September 10, 2005
A few classes for multicast event handling have been previously released, and here's another. I think this one's unique in that it allows you to use both delegates and function pointers (although I may be wrong).

Read more or download it here: http://www.paperocean.org/cirrus.core.events.html

If you find any bugs, please let me know.

BTW: I've added various ways of chaining together delegates/function (opAddAssign/opSubAssign, opCatAssign, add/remove method calls). This way you can use the one you prefer.

John.


September 11, 2005
John C wrote:
> A few classes for multicast event handling have been previously released, and here's another. I think this one's unique in that it allows you to use both delegates and function pointers (although I may be wrong).

dcouple at dsource has them too :-)

What do you do when a recipient gets deleted?

Bastiaan.
September 11, 2005
"Bastiaan Veelo" <Bastiaan.N.Veelo@ntnu.no> wrote in message news:dg1e0u$2ihm$1@digitaldaemon.com...
> John C wrote:
>> A few classes for multicast event handling have been previously released, and here's another. I think this one's unique in that it allows you to use both delegates and function pointers (although I may be wrong).
>
> dcouple at dsource has them too :-)
>
> What do you do when a recipient gets deleted?

Good question ... my test program crashes and there doesn't appear to be a way to check if the delegate still exists - testing for null doesn't work. Also, I was unable to catch any exceptions. Ideas?

As a workaround, you can call 'remove' to break the connection in the receiver's destructor. But I'm not sure how this can be done when a function literal is attached, though.

I've noticed that Ben Hinkle's MultiDelegate exhibits the same problem. I've not tested dcouple.

Shouldn't the compiler null out delegates that reference nonexistent methods?


September 11, 2005
In article <dg1r4v$2tcj$1@digitaldaemon.com>, John C says...
>
>"Bastiaan Veelo" <Bastiaan.N.Veelo@ntnu.no> wrote in message news:dg1e0u$2ihm$1@digitaldaemon.com...
>> John C wrote:
>>> A few classes for multicast event handling have been previously released, and here's another. I think this one's unique in that it allows you to use both delegates and function pointers (although I may be wrong).
>>
>> dcouple at dsource has them too :-)
>>
>> What do you do when a recipient gets deleted?
>
>Good question ... my test program crashes and there doesn't appear to be a way to check if the delegate still exists - testing for null doesn't work. Also, I was unable to catch any exceptions. Ideas?

I'm not sure if this applies directly, but I use proxy classes for this purpose in C++.  The basic idea is something like this:

# class Proxy(Dest) {
#     this() {}
#     this( Dest d ) { attach( d ); }
#     void attach( Dest d ) { m_dest = d; }
#     void detach() { m_dest = null; }
#     void opCall() { if( m_dest ) m_dest(); }
# private Dest m_dest;
# }
#
# class C {
#     ~this() {
#         foreach( Proxy!(C) p; m_proxies )
#             p.detach();
#     }
#     void doSomething() {
#         Proxy!(C) p = new Proxy!(C)( this );
#         // pass proxy to callback
#         m_proxies ~= p;
#    }
# private Proxy!(C)[] m_proxies;
# }

>As a workaround, you can call 'remove' to break the connection in the receiver's destructor.

Sounds like you're doing something like the above.

>But I'm not sure how this can be done when a function literal is attached, though.

A delegate you mean?  There's not much you can do aside from perhaps putting the call in a try{}catch(Object){} block, as I think this will trap the access violation if the referenced object has been destroyed.

>I've noticed that Ben Hinkle's MultiDelegate exhibits the same problem. I've not tested dcouple.
>
>Shouldn't the compiler null out delegates that reference nonexistent methods?

If so then it may as well do the same for references to objects that have been cleaned up.  I think the issue is that this is simply too expensive to be worthwhile in most cases.


Sean


September 11, 2005
"John C" <johnch_atms@hotmail.com> wrote in message news:dg1r4v$2tcj$1@digitaldaemon.com...
> "Bastiaan Veelo" <Bastiaan.N.Veelo@ntnu.no> wrote in message news:dg1e0u$2ihm$1@digitaldaemon.com...
>> John C wrote:
>>> A few classes for multicast event handling have been previously released, and here's another. I think this one's unique in that it allows you to use both delegates and function pointers (although I may be wrong).
>>
>> dcouple at dsource has them too :-)
>>
>> What do you do when a recipient gets deleted?
>
> Good question ... my test program crashes and there doesn't appear to be a way to check if the delegate still exists - testing for null doesn't work. Also, I was unable to catch any exceptions. Ideas?
>
> As a workaround, you can call 'remove' to break the connection in the receiver's destructor. But I'm not sure how this can be done when a function literal is attached, though.
>
> I've noticed that Ben Hinkle's MultiDelegate exhibits the same problem. I've not tested dcouple.

Correct. When a programmer deletes an object by hand it is a promise that no other live references exist. Violating that promise results in crashes.

> Shouldn't the compiler null out delegates that reference nonexistent methods?


