Inheritance bug?

BlitzMax Forums/BlitzMax Programming/Inheritance bug?

therevills

(Posted 2011) [#1]

When I run this code:

Print "Start"

New MyGame.MyGameNew()

Type MyGame Extends App
	Method MyGameNew()
		Super.AppNew()
	EndMethod
EndType

Type App
	Method AppNew()
		New AppDevice.AppDeviceNew()
	EndMethod
EndType

Type AppDevice Extends GxtkApp
	Method AppDeviceNew()
	EndMethod
	
	Method Update()
		Print "AppDevice - Update"
	EndMethod
EndType

Type GxtkApp
	Method New()
		Update()
	EndMethod
	Method Update()
		Print "GxtkApp - Update"
	EndMethod
EndType

It outputs:

Start
GxtkApp - Update

I would have thought it would output:

Start
AppDevice - Update

Bug?!

beanage

(Posted 2011) [#2]

Interesting! Maybe the vmt is not fully setup when you hit the ctor of GxtkApp? What happens if you make GxtkApp::Update() abstract?

H&K	(Posted 2011) [#3]

Well you would be wrong in your thinking (Although understandably).

Update is called at the level of GxtApp, so its update on that level which is called.
That is it isn't an inherited object calling New, it is the base object

Oh and give better class names next time, like one two three or similar

therevills

(Posted 2011) [#4]

In Java:

package test;

public class Test
{
  public static void main( String s[] )
  {
    System.out.println("Start");
    new MyGame().MyGameNew();
  }
}

class MyGame extends App
{
  public MyGame( )
  {}

  public void MyGameNew( )
  {
    super.AppNew();
  }
}

class App
{
  public App( )
  {}

  public void AppNew( )
  {
    new AppDevice().AppDeviceNew();
  }
}

class AppDevice extends GxtkApp
{
  public void AppDeviceNew( )
  {}

  public void Update( )
  {
    System.out.println("AppDevice - Update");
  }
}

class GxtkApp
{
  public GxtkApp( )
  {
    Update();
  }

  public void Update( )
  {
    System.out.println("GxtkApp - Update");
  }
}

This prints:

Start
AppDevice - Update

ActionScript does the same... its only BlitzMax which runs the GxtkApp update...

BTW this is the Monkey generated code...

Czar Flavius

(Posted 2011) [#5]

In some languages, such as C++, calling a function in a constructor always calls the function assosiated with THAT class's constructor, even if the function is overridden in a derived class, and you have created a new instance of that class.

In other languages such as Java, it always calls the function assosioated with the actual type you are creating, even in a base constructor. I think C# also does the same.

This is a bit suprising - I thought Blitz would follow the latter, but it seems it's the former. So this isn't so much a bug as it is a gotcha.

You can get around it by creating your types from create functions and call the functions you want from them.

therevills

(Posted 2011) [#6]

You can get around it by creating your types from create functions and call the functions you want from them.

Yep, that's what I ended up doing, but it meant changing the Mojo's app.monkey file, which i did not want to do :(

Czar Flavius

(Posted 2011) [#7]

I didn't know monkey could generate Blitz code.

Yasha

(Posted 2011) [#8]

Perhaps this is relevant: http://www.blitzbasic.com/Community/posts.php?topic=66144#739775

Old, but it does address this topic.

Jesse

(Posted 2011) [#9]

I didn't know monkey could generate Blitz code.

That's therevills baby. He is actually creating the parser and translator(or maybe just the translator).
I think that's pretty neat.

Last edited 2011

Czar Flavius

(Posted 2011) [#10]

Thanks for the link Yasha. I'm sure I've read Java and maybe even C# don't follow the same behavior as C++ on this rule.

Sometimes it can be useful for the functions called in the constructor to be the derived type, but if the function uses fields which aren't set up yet it can be dangerous. It's tricky in C++ as you use the constructors a lot, and it would be nice to be able to choose whether to call the function virtually or not in the constructor.

In Blitz it's less of a problem. It's very common to use some kind of create function, which does its stuff after the object has been new-ed already. Blitz's new isn't as flexible as a C++ constructor (no parameters etc) so I think it's relied upon less than in C++ style.

I'm a big fan of OO and use it in all my projects, but I have to say the code above is an example of OO gone mad in my opinion. Impossible to follow!

Last edited 2011

Czar Flavius

(Posted 2011) [#11]

That's cool therevills! How are you solving problems of features available in monkey but not in BMax? Such as constructor parameters?