|
When reading about IDisposable, I get the impression that the consensus is that once you introduce it, everything that's even heard about it need to be made IDisposable too. Can this really be correct?
I have a class DriverWrapper that holds an IntPtr to a DLL with a C interface. This class is IDisposable, and implements the standard Dispose pattern, with a finaliser. It also has a number of abstract functions, some delegate definitions, and a helper function to get a delegate from the loaded DLL.
From this class, I derive some other classes, which implements the abstract functions by creating delegates to DLL functions. (The reason for this is that I have three related DLLs, all doing effectively the same thing to different bits of hardware, so to a user in the application it makes sense that they all have the same, shared interface. Unfortunately, the DLLs have different names, and in some cases different signatures, for their functions.)
Now, the derived classes are already IDisposable from DriverWrapper, but if they only hold delegates, which aren't IDisposable themselves, is there any need to override Dispose(bool) in these? I think not, but I'm no expert on C#.
(Oh, and I'm not concerned with the using idiom, as these will be alive throughout the lifetime of their owner. Neither am I interested in whether I should implement it in case I will at some future point have a managed and disposable member, or hypotheticals like that.)
Next question:
There is an abstract base class Model which takes a DriverWrapper as a constructor parameter, and takes ownership of it. The classes derived from this only override one function, and are just meant to provide different types of DriverWrapper-derived classes. I'll need to have access to the DriverWrapper for the whole lifetime of the Model.
abstract class Model
{
DriverWrapper theDriver;
Model(DriverWrapper driver)
{
theDriver = driver;
}
...
}
class DriverWrapper : IDisposable
{
~DriverWrapper()
void Dispose()
void Dispose(bool)
}
class DriverA : DriverWrapper
{ ... }
class DriverB : DriverWrapper
{ ... }
class DriverC : DriverWrapper
{ ... }
class ModelA : Model
{
public ModelA()
: base (new DriverA())
{}
}
Would Model have to be IDisposable? Surely the DriverWrapper is now managed, and will be tidied away when the Model is? If not, would ModelA-C need to override Dispose even if they have no further data?
(Honestly, at times like this I really miss RAII and typedefs from C++)
|
|
|
|
|
Any classes derived from a disposable class are already disposable so there is no need to override Dispose(bool) unless it holds additional resources, in which case it should override, free the additional resources then call base.Dispose(disposing).
Any wrapper classes that wrap a disposable class should be disposable so they can dispose the wrapped object properly.
|
|
|
|
|
Orjan Westin wrote:
When reading about IDisposable, I get the impression that the consensus is that once you introduce it, everything that's even heard about it need to be made IDisposable too. Can this really be correct? |
No, otherwise the whole framework would implement it.
Orjan Westin wrote: Would Model have to be IDisposable?
No, what you have is perfectly reasonable.
What can happen is that in taking control of when an object is disposed, it results in the impulse to chain IDisposable all the way up calling classes. Really you you only need to implement IDisposable where it makes sense: e.g. for memory performance or objects interacting with IO of some sort. As an example, I often write repository classes for EF/Linq TO SQL as the Datacontext is IDisposable and I want to pass the control of the datacontext's disposal up to the repository's controlling class I make the repository IDisposable but the calling class generally uses the using statement, or explicitly calls the destructor depending on the operations I am actually performing against the backing store.
|
|
|
|
|
Adding IDisposable never hurt anyone. A big part of me wishes Object had a Dispose and there was no interface -- I think that would be cleaner.
(This is not likely to be a popular opinion.)
|
|
|
|
|
PIEBALDconsult wrote: Adding IDisposable never hurt anyone. A big part of me wishes Object had a Dispose and there was no interface -- I think that would be cleaner.
May be not. You don't have to Dispose() all the objects. Only those which holds up resources that needs deterministic cleanup.
PIEBALDconsult wrote:
(This is not likely to be a popular opinion.)
True 
Best wishes,
Navaneeth
|
|
|
|
|
Adding IDisposable never hurt anyone
Actually, it does, by adding unneeded clutter and complicating the class.
|
|
|
|
|
Orjan Westin wrote: When reading about IDisposable, I get the impression that the consensus is that once you introduce it, everything that's even heard about it need to be made IDisposable too. Can this really be correct?
Not really. It depends on how you want to maintain lifetime of your resources. Only those who take ownership of resources should care about Dispose().
Orjan Westin wrote:
Now, the derived classes are already IDisposable from DriverWrapper, but if they only hold delegates, which aren't IDisposable themselves, is there any need to override Dispose(bool) in these? I think not, but I'm no expert on C#.
If it has nothing to dispose, you should be good with the base class implementation.
Orjan Westin wrote:
Would Model have to be IDisposable? Surely the DriverWrapper is now managed, and will be tidied away when the Model is? If not, would ModelA-C need to override Dispose even if they have no further data?
You can model it either way. If model is responsible for managing lifetime of a DriverWrapper, I'd put Dispose() to model which calls DriverMapper's Dispose(). But since model gets DriverWrapper through constructor, you would be constructing it somewhere else. In this, it'd be cleaner to model it like who creates DriverWrapper will dispose it when done.
