C / C++ / MFC

Silly RichEdit question...

11-Jan-06 16:23

With "Debugging Tools For Windows" - The debugging tools from Microsoft. It contains a utility called "UMDH" can be used find leaks and the like in vanilla binaries.

Steve

KellyR11-Jan-06 12:36

KellyR

11-Jan-06 12:36

I'm trying to work with a RichEdit, and the code looks like this:

chat_window_rich.SetSel(nLength, nLength);
chat_window_rich.SetSelectionCharFormat(cf);
chat_window_rich.ReplaceSel(buff);

I don't want the edit window to automatically scroll to the bottom when a user is trying to scroll up and look at previously added text. For some reason, whenever a user has clicked in the edit window and a new chat message is received, the window is automatically scrolled to the bottom.

There are no calls to any scrolling functions anywhere.

If the user has scrolled up in the chat window but has clicked in some other window, for instance the typing window, the chat will NOT automatically scroll, it only happens when the mouse focus is on the window.

Anyone know what's wrong or how I can fix? Frown | :(

Kelly Ryan

Stephen Hewitt11-Jan-06 13:04

11-Jan-06 13:04

I'm no rich edit expert (I've never used one) but the general solution to this problem (you need to call a method that scrolls the window but you don't want the scrolling) is to turn off redraw using WM_SETREDRAW (which MFC wraps with CWnd::SetRedraw), call the functions you like, restore the position to where you want it then re-enable redraw. Not pretty but I've had to resort to this technique before.

There may be some easy way in this case however.

Steve

KellyR11-Jan-06 14:12

KellyR

11-Jan-06 14:12

I gave it a try, I can set the scrollbar in the right position but I can't get the darned window to scroll accordingly... ScrollWindow messages just don't seem to affect it, WM_VSCROLL messages don't affect it.. it just won't move.

Is there an easy way I can just tell it to set itself to the position dictated by its scrollbar?

Kelly Ryan

Stephen Hewitt11-Jan-06 14:29

11-Jan-06 14:29

I believe the EM_SETSCROLLPOS and EM_GETSCROLLPOS are what you're after.

Steve

Design patterns - simple questions.

KellyR11-Jan-06 14:38

KellyR

11-Jan-06 14:38

That worked, thank you very much Smile | :)

Kelly Ryan

Nick_Kisialiou11-Jan-06 12:08

Nick_Kisialiou

11-Jan-06 12:08

Assumptions and constraints:
I am working with C++, Object-oriented design. I am using two layers - UI and domain layer which contains application logic.

Question 1.

Problem formulation:
At the architectural level I need to separate UI from the application logic. At the design level I am using a facade/controller object in the domain layer. To track the updates of the state of domain objects I use Publisher/Subscriber design pattern, so that domain objects remain decoupled from UI but UI objects received notifications when a domain object state changes.

Question:
Is Mediator pattern more appropriate for this purpose?

Justification:

The downsides of Publisher/Subscriber pattern:

a) Domain object is coupled to the Publish interface.
b) When moving UI to a difference process or a different remote computer in the network, the publishing logic must be changed.
c) UI objects are coupled to the Publisher.

How do I suggest applying Mediator (Mediator-Singleton)? (There may be several different Mediators serving this purpose.)

a) During initialization stage each object registers with the Mediator to either send or receive a message with a certain MessageID. (Since this specific Mediator is Singleton it can easily be made visible from all domain and UI objects.)
b) When state of the domain object changes, the latter sends a message to the Mediator to inform it about the change.
c) Mediator stores the received message in its storage and forwards the forwards the message to all listeners waiting for the corresponding MessageID.
This scheme achieves the separation of concers - concern of message passing is moved to a separate class instead of domain objects. All objects are decoupled except for the Mediator. This design supports the reuse, and can be easily extended to a network application. For network applications it is a concern of two Mediators in both computers to exchange the messages, thus the design is connection independent.

So why is it not used, but Publisher/Subscriber is used instead?

Question 2.

Problem formulation:
Separation of UI and domain objects is usually implemented with ONE facade/controller object in the domain layer. For example, see Craig Larman, "Applying UML and patterns".

