Interfaces, Big Deal!!

[Franco Gustavo]

Hi,



Please help me to understand this, because I don't see what I'm missing.



I was reading a lot of examples on Internet that explain that C# doesn't
implement multiple inheritance it implement multiple interfaces.



So, I keep reading and reading and reading to see how can I resolve my
problem with interfaces.

The conclusion is I'm writing to all of you because I couldn't find the way
to resolve my problem.



If I can create multiples interfaces and I have to implement the methods on
the derived class, then what's the big deal.



I know if I declare a class that inherit from one interface is useful
because for example

I can create an Interface called Shape, and 3 Classes called Circle,
Rectangle, and Line.

If on my interface I declare my method Draw(); and I implement on each
derived class then I could do the next.



IShape s = (IShape) AnyShapeObject;

s.draw();



I think is very useful and I used many times. But now the question is what
is the similitude between Multiple Inheritance and Multiple Interfaces???? I
don't see too much



Why on Internet always speak for Multiple Inheritance and Multiple
Interfaces like they have a relation???



My Case:



I have many Windows Control and I'm inherit each one for create my own
Controls.

For example I want to add method called getVirtualSize();



I create and interface

IVirtualSize



It contain

public getVirtualSize();



So, on my derived class I MUST implement my method getVirtualSize and that's
it.



Very nice!!!



Now the real question is, I create a lot of logic on my paint method; this
logic is exactly the same for all derived classes, why do I have to
implement this entire logic on each Control???



With multiple inheritances you just create a class
ExtendedControlFuncionality and Implement a method OnPaintEx, and really I
have a nice source code, because my class ExtendedControlFuncionality has
all the new extended functionality.



Then how can I do the same with interfaces????



I don't want to put my piece of code on every derived class, it is a mess
and very difficult to maintain, I want to centralize my source code and in
the same time I want to use many methods of the base classes, that means I
want to use all the protected methods that the controls offer to me because
OnPaintEx needs those.



Really, here is only one example, the reality is that is a lot of source
code that I "should" implement on every Control, I don't want to do that.



Thanks, the answers will be really appreciated.

Gustavo.

 

[Sherif ElMetainy]

Hello

Unlike C++, C# doesn't support multiple implementaion inheritance. But your
problem can be solved by deriving the class ExtendedControlFuncionality from
Control and deriving your class from ExtendedControlFuncionality.

Best regards,

Sherif

 

[Franco Gustavo]

But Control doesn't do anything that I want.

Basically, I want derive from Label, Panel, GroupBox, CheckBox, RadioButton,
etc. and add my extended funcionalities

The Class control doesn't have any funcionality.

On that case if I derived from Control I have to rewrite each windows
control from scratch wich is "not possible", doesn't make sense.

Thanks,
Gustavo.

 

[Kieran Benton]

So just have a class that has private members of all those types. Inheriting
from all of the GUI classes does not make sense as they are all functionally
different.

Kieran

 

[Anon]

Can you implement the logic into a function (if VB) or static method in a
class (VB or C#). Then for a control, implement your new interface and in
the method of the implemented interface just pass the call onto the function
or shared method?

 

[Fergus Cooney]

Hi Kieran,

|| So just have a class that has private members of all those types.

I'm not sure how that would work.

|| Inheriting from all of the GUI classes does not make sense as
|| they are all functionally different.

That's exactly the point - disparate objects but which require a <common
Capability>. Franco would have a FrancoLabel which would inherit from Label
and CapCtrlExtensions, and a FrancoPanel - Panel and CapCtrlExtensions, etc,
etc.

Using the interface approach means that the code has to be duplicated in
each of the ingheriting classes. Using multiple inheritance means bolting the
Capability on. Write once, re-use - it's what OOP shouts from the rooftops.

Regards,
Fergus

 

[Franco Gustavo]

Yep, I tried that, but if you pass the control to a function on different
class (static or not static) then you can't get access to the protected
methods from the control.



Basically, My application is skinned, and I needed "true" transparent
controls.



Windows Problem

1) On Many controls you can't set Background like transparent,
(Color.Transparent)

2) Controls like GroupBox they capture the background wrong (if you have a
gradient background on the parent windows and the GroupBox is on the
position 50,50 then the background is capture from the Parent from the
position 0,0 which is wrong, because the GroupBox doesn't looks like
transparent in at all, has the gradient translated)

3) On the Controls that accept Color.Transparent, if you have a control on
Color.Transparent color and there is another control behind it (is not child
or parent, they have the same parent), Windows just capture the parent
Background which is wrong again, because you can't see the control that is
behind him.



So, what I did is inherit every control of windows, create my own BackBuffer
where I keep all my window status and changes, avoid call to
OnPaintBackground, and I take the process to draw the portion that windows
ask me to redraw with my own BackBuffer.



