Re: Dynamic Table Creation on Database Using SQL Datatypes from a Dataset, Possible? - C# Discussion Boards

Help with simple user control

8-Dec-03 10:34

All you need to do is use a collection of controls (these timer controls you mentioned) and add them sequentially to the parent control's Controls collection, while setting their Size and Location properties. Presumably, their Size would be based on the parent's Width and the control's Height, and you could set the X and Y coordinates in the Location property programmatically when you add the controls to the parent control collection (incrementing and keeping track of the last Y position used, for instance.

If you're talking about the Repeater control in .NET, however, that is only for ASP.NET and won't work for Windows Forms controls since it merely generates HTML.

Charlie Williams8-Dec-03 9:05

Charlie Williams

8-Dec-03 9:05

I am reasonably competent with C# (please note that "reasonably" is a very subjective term), but don't have much (read: any) experience building user controls.

The (very simple) user control I am having problems with contains two panels, three labels and a combobox. Which of the two panels is visible depends on whether the control is currently in edit mode. Currently I have the control go into edit mode and thus change which panel is visible when any control in the user control is clicked. I am looking for a clean way of doing the opposite: I can't seem to find an event that will fire when a user clicks on another part of the form. Therefore, the control still looks like it's being edited while the user is off doing something else entirely.

I'm sure the answer will be sufficiently simple so as to make me feel foolish for asking, but it's really starting to irritate me.

Any help would be appreciated.

Charlie

Here I am. Love me.

Re: Help with simple user control

Heath Stewart8-Dec-03 10:25

Re: Help with simple user control

8-Dec-03 10:25

You should be able to handle the LostFocus event (or Leave event) of the UserControl itself, since the GotFocus and Enter events bubble up. This way, when any control in your UserControl (and thus the UserControl itself) loose focus, you can switch your panels' visibility.

Charlie Williams8-Dec-03 12:56

Charlie Williams

8-Dec-03 12:56

Thank you for the response. It turned out that by making the panel visible, but not giving focus to any of the controls on it, that the original control still had the focus and as such, no LostFocus event would fire Blush | :O

Your reinforcement that I was handling the correct event allowed me to stop looking araound aimlessly and figure out what the problem was.

Thank you.

Charlie

Here I am. Love me.

Dynamic Table Creation on Database Using SQL Datatypes from a Dataset, Possible?

Paul Evans8-Dec-03 7:46

Re: Dynamic Table Creation on Database Using SQL Datatypes from a Dataset, Possible?

8-Dec-03 7:46

"Dynamic Table Creation on Database Using SQL Datatypes from a Dataset" quite a mouth full, but that's what we are trying to do!

We have a source database which is of one data-provider, and a destination which is of another provider. The source database is in structural flux, so we need to be able to dynamically create a destination database from a dataset populated by a select * from table. Then read off the column names and types and recreate it on a different database.

There are about 30 million examples of Datasets and datagrids, but nothing that actually has the same effect as an CREATE TABLE as we've seen.

When interrogating the dataset we get the .Net datatypes. What we need is the SQL datatypes associated with those .Net datatypes for that particular provider (probably via the SqlDataAdapter?) so we can build a decent CREATE TABLE query that the provider will understand from the contents of the dataset.

Or better yet just get .Net to do it, because it obviously already knows the mappings or it wouldn't be able to translate datasets into .Net types from SQL types when reading, and back again when updating!

Any ideas or avenues of investigation gratefully received!

/**********************************
Paul Evans, Dorset, UK.
Personal Homepage "EnjoySoftware" @
http://www.enjoysoftware.co.uk/
**********************************/

Heath Stewart8-Dec-03 8:01

Re: Dynamic Table Creation on Database Using SQL Datatypes from a Dataset, Possible?

8-Dec-03 8:01

No, a database can't be created from a DataSet - at least not with classes in the BCL. There might be some third-party library out there that can do it, though. Mapping from a database (table, view, or stored proc) to a DataSet is easy but there just isn't support for the other way because it typically is not done (backends are usually written and mostly solidified when application coding begins).

