\firefox.exe","<a href="http://localhost/&quot;);">http://localhost/");</a>
    selenium = new DefaultSelenium("localhost", 4444, @"*iexplore", "<a href="http://localhost/&quot;);">http://localhost/");</a>
    //selenium = new DefaultSelenium("localhost", 4444, @"*iexploreproxy", "<a href="http://localhost/&quot;);">http://localhost/");</a>
    selenium.Start();
    verificationErrors = new StringBuilder();
    }
    
    [TearDown]
    public void TeardownTest()
    {
    try
    {
    selenium.Stop();
    }
    catch (Exception)
    {
    // Ignore errors if unable to close the browser
    }
    Console.WriteLine(verificationErrors.ToString());
    Assert.AreEqual("", verificationErrors.ToString());
    }
    
    [Test]
    public void FindTextAfterClickOnButton()
    {
    
    selenium.Open("/<strong>your application</strong>");
    selenium.Focus("");
    selenium.WindowMaximize();
    selenium.Type("the textbox", "the text");
    selenium.Click("button ");
    selenium.WaitForPageToLoad("30000");
    try
    {
    Assert.IsTrue(selenium.IsTextPresent("<strong>new text from your application</strong>"));
    }
    catch (AssertionException e)
    {
    verificationErrors.Append(e.Message);
    Console.WriteLine(e.Message);
    }
    }
    
    }
    }
    

Entity Framework profiler

Many times I’ve had problem with the following error when inserting objects with dates with Entity Framework :

System.Data.UpdateException An error occurred while updating the entries. See the inner exception for details.SqlDateTime overflow. Must be between 1/1/1753 12:00:00 AM and 12/31/9999 11:59:59 PM.

Ok, it’s my faute – but to remember each one date is too much for me…

The usual method was to start SqlProfiler, monitor the database and see how the sql is constructed. However, the database being used by all developers, it was not so simple to differentiate between all sql’s.

Other alternative was to log the entity framework generated sql’s . I have discovered Ayende Rahien Entity Framework profiler . Simple to use , as is wrote here in 2 simple steps

  1. add reference to HibernatingRhinos.Profiler.Appender.dll
  2. put this
HibernatingRhinos.Profiler.Appender.EntityFramework.EntityFrameworkProfiler.Initialize();

and start the exe. That will be all.

Do not forget to remove it on release version!

Pros:

Easy to use, valuable information, good!

Cons :

Not free …

Caching in .NET

In every application you have some data that is more read-more, write-once or twice. For example you can have the list of Cities of a country, the list of Countries of the world or list of exchange currency. This data is modified rarely. Also, you can have data that is not very sensitive to be real-time , such as the list of invoices for the day.
In .NET 3.5 you have several options
1. ASP.NET caching – and implementing in other applications with HttpRuntime  ( even if MS says “The Cache class is not intended for use outside of ASP.NET applications”)

2. Enterprise caching block : hard to configure

3. memcached , velocity, sharedcache and other third party providers – that comes with their options and configuration

In .NET 4.0 you have a new kid : objectcache and , more important , an implementation : MemoryCache

http://msdn.microsoft.com/en-us/library/system.runtime.caching.memorycache%28v=VS.100%29.aspx

What is very good is that now you can cache in Memory what do you want – and apply easily to your Business Layer. More, the object is a singleton for the application – that is even better (see the test on the final of the post)

What it is missing is an easy implementation for List and an implementation to remove data after a defined time.