So, All this process is standard for every control and work really good,
then I can get "true" transparency and you don't see the difference on
performance.



But the code is ugly, because I have the same method implementation like 10
times!!!



Then the only that I need is consolidate my functionality on just one single
Class, Interface, Method or whatever I need.

On C++ if you do multi-inheritance then I can get what I'm looking for.



So, there is some way to emulate multi-inheritance on C#?????



I though interface was the answer. I see may be interface can't do too much
here, then some idea???



Thanks,

Gustavo.



P.S.: I think this is a very commun problem where you have to add
funcionality to thirdparty classes where you have to inherit from each one
of them and you have to share code for all of them.

 

[Fergus Cooney]

Hi Anon,

In this instance, the problem with off-loading the functionality to
another Class or Module and passing the target object in, is that it would not
have access to the target's protected and/or private members.

In any case, that is just one example that frustrates Franco - as a
general solution there would be many cases where it's a workable but messy.

In my main response to Franco, I mention a linked-list Capability that I
have had to implement using an Interface. This does actually work by sending
the object to a worker Class to traverse the list. The method works, but it
suffers because I have to make the actual link fields, Prev and Next, public,
as my worker needs access to them. There are also issues with casting and
type-safety. I can't prevent a programmer from inserting a Widget into a
linked list of Thingies. As I'm the programmer, that's not going to happen,
but I would prefer to have my carpets tacked down rather than watch my feet.

Regards,
Fergus

 

[Fergus Cooney]

Hi Franco,

I'm entirely in agreement with you on this issue. Yes, Interfaces have
their (many) uses. But so does multiple inheritance.

My case in point - I create many objects with a totally different nature.
But I want them all to be capable of being in a linked list. I don't want to
have to create a separate class to manage these links. I don't want a
collection or an ArrayList. I simply want a Next and a Prev and a couple of
other trivial bits.

I'd like to define a small linked-list Capability class which I bolt-on to
my other classes through multiple inheritance.

But I can't. :-((

My only recourse is to define an interface, CapLinkable, and <copy> the
code (it's perfectly generic - needs not even a single character to be
changed) into <every> class that is to have this Capability. [I only have this
in C# because I can do it in about 15 lines - in VB it would probably be about
75!!]

But it's a maintenance nightmare, for I cannot even do it with <include>
files - it has to be a textual copy each time!!

There are many small pieces of functionality that would work for many,
many disparate objects without code change required - or in your case, minimal
changes.

Another example is the ability to Serialise. This could be done through
the inheritance of Capabilities. But no, it has to be done by setting an
Attribute and having the serialization done behind the scenes as a fiddle by
the Attribute in conjunction with a Formatter. Why can't an object serialise
<itself> by inheriting from a CapSerializable Capability?

It would be very useful if there were a mechanism for bolting these
Capabilities onto a class without having to reimplement the wheel each time.

Regards,
Fergus

 

[Jay B. Harlow [MVP - Outlook]]

Franco,
As Sherif stated, .NET does not have Multiple Implementation Inheritance,
one common method to get around this is to have an Implementation class
along with your interface. You would have an IShape interface along with a
ShapeImpl implementation class. Your Circle class would implement the IShape
interface, then contain a ShapeImpl object as a field, the IShape method
would delegate to the ShapeImple class. NOTE, Circle would NOT inherit from
ShapeImpl.

Have you considered using creating an Extender Provider class instead?

An Extender Provider is a class (can be either a Component or a Control)
that implements the System.ComponentModel.IExtenderProvider interface
allowing you to add new properties to controls. Along with these properties
you can attach event handlers to the targeted controls enabling you at add
functionality.

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

http://msdn.microsoft.com/library/d...de/html/cpconimplementingextenderprovider.asp

http://msdn.microsoft.com/library/d.../html/cpconwinformsextenderprovidersample.asp

Hope this helps
Jay

 

[Franco Gustavo]

Exactly, You know perfect what I'm talking about.



In fact on my example all the base classes have a class in common,

"Control", and that is exactly what I need to change because all the OnPaint
and OnPaintBackground are coming from Control Object.



So is not crazy the idea that if I have Label, Panel, and GroupBox coming
from a common class "Control" have the possibilities to change the behavior
of the Base class.



And having this functionality still I think is good OOP



Thanks,

Gustavo.

Hi Franco,

I'm entirely in agreement with you on this issue. Yes, Interfaces have
their (many) uses. But so does multiple inheritance.

My case in point - I create many objects with a totally different nature.
But I want them all to be capable of being in a linked list. I don't want to
have to create a separate class to manage these links. I don't want a
collection or an ArrayList. I simply want a Next and a Prev and a couple of
other trivial bits.