It's obvious that one of these is a SQL Server installation, so why not just use a DTS transform? For any database with an OLE DB provider, you can create appropriate table structures, transform (if necessary) and copy data, and some other things. You even get a nice graphic utility to help you with the transform and these DTS transforms can be saved and executed again and again (hint, you can use this technique with the SQL DMO library, either using a script, native COM client, or .NET interop assembly). This is by far a better - and faster - way of doing it.

Paul Evans8-Dec-03 8:19

Re: Dynamic Table Creation on Database Using SQL Datatypes from a Dataset, Possible?

8-Dec-03 8:19

Is it obvious? We could very easily be using Ingres and Oracle. Don't take too much for granted!

I know we could do a ddl export or similar from sql server - but it's not from that side. Currently we are using ODBC, but that will change over to using the proper Ingres .net stuff when we get our hands on it. We are currently looking at a couple of different platforms because in the end some of it is going to be running on a PDA.

Your answer is fair, but is actually to a different question I didn't ask.

To restate what I meant because I didn't put myself over that clearly:

1. Dynamic SQL Dataprovider <-> .Net datatype mapping - so we can dynmacially construct "CREATE TABLE" commands with the proper datatypes for that provider from column headings and .Net datatypes from the Dataset.

2. Possibility of using a Dataset to actually create a table rather then updating / deleting entries. (Long shot and probably not possible in anyway).

I think you gave your opinon on 2 - which is that it is impossible. Smile | :)

(1) Is fustrating because MS even publish a lovely table mapping types, and as I said the framework itself must know what .Net type relates to which SQL type for the provider or the Dataset wouldn't work in the first place:

http://msdn.microsoft.com/library/default.asp?url=/library/en-us/cpguide/html/cpconmappingnetdataproviderdatatypestonetframeworkdatatypes.asp

/**********************************
Paul Evans, Dorset, UK.
Personal Homepage "EnjoySoftware" @
http://www.enjoysoftware.co.uk/
**********************************/

Heath Stewart8-Dec-03 10:04

8-Dec-03 10:04

Actually, I said it's not possible with the classes in the BCL (base class library). You could make (or a third-party library may already exist) a class (or classes) that takes those mappings and generates and / or executes the appropriate CREATE TABLE or ALTER TABLE commands.

It was obvious you were using SQL Server because you were refencing the System.Data.SqlClient classes! Smile | :)

They only work for SQL Server (specifying any other OLE DB provider throws an exception when initializing a SqlConnection. If you don't want to use SQL Server, you MUST use the System.Data.OleDb namespace classes (or a custom provider implementation, such as ODBC.NET or Oracle.NET - others also exist in beta form last time I checked-up on them).

The SQL classes do understand the data-types as you suspect (and it is well-documented). This comes at an expense, though. Instead, you should consider using strongly-typed DataSet classes when possible. This makes table and column lookups - as well as field resolution - must faster and more type-friendly. For instance, many people always use DataRow[string] to get a column value, but this is very expensive because it must enumerating the columns each time to find the column with that name. There are ways to make this faster (such as caching the column index for that name), but strongly-typed DataSet classes do something similar automatically (technically, they already know the column indexes).

Thread-Safe Singleton

bzurer8-Dec-03 5:18

8-Dec-03 5:18

This is the code I have always used for a singleton.

public sealed class Singleton
{

private Singleton(){}
public static readonly Singleton Instance = new Singleton();

}

I am just getting into multithreading. I have read conflicting opinions on whether or not the code above is thread-safe?