More than that, if your DriverWrapper will be initialized when your application starts and needs to be alive till it ends, and in Dispose() all you do is reclaiming native allocated memory, then you probably shouldn't worry about dispose pattern because when the application exits, OS will reclaim all it's memory. And implementing dispose pattern would be an overkill.
Orjan Westin wrote: Honestly, at times like this I really miss RAII
That's what using keyword tries to do. But dispose pattern is different than RAII and not sure if both can be compared. A Dispose() call with never destruct a class. All it does is to provide a mechanism to release resource before GC does it by calling Finalize(). Where in RAII, the object will be removed and memory wlll be reclaimed. In .NET there is no way to do deterministic cleanup.
Best wishes,
Navaneeth
|
|
|
|
|
Orjan Westin wrote: Now, the derived classes are already IDisposable from DriverWrapper, but if they only hold delegates, which aren't IDisposable themselves, is there any need to override Dispose(bool) in these? I think not, but I'm no expert on C#.
No. As you say they already implement IDisposable through inheritance, and if they don't specify any expensive resources that you need to clean up before the GC gets around to collecting your object, then you don't need to specify additional dispose behaviour.
Would Model have to be IDisposable?
If you want it to be cleaned up before the GC notices it, then yes. And since it's wrapping a class which you already decided should be disposable, then so should the wrapper, and its Dispose should call the driver's Dispose. These two sets of classes seem so closely related that I wonder if you need to have two sets of classes at all, though.
If not, would ModelA-C need to override Dispose even if they have no further data?
No (see first point).
You need to implement IDisposable if you are allocating resources that you want to have guaranteed return at a known time on, but you don't have sufficient control over the context to explicitly call a close method to release them. It's also convenient to implement it when you do have that context, but you want to use the class in a using(x) block, or if you might forget you're supposed to call a close method (as Dispose will get called in GC cleanup too).
You don't need to implement it if the resources you're allocating in a class are so small and unlimited in number available that you don't mind if they don't get collected for a while. That applies to most managed objects, unless they're really huge and there's a memory impact (though the GC will notice this and run a collection sooner than otherwise), and to unmanaged resources that aren't limited (e.g. non-locking file handles, sockets, bitmap handles etc). The other case is that if you know you need a resource for the entire lifetime of your process you don't need to implement IDisposable either, as the one time that finalisers are guaranteed to be called is in process shutdown (assuming it's a clean shutdown), and even if it doesn't get run, the OS will clean up after you.
|
|
|
|
|
In my winforms project I have a PropertyGrid which can be changed at runtime. I need to capture the double mouse click events to open a dialog window where you entered the code of the event. Can anyone help me?
|
|
|
|
|
|
Hi Richard,
I am testing various methods and events of the PropertyGrid class but could not capture the double click an event, nor the automatic generation of the event name using the method CreateUniqueMethodName.
I need to intercept this action to open my dialogue will be informed where the event code.
Can you be more specific?
Thanks
|
|
|
|
|
Your original question stated: I need to capture the double mouse click events
If you check the documentation there is a method called OnMouseDoubleClick[^] and an event called MouseDoubleClick[^]. Is there some reason why neither of these will do the job?
|
|
|
|
|
I had already tested the events MouseDoubleClick and both DoubleClick and not fired.
myPropertyGrid_MouseDoubleClick void (object sender, MouseEventArgs e)
{
MessageBox.Show ("MouseDoubleClicked");
}
myPropertyGrid_DoubleClick void (object sender, MouseEventArgs e)
{
MessageBox.Show ("DoubleClicked");
}
|
|
|
|
|
It would appear that the documentation is not very clear. Since almost everything within the property grid is a control, they are all consuming the mouse clicks. I found the only place where it responds is by clicking in the small space between the main property table and the description below it.
|
|
|
|
|
Sorry Richard, but did not. I clicked all over the place and does not fire any of the two events (DoubleClick / MouseDoubleClick).
|
|
|
|
|
It's a very small gap between the two windows. When the mouse is over it the cursor changes to a little dark horizontal line with an up and down arrow on it.
|
|
|
|
|
Really, I see! Among the properties panel and the details of the component
|
|
|
|
|
Hello
I need some help. I am newcomer in C#.
The application is the PhoneBook. All the data serializable and stored in the List<>. Ithe main idea is to realize MVC pattern.
The question is how to implement events for deleting and editing contacts? Event for adding a contact is already exist and works.
To delete contact - inside the buttonDeleteContact_Click handler I need to get a selected object, than wrap it in DeletePersonEventArgs
and call the event (identical to AddClicked) DeleteClicked that will contail a code that delete a contact.
1. How to describe a type of Delete event ?
internal class DeletePersonEventArgs : EventArgs
2. What should be in
buttonDeleteContact_Click handler ?
3. What should be inside controller's method that handles the DeleteClicked event?
void form_DeleteClicked(object sender, DeletePersonEventArgs e)
The same issues for Update code.
