Forum: JRuby extension of java class throws internal jruby exception: HELP

Ok, there is a bug/feature in jruby such that the # of arguments to a 
ruby class ctor must equal the number of args for the java superclass, 
assuming one is extending a java super-class.   In:

http://jira.codehaus.org/browse/JRUBY-2457

it indicates that overriding self.new is a work-around.   The ruby side 
of this works, however when the interpreter tries to create a java proxy 
for the ruby obj it generates a null pointer exception.   Is there 
something special that needs to be done in the self.new initialization 
to make this work?   More generally, how can I subclass a java class 
where the ruby class has more ctor arguments than the java class?

The stack trace is:

Exception in thread "main" java.lang.NullPointerException
  at org.jruby.java.proxies.JavaProxy.getObject(JavaProxy.java:62)
  at org.jruby.java.proxies.JavaProxy.toJava(JavaProxy.java:335)
  at 
org.jruby.javasupport.JavaEmbedUtils.rubyToJava(JavaEmbedUtils.java:283)
  at 
org.jruby.embed.ScriptingContainer.runUnit(ScriptingContainer.java:458)
  at 
org.jruby.embed.ScriptingContainer.runScriptlet(ScriptingContainer.java:450)
  at test.nonunit.TestRubySubclassing.main(TestRubySubclassing.java:66)

And the code:

package test.nonunit;

import org.jruby.embed.ScriptingContainer;


public class TestRubySubclassing
{
  public static class Foo
  {
    public Foo (double a)
      { _a = a; }

    public double f(double a, double b)
    {
      return a*b;
    }

    private double _a;
  }

  public static void main (String[] argv)
  {
    final String code1 =
      "class A < Java::test.