I'd like to define a small linked-list Capability class which I bolt-on to
my other classes through multiple inheritance.

But I can't. :-((

My only recourse is to define an interface, CapLinkable, and <copy> the
code (it's perfectly generic - needs not even a single character to be
changed) into <every> class that is to have this Capability. [I only have this
in C# because I can do it in about 15 lines - in VB it would probably be about
75!!]

But it's a maintenance nightmare, for I cannot even do it with

 

[Lloyd Dupont]

I 100% agree Jay B. Harlow, It's my solution either.

let's speak about your transparency problem now, you have 2 things to set
(not one)
1. a transparent background color
2. SetStyle() there is a transparent flag.

now if you have a top panel which is transparent, and use .NET 1.1 all your
child control would be correctly transparent.

 

[Franco Gustavo]

As Sherif stated, .NET does not have Multiple Implementation Inheritance,

one common method to get around this is to have an Implementation class
along with your interface. You would have an IShape interface along with a
ShapeImpl implementation class. Your Circle class would implement the IShape
interface, then contain a ShapeImpl object as a field, the IShape method
would delegate to the ShapeImple class. NOTE, Circle would NOT inherit from
ShapeImpl.

Ok, I see, on this case Interface and Implementation are coming on couple
and I have just one implementation for my code.
But still I don't see how ShapeImpl will access to my protected methods and
private variables from Circle because is a object into the Circle object.

Have you considered using creating an Extender Provider class instead?

An Extender Provider is a class (can be either a Component or a Control)
that implements the System.ComponentModel.IExtenderProvider interface
allowing you to add new properties to controls. Along with these properties
you can attach event handlers to the targeted controls enabling you at add
functionality.

I didn't know about that and how it works, I'll read that and try a shot.

Thanks very much,
Gustavo.

 

[Franco Gustavo]

Wow, do they fixed??

I'm using 1.0 and I set for my controls

this.SetStyle(ControlStyles.AllPaintingInWmPaint, true);
this.SetStyle(ControlStyles.UserPaint, true);
this.SetStyle(ControlStyles.Opaque, true);

And doing all the logic.

So, that means if I use VS2003 wich compile with 1.1 can I :
Create a Form, put a Background Image on the Form
Create a Panel and set Color.Transparency like a child of the Form
Create a Radio button like Child of Panel

And the background for the radio button will be the form Backgorund????????

May you confirm that? if that's true is a excelent news and I'll switch to
VS2003 the soon as possible

Thanks,
Gustavo.

 

[Rick]

Check out the "decorator pattern". That might be what you're looking for
(at least how I interpretted your question)

Rick

 

[Jay B. Harlow [MVP - Outlook]]

Franco,

Ok, I see, on this case Interface and Implementation are coming on couple
and I have just one implementation for my code.
But still I don't see how ShapeImpl will access to my protected methods and
private variables from Circle because is a object into the Circle object.

Are the protected methods inherited or added to Circle. If you added them to
Circle should they have been added to ShapeImpl instead?

Are they private variables added to Circle itself, should they have been
added to ShapeImpl instead?

For methods you will need to structure the method such that the
Implementation class (ShapeImpl) will be able to see it. Either via an
interface of Friend.

Alternatively ShapeImpl can have call backs (either delegate fields or full
events) that it uses to call into Circle when it needs information from
Circle.

For variables I would either make them a member of the Implementation class
itself or pass them as parameters when I call into the Implementation class.
Normally I would set it up such that each instance of Circle had its own
ShapeImpl, which means that ShapeImpl can hold per instance variables of
Circle, in which case ShapeImpl would not need to access details of Circle.

Hope this helps
Jay

 

[Franco Gustavo]

Are the protected methods inherited or added to Circle. If you added them
to

Circle should they have been added to ShapeImpl instead?

Are they private variables added to Circle itself, should they have been
added to ShapeImpl instead?

No, they belong to the base class.



I see how to do it with callback functions, is not so messy and can work
fine.



Anyway all those ways are an alternative, is like there are no a clear way
how to do it without insert a trick o working around.



So, Then if multiple inheritance and multiples interfaces implementation "I
see" are really different, why on 95% on the web site speak like they have a
relation?

Thanks very much, and I'll try to do it on the clear possible way.

Franco,

Ok, I see, on this case Interface and Implementation are coming on couple
and I have just one implementation for my code.
But still I don't see how ShapeImpl will access to my protected methods and
private variables from Circle because is a object into the Circle object.

Are the protected methods inherited or added to Circle. If you added them to
Circle should they have been added to ShapeImpl instead?

Are they private variables added to Circle itself, should they have been
added to ShapeImpl instead?