Could you please provide a code example for this events and help me to realize this logic.
https://skydrive.live.com/redir?resid=45CC0C0F1CD563C9!281&authkey=!AIbigiVNGjyCQ9w here is the link for the source code - VS 2012
|
|
|
|
|
I am using distance method in the DBGeography to find the places within given raduis... but now I need to figure out whether the given geo coordinates (Lat, longitude) are whether within given geo fence or not... geo fence could be any shape (circle, square, ploygon etc... My question is that is there any classes or methods in C# to do that?
Thanks.
|
|
|
|
|
Effectively you're just going to do a PointIn... test. Given that you can represent pretty much anything as a polygon, you could use the techniques presented in this[^] article. It's fairly straightforward to create your own implementation.
|
|
|
|
|
this is in my c++ file
public ref class Class1
{
public:
int read_file
(const char * file_path,
hdf_call_vars_t & ret_vals) ;
};
and from c# I call it like so (unsafe code)
hdf_call_vars_t ret_vals;
string str = "C:\\a.h5";
byte[] bytes = Encoding.ASCII.GetBytes(str);
fixed (byte* p = bytes)
{
sbyte* sp = (sbyte*)p;
DoAT.Class1 cl = new DoAT.Class1();
cl.read_file(sp, ref ret_vals);
It does not compile sayiing the c++ is expecting a pointer to a structure. I thought making it an & is the same as passing by ref?
Input appreciated.
thanks,
sb
p.s. the c++ is a managed dll and I add it as a reference in the c# program.
modified 12-Feb-13 23:53pm.
|
|
|
|
|
bonosa wrote: It does not compile sayiing the c++ is expecting a pointer to a structure. I thought making it an & is the same as passing by ref?
It is indeed passing by ref. How is your hdf_call_vars_t looks like? For me the following code compiles and runs fine.
struct hdf_call_vars_t {
int foo;
};
public ref class Class1 {
public:
int read_file(const char * file_path, hdf_call_vars_t & ret_vals);
};
int Class1::read_file(const char * file_path, hdf_call_vars_t & ret_vals)
{
Console::WriteLine(ret_vals.foo);
return 1;
}
int main(array<System::String ^> ^args)
{
hdf_call_vars_t instance;
instance.foo = 10;
Class1^ c = gcnew Class1;
c->read_file("A string", instance);
return 0;
}If you are using C++/CLI, you don't have to use unsafe code in C#. Just wrap the C++ specific code in C++/CLI and return well defined managed objects which can be consumed directly from C#. After all, that's what C++/CLI is best at.
Best wishes,
Navaneeth
|
|
|
|
|
thank you Navaneeth.
The struct is
[StructLayout(LayoutKind.Sequential, Pack = 1)]
/* Returned values from read_file */
unsafe struct hdf_call_vars_t {
public channel_vars p_vars;
public channel_vars s_vars;
FILE_VERSION file_vers;
public int fetch_n;
public s_line_header_t * n_addr; /* malloc'd address of n data */
public UInt32 n_lines;
c_file_header_t hdr;
}
I added the structlayout line in the c# world.
Mine still doesn't compile.
I'm new to this so I don't know what is meant by wrapping unsafe code in c++ yet.
thanks so much,
saroj
|
|
|
|
|
Wrapping native objects means, you create a managed class in C++/CLI and hide your hdf_call_vars_t struct inside this managed class. Your library will only deal with this managed object and at the C# side you only have to deal with this managed object. This approach hides all the complexities of using native structs and simplifies your library's interface.
Here is working example of wrapping native object in a managed object.
using namespace System;
namespace Example {
struct hdf_call_vars_t {
int foo;
};
public ref class HdfCallVars
{
private:
hdf_call_vars_t* native;
public:
HdfCallVars()
{
native = new hdf_call_vars_t;
}
property int Foo
{
int get()
{
return native->foo;
}
void set(int value)
{
native->foo = value;
}
}
};
public ref class YourLibrary {
public:
int ReadFile(String^ file_path, HdfCallVars^ ret_vals)
{
return ret_vals->Foo;
}
};
}To use this library from a C# application,
using System;
using Example;
namespace ConsoleApplication1
{
class Program
{
static void Main(string[] args)
{
HdfCallVars vars = new HdfCallVars();
vars.Foo = 10;
YourLibrary library = new YourLibrary();
int status = library.ReadFile("filepath", vars);
Console.WriteLine(status);
Console.Read();
}
}
}
Hope that helps.
Best wishes,
Navaneeth
|
|
|
|
|
Thank you so very much for taking the time to answer in such detail, Navaneeth! I learned a lot from your code. I do have a question though. Since my hdf_call_vars_t is a complex structure containing members whiuch are instances of other structs, how do I treat those? For example, it has a member 'struct channel_vars p_data.' A channel_vars type struct instance contains in it ints, doubles, enums and a void * . I'll of course create get set accessors for Channel_vars as you have shown for the int foo in your example.But do I need to make Channel Vars a wrapper class also?
thank you,
saroj
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