So I decided to do my implementation for that (ok, it is wrong to have both implementations in a single class – but you can separate easily )

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Runtime.Caching;

namespace CachingData{

    /// <summary>
    /// List<int> i = new List<int>() { 1, 10, 100 };
            //CacheData_List<List<int>, int> data = new CacheData_List<List<int>, int>(2);
            //data.Add(i);
            //Assert.AreEqual(3, data.Items().Count, "must be 3");
            //data = new CacheData_List<List<int>, int>();
            //Assert.AreEqual(3, data.Items().Count, "must be 3");
            //Assert.IsTrue(data.ContainsData, "must have data");
            //Thread.Sleep(1000 * 3);
            //Assert.IsFalse(data.ContainsData, "must not have data");
    /// </summary>
    /// <typeparam name="T">and generic ILIST </typeparam>
    /// <typeparam name="U"></typeparam>
    public class CacheData_List<T, U>
        where T:class,IList<U>, new()

    {
        /// <summary>
        /// use this for adding in cache
        /// </summary>
        public event EventHandler RemovedCacheItem;

        private System.Timers.Timer timerEvent;
        private MemoryCache buffer;
        private int TimeToLiveSeconds;

        private DateTimeOffset dto;
        private string Key;
        /// <summary>
        /// default constructor - cache 600 seconds = 10 minutes
        /// </summary>
        public CacheData_List()
            : this(600)
        {
        }
        /// <summary>
        /// constructor cache the mentioned TimeSeconds time
        /// </summary>
        /// <param name="TimeSeconds">value of time for cache</param>
        public CacheData_List(int TimeSeconds)
        {
            TimeToLiveSeconds = TimeSeconds;
            timerEvent=new System.Timers.Timer(TimeToLiveSeconds * 1000);
            timerEvent.AutoReset = true;
            timerEvent.Elapsed += new System.Timers.ElapsedEventHandler(timerEvent_Elapsed);
            dto = new DateTimeOffset(DateTime.UtcNow.AddSeconds(TimeToLiveSeconds));
            Key = typeof(T).ToString();
            buffer = MemoryCache.Default;
        }

        void timerEvent_Elapsed(object sender, System.Timers.ElapsedEventArgs e)
        {
            if (RemovedCacheItem != null)
            {
                RemovedCacheItem(this, EventArgs.Empty);
            }
        }
        /// <summary>
        /// remove item from cache
        /// </summary>
        public void Remove()
        {
            if (buffer.Contains(Key))
            {
                buffer.Remove(Key);

            }
            dto=new DateTimeOffset(DateTime.UtcNow.AddSeconds(TimeToLiveSeconds));
        }
        /// <summary>
        /// add multiple items to cache
        /// </summary>
        /// <param name="items">items to add to the cache</param>
        public void Add(T items)
        {

            if (buffer.Contains(Key))
            {
                T data = Items();
                foreach (var t in data)
                {
                    items.Add(t);
                }
                buffer.Remove(Key);
            }
            buffer.Add(Key, items, dto);
        }
        /// <summary>
        /// add a item to the IList of the cache
        /// </summary>
        /// <param name="item">an item to add</param>
        public void AddItem(U item)
        {

            T data=new T();
            if (buffer.Contains(Key))
            {
                data = buffer.Get(Key) as T;
                buffer.Remove(Key);
            }

            data.Add(item);
            buffer.Add(Key, data,dto);
        }
        /// <summary>
        /// usefull if you do not intercept the removed event
        /// </summary>
        public bool ContainsData
        {
            get
            {
                return buffer.Contains(Key);
            }

        }
        /// <summary>
        /// retrieve items
        /// </summary>
        /// <returns></returns>
        public T Items()
        {
            if (!buffer.Contains(Key))
                return null;

            return buffer.Get(Key) as T;
        }
    }
}

Please note that the test for usage is in the summary :

</pre>
List i = new List() { 1, 10, 100 };
CacheData_List <List<int>, int> data = new CacheData_List<List<int>, int>(2);
data.Add(i);
Assert.AreEqual(3, data.Items().Count, "must be 3");
data = new CacheData_List<List<int>, int>();
Assert.AreEqual(3, data.Items().Count, "must be 3");
Assert.IsTrue(data.ContainsData, "must have data");
Thread.Sleep(1000 * 3);
Assert.IsFalse(data.ContainsData, "must not have data");

You can download the file from

http://msprogrammer.serviciipeweb.ro/wp-content/uploads/2010/05/CachingData.zip

(Last Note : for synchonization maybe it was better to lock on readerwriterlockslim