September 11, 2005
"Ben Hinkle" <bhinkle@mathworks.com> wrote in message news:dg24mg$2r4$1@digitaldaemon.com...
>
> "John C" <johnch_atms@hotmail.com> wrote in message news:dg1r4v$2tcj$1@digitaldaemon.com...
>> "Bastiaan Veelo" <Bastiaan.N.Veelo@ntnu.no> wrote in message news:dg1e0u$2ihm$1@digitaldaemon.com...
>>> John C wrote:
>>>> A few classes for multicast event handling have been previously released, and here's another. I think this one's unique in that it allows you to use both delegates and function pointers (although I may be wrong).
>>>
>>> dcouple at dsource has them too :-)
>>>
>>> What do you do when a recipient gets deleted?
>>
>> Good question ... my test program crashes and there doesn't appear to be a way to check if the delegate still exists - testing for null doesn't work. Also, I was unable to catch any exceptions. Ideas?
>>
>> As a workaround, you can call 'remove' to break the connection in the receiver's destructor. But I'm not sure how this can be done when a function literal is attached, though.
>>
>> I've noticed that Ben Hinkle's MultiDelegate exhibits the same problem. I've not tested dcouple.
>
> Correct. When a programmer deletes an object by hand it is a promise that no other live references exist. Violating that promise results in crashes.

Sounds like a fair rule. After all, if you're explictly calling delete you'll probably want to ensure everything else is cleaned up too.

>
>> Shouldn't the compiler null out delegates that reference nonexistent methods?
>
> 


September 12, 2005
John C wrote:

>>What do you do when a recipient gets deleted?
> 
> 
> Good question ... my test program crashes and there doesn't appear to be a way to check if the delegate still exists - testing for null doesn't work. Also, I was unable to catch any exceptions. Ideas?
> 
> As a workaround, you can call 'remove' to break the connection in the receiver's destructor. But I'm not sure how this can be done when a function literal is attached, though.
> 
> I've noticed that Ben Hinkle's MultiDelegate exhibits the same problem. I've not tested dcouple.


Dcouple solves this with Slot objects and managing code. I am ioning out some segfaults in version 0.3, but 0.2 should work.


> Shouldn't the compiler null out delegates that reference nonexistent methods? 


That would be nice, but how would the compiler know? There is a concept called weak references, that has been brought up a couple of times in this context. See the post "Resources" on the dcouple forum at dsource.


Bastiaan.
September 13, 2005
Bastiaan Veelo wrote:

> Dcouple solves this with Slot objects and managing code.

Version 0.3 (currently in the svn trunk) works per today (http://svn.dsource.org/projects/dcouple/trunk/managed/dcouple/). The dcouple version of Ben's example from multidg.d is given below

#import dcouple.connect;
#import dcouple.release;
#import dcouple.signalslot;
#
#class Foo : SignalSlotManager
#{
#  mixin SignalSlotManagement;
#
#  Signal!(int) valueChanged;
#  Slot!(int) changeValue;
#
#  private int val;
#
#  this()
#  {
#    valueChanged = new Signal!(int)(this);
#    changeValue = new Slot!(int)(this, &value);
#  }
#
#  ~this()
#  {
#    deleteSignals();
#    deleteSlots();
#  }
#
#  int value()
#  {
#    return val;
#  }
#
#  void value(int v)
#  {
#    if (val != v) {
#      val = v;
#      valueChanged.emit(v);
#    }
#  }
#}
#
#import std.stdio;
#int main() {
#  writefln("Compiled with ", dcoupleVersion() );
#  Foo a = new Foo;
#  Foo b = new Foo;
#  connect(a.valueChanged, b.changeValue);
#  writefln(a.value," ",b.value);
#  b.value = 11;
#  writefln(a.value," ",b.value);
#  a.value = 79;
#  writefln(a.value," ",b.value);
#  delete b;                                      // Note delete!
#  a.value = 100;
#  writefln(a.value);
#  return 0;
#}

(also available here http://svn.dsource.org/projects/dcouple/trunk/managed/examples/delete_example.d)

output is

Compiled with dcouple version 0.3, copyright 2004, 2005 Bastiaan Veelo.
0 0
0 11
79 79
100


So dcouple lets you tie together objects, and handle them any way you like regardless what ties exist. If you delete an object, existing ties
clean themselves up.

Bastiaan.
September 13, 2005
> #  ~this()
> #  {
> #    deleteSignals();
> #    deleteSlots();
> #  }

Does deleteSignals and deleteSlots reference other GC-managed objects or arrays? If so then you'll get random seg-v's during a GC. Objects are collected in random order and so the other objects might be gone by the time ~this runs. The details (as they are) are in http://www.digitalmars.com/d/class.html#destructors


September 14, 2005
Ben Hinkle wrote:
>>#  ~this()
>>#  {
>>#    deleteSignals();
>>#    deleteSlots();
>>#  }
> 
> 
> Does deleteSignals and deleteSlots reference other GC-managed objects or arrays?

Yes, but only to objects (Signals and Slots) that know about each other and about their managers (the ones calling deleteSignals and/or deleteSlots). In their destructor they deregister themselves with these referencing objects. So at the time of destruction, only references exist to objects that have not been destructed yet. Therefore, the order of destruction does not matter.

Bastiaan.
« First   ‹ Prev
1 2
Top | Discussion index | About this forum | D home