Robert Zurer
robert@zurer.com

Heath Stewart8-Dec-03 6:07

8-Dec-03 6:07

statics are initialized with the Type, but the above code won't get you a singleton. The code below is oft-used and recommended by Microsoft for the .NET Framework:

public sealed class Singleton
{
  private static Singleton instance;
 
  private Singleton() {}
  // Makes sure that only one instance exists.
  private static Singleton Instance
  {
    get
    {
      // Thread-safe initialization.
      if (instance == null)
        lock (typeof(Singleton))
          if (instance == null)
            instance = new Singleton();
      return instance;
    }
  }
  private void doSomething()
  {
    Console.WriteLine("Something done.");
  }
  public static void DoSomething()
  {
    Instance.doSomething();
  }
}

Notice how the public methods are static? In most cases, this is what you want because you control what the caller can do. The Instance property is a thread-safe initializer that makes sure that only one instance is created. Your public static methods use that property (which will initialize the class instance if necessary) to call private instance methods or use private instance properties or fields. That is a singleton.

Keep in mind that the methods themselves are not thread-safe this way, though. To do that, you'll either want to use the lock keyword (on a static object, like the typeof(Singleton) I used, or this for instance methods / property accessors).

Now, you can accomplish this another way by using a ContextBoundObject where you can intercept calls but to the caller - it's just creating a new instance (where you return an existing instance). This gets into remoting issues which are more advanced, so you should probably try this out first. It's by far easier and accomplishes the same task.

Paul Evans8-Dec-03 7:37

8-Dec-03 7:37

good call! Like that.

You may want to look at the lock hint pattern, so it doesn't have to do the mutex lock each time.

/**********************************
Paul Evans, Dorset, UK.
Personal Homepage "EnjoySoftware" @
http://www.enjoysoftware.co.uk/
**********************************/

Heath Stewart8-Dec-03 7:56

8-Dec-03 7:56

Not familiar with that, but in this case the lock is not performed each time. Once the instance is non-null, the rest of the condition block isn't executed.

Paul Evans8-Dec-03 8:05

8-Dec-03 8:05

you kind of did it, but without the boolean flag.

It's name is the "The Double-Checked Locking Pattern"

Example implementation here:
http://www.cuj.com/documents/s=8047/cuj9910ringle/

It comes from this academic paper (although the link is dead you can find it if you hunt about a bit):
[5] Douglas Schmidt and Tim Harrison. "Double-Checked Locking: An Optimization Pattern for Efficiently Initializing and Accessing Thread-safe Objects," http://www.cs.wustl.edu/~schmidt/DC-Locking.ps.gz

/**********************************
Paul Evans, Dorset, UK.
Personal Homepage "EnjoySoftware" @
http://www.enjoysoftware.co.uk/
**********************************/

Heath Stewart8-Dec-03 8:12

8-Dec-03 8:12

I did do the double-checked locking mechanism! Look at my code again:

if (instance == null)
  lock (typeof(Singleton))
    if (instance == null)
      instance = new Singleton();

Looks like double-checked locking to me, and was called such (couldn't remember what the "technical" term) in a Microsoft article published in the 1.0 beta days.

Paul Evans8-Dec-03 8:20

8-Dec-03 8:20

Indeed, as I said, you did it without the boolean flag. Kudos.

/**********************************
Paul Evans, Dorset, UK.
Personal Homepage "EnjoySoftware" @
http://www.enjoysoftware.co.uk/
**********************************/

Heath Stewart8-Dec-03 11:08

8-Dec-03 11:08

Sorry, I misunderstood one of your posts (well, maybe). As far as a boolean flag goes, though, the way I used a conditional would result in few instructions and field members. Having a separate field just to remember this would increase the size of the Type, and having to set that in addition to instantiating the singleton class is yet a few more instructions:

ldarg.0
ldc.i4.1
stfld Singleton.initialized

bzurer8-Dec-03 8:07

8-Dec-03 8:07

You mention that the methods themselves are not thread-safe. Do you mean that the method

public static void DoSomething()  <br />
{    <br />
    Instance.doSomething();  <br />
}

Should be

public static void DoSomething()  <br />
{    <br />
    lock (typeof(Singleton))<br />
    Instance.doSomething();  <br />
}

even though Instance itself is protected by a lock?