For methods you will need to structure the method such that the
Implementation class (ShapeImpl) will be able to see it. Either via an
interface of Friend.

Alternatively ShapeImpl can have call backs (either delegate fields or full
events) that it uses to call into Circle when it needs information from
Circle.

For variables I would either make them a member of the Implementation class
itself or pass them as parameters when I call into the Implementation class.
Normally I would set it up such that each instance of Circle had its own
ShapeImpl, which means that ShapeImpl can hold per instance variables of
Circle, in which case ShapeImpl would not need to access details of Circle.

Hope this helps
Jay


with

a

Ok, I see, on this case Interface and Implementation are coming on couple
and I have just one implementation for my code.
But still I don't see how ShapeImpl will access to my protected methods and
private variables from Circle because is a object into the Circle object.


I didn't know about that and how it works, I'll read that and try a shot.

Thanks very much,
Gustavo.

with

a

http://msdn.microsoft.com/library/d.../html/cpconwinformsextenderprovidersample.asp

resolve on
each question
is my
own getVirtualSize
and

is

a

 

[Jay B. Harlow [MVP - Outlook]]

Franco,

Anyway all those ways are an alternative, is like there are no a clear way
how to do it without insert a trick o working around.

If multiple implementation inheritance is important to you, and you prefer
to target the .NET platform you may want to consider switching to Eiffel
http://www.eiffel.com. It supports multiple implementation inheritance on
the .NET platform. Of course using Eiffel's multiple implementation
inheritance may limit your class libraries to Eiffel. I don't use Eiffel, I
have just read about it on occasion.

So, Then if multiple inheritance and multiples interfaces implementation "I
see" are really different, why on 95% on the web site speak like they have a
relation?

As has been demonstrated Multiple interfaces & delegation is how you
implement multiple inheritance in .NET. So obviously they are related.

If you take implementation out of the picture. In C++ Multiple Class
Inheritance defines a derived type that is composed of multiple base types.
In .NET Multiple Interface Inheritance defines a derived type that is
composed of multiple base types.

In both cases you are defining a derived type that is also "is a" more than
one base type.

In either case you can use the derived type anyplace any of the base types
can be used.

So I'm really not seeing how you consider them not to be related!

Implementation is just one facet of the equation.

Hope this helps
Jay

 

[Fergus Cooney]

Hi Jay,

|| As has been demonstrated Multiple interfaces & delegation is
|| how you implement multiple inheritance in .NET. So obviously
|| they are related.

They <are> related in that they use the words multiple and inheritance but
the <quality> of that inheritance is radically different. As you say above
".... is how you implement...". With multiple inheritance of base classes you
<don't> implement - the implementation is carried forward from the base class.

|| If you take implementation out of the picture.

Exactly. But what a big if!!

|| Implementation is just one facet of the equation.

Implementation is the first-order component in the equation!!

To me, the the whole point of multiple class inheritance is bolt-on
implementation. I define my Widget class with three words: Inherits
BaseWidget, CapLinkable. Just that, and I've got a Widget which has
linked-list capabilities and I haven't had to mess around with worker classes,
delegates or anything else.

Using a ShapeImpl, or another worker/helper, with call-backs to get
protected and private members will work, sure, but it's a workaround - around
the gap that multiple class inheritance would have filled.

You have to define the IShape interface. Then you have to implement it,
and, as it's going to be reuseable you implement it in ShapeImpl. But every
class that wants it still has to implement IShape to instantiate and call its
ShapeImpl. Then everything that ShapeImpl needs to do that accesses
private/protected members requires either a Delegate or a part of the IShape
interface - dedicated just to this inter-object communication - while the
target object has to have receiving methods which do their thing and package
up the results and passes them back.

All of the above compared with: Inherits BaseClass, ShapeExtensions,
OtherUsefulStuff.

Hey, I added an extra word and got a whole bunch of other useful stuff.
;-)

Well, that's my moan about it. I'd like the language/framework/compiler
people to put the hard work in so that I can have reusable Capabilities. I'm
not asking for much am I?**

Regards,
Fergus

** Yes, I know - I am! :-D

 

[Jay B. Harlow [MVP - Outlook]]

Fergus,
I am not disagreeing with you or Franco. Multiple Implementation Inheritance
is a very useful tool.

The problem is .NET does not offer it (1.0 or 1.1). Period. Based on the
published roadmap, it will not be offered in .NET 2.0.

Whether it will be offered in .NET 3.0 is anyone's guess.

Franco is trying to understand how in .NET they are related, I am trying to
explain to him how they are related.

So there is no need to get in some academic debate about how great & wonder
Multiple Implementation Inheritance is and how Multiple Interface
Inheritance is not related. I don't buy it.

I'm out of this conversation.

Hope this helps
Jay