Question:
Do I need a similar facade/controller object in the UI layer? I feel like I do, but I have not seen people suggesting it, so I may be wrong.

Justification:
If UI is moved to a remote computer then all requests from UI objects will go through network, hence, all UI will need to be changed to accomodate for TCP/IP requests.
When UI grows large, when we use Operating System SendMessage(.) etc. the UI becomes a mess if we don't have a control over the sequential order of messages.

Do I make sense with my solutions or there is something that I have missed? Would appreciate any suggestions.

Thanks!

Nick Kisialiou

0x2B | -0x2B

Re: Design patterns - simple questions.

Curtis Schlak.11-Jan-06 13:53

Curtis Schlak.

11-Jan-06 13:53

My response for Question 1
The Mediator design pattern gets used to coordinate communication between many colleagues. This would apply to your situation if that communication needed to exist. The Publisher-Subscriber pattern exists for one-way communication, from the Publisher to the Subscriber. From your description of the application, I agree that you have used the correct design pattern for the situation. However, you may need to reconsider your implementation since so much depends on your Publisher interface.

My response for Question 2
Your Publisher-Subscriber framework should handle this. Instead of publishing to a local object (in the same process), it uses some sort of IPC. How this gets done depends on your implementation. However, if you've defined your Publisher-Subscriber interfaces correctly, you should have the ability to replace the classes that implement intraprocess communication with interprocess communication. Many RPC frameworks exist; for a starting point, check out Wikipedia[^].

"we must lose precision to make significant statements about complex systems."
-deKorvin on uncertainty

Problem with instantiating ActiveX Variant* Mething

godspeed12311-Jan-06 11:43

Re: Problem with instantiating ActiveX Variant* Mething

11-Jan-06 11:43

Hi,

I am running into some problems with this ActiveX function, I use the Class View to generate the method, it is all done through Visual C++ Wizards. Everything compiles, so say I create a method through the wizard as follows:

void CTestCtrl::Testing(VARIANT* test)
{
AFX_MANAGE_STATE(AfxGetStaticModuleState());

}

So I have not entered anything into the code.

Now I'm trying to access this function in Visual Basic .NET so I do the following

Private Sub Form1_Load(ByVal sender As System.Object, ByVal e As System.EventArgs) Handles MyBase.Load

Dim test123 As Object

AxTest.Testing(test123)

End Sub

When I run the VB program it crashes.

Any help would be greatly appreciated.

Stephen Hewitt11-Jan-06 12:04

Re: Problem with instantiating ActiveX Variant* Mething

11-Jan-06 12:04

Have you got any info on the crash? A stack trace and the exception generated would be a bug help.

Steve

godspeed12311-Jan-06 13:12

Re: Problem with instantiating ActiveX Variant* Mething

11-Jan-06 13:12

This is the error I'm getting

An unhandled exception of type 'System.NullReferenceException' occurred in mscorlib.dll

Additional information: Object reference not set to an instance of an object.

Stephen Hewitt11-Jan-06 13:16

Re: Problem with instantiating ActiveX Variant* Mething

11-Jan-06 13:16

Can I have a look at the .ODL for the method in question (I want to see that attributes are in place).

Steve

Stephen Hewitt11-Jan-06 13:37

Re: Problem with instantiating ActiveX Variant* Mething

11-Jan-06 13:37

My VB lets me down here but I suspect nothing is wrong. The Dim statement defaults to "Nothing" so all is probably working as it is meant to. I suspect is you changed to "As Object" to "As Integer" all would work well.

Steve

godspeed12311-Jan-06 16:59

Re: Problem with instantiating ActiveX Variant* Mething

11-Jan-06 16:59

Here is the part of the ODL file, this is in the method part

[id(9), helpstring("method Testing")] void Testing(VARIANT* test);

I also tried putting the Integer value there, but the same problem.

Stephen Hewitt11-Jan-06 17:10

Re: Problem with instantiating ActiveX Variant* Mething

11-Jan-06 17:10

I notice there are no attributes on "test". Given that you're using a pointer to a variant I would expect this to look like the following:

[id(9), helpstring("method Testing")] void Testing([in, out]VARIANT* test);

Steve

godspeed12311-Jan-06 17:56