Heath Stewart8-Dec-03 8:10

8-Dec-03 8:10

The instance is not protected by a lock - the initialization of the Singleton was protected by a lock, but that was just to make sure that only one instance of the Singleton class was initialized. See the discussion Paul and I are having on this thread for more info.

Also, it is not necessary to lock every static method - only if you need to synchronize access to a resource. For doing something as simple as a calculation that doesn't use instance resources, you don't need to, like:

public static int Add(int a, int b)
{
  return a + b;
}

bzurer8-Dec-03 8:56

8-Dec-03 8:56

In this case, my Singleton will be used to synchronize access to a centralized Hashtable which holds the 'clean' or saved values of properties of objects while the actual objects are being validated, persisted and then finally updated if all goes well. After the update succeeds the values are removed from the hash table.

The aim is to allow multiple threads to be able to have read consistency while a write is going on. I want to accomplish this by deriving all my business from ContextBoundObject.All calls to set methods would be intercepted and the existing values added to this Hashtable(keyed to a global object id).

When another thread wants to read, the get method is intercepted, and, in SyncProcessMessage, the MessageSink checks for an entry in this Hashtable, if one is found, the value of the property is taken from the Hashtable and passed to the caller, otherwise it passes the IMessage unaltered.

I assume this Hashtable qualifies as a resource.

But this brings up another two questions.

1 - By using ContextBoundObject am I in some way automatically serializing access? and if not
2 - Must I place a lock within SyncProcessMessage?

Heath Stewart8-Dec-03 10:14

8-Dec-03 10:14

I don't know if you want to mess with a ContextBoundObject in this case. It's a lot of overkill. For instance, you want the value to always be available whether or not it exists first (or something like that). Use myHashtable[key] = value. If the key exists, the value is changed. If it doesn't exist, the key is added and the value is still set. Using a ContextBoundObject also means that the objects you're passing around have to be serializable and this mechanism is a little slower (as you have probably figured out) that direct calls.

As far as threading the Hashtable goes, you should actually lock against Hashtable.SyncRoot, or use the synchronized Hashtable returned from the static method Hashtable.Synchronized(myHashtable).

Zhang Ye Fan8-Dec-03 8:33

Zhang Ye Fan

8-Dec-03 8:33

By why do we need the "instance" in a singleton at all? Making the methods static seems to be enough:

<br />
public sealed class SingletonNoInstance<br />
{<br />
    private static Object m_objInit = new InitObj();<br />
<br />
    private static Object InitObj()<br />
    {<br />
        // do intialization here<br />
        // ...<br />
        return new Object();   <br />
    }<br />
  <br />
    public static void ThreadSafeDoSomething()  <br />
    {    <br />
        lock(typeof(SingletonNoInstance))<br />
        {<br />
            // do something here<br />
            // ...<br />
        }<br />
    }<br />
}<br />

The private member m_objInit and method InitObj() can be removed if no initialization of the class is required.

Heath Stewart8-Dec-03 10:08

8-Dec-03 10:08

There's a difference between singletons and a class with static methods. If you just have simple static methods (like the System.Math class), you don't need to construct a singleton. Singletons are useful when you want lots of different callers to use the same resource without having multiple instances of that resource (or the class that has that resource). For instance, all throughout the .NET BCL (and our application I architected), most classes in an assembly use a internal singleton that has a single instance of a ResourceManager that contains strings (and other resources at times) for all classes to use (like common exception messages or property categories and descriptions). It would be horribly inefficient to have an instance of this ResourceManager for each class (and who knows how many instances of each class?!), so they use a singleton with a single instance of the ResourceManager.

There are many other reasons for singletons (if you know anything about remoting, there's some good examples for that, too). This is just a common example I think of because I use it a lot and it is far more efficient to do it this way that for each instance of a class to have it's own - especially when you're sharing string resources amongst other classes.

bzurer11-Dec-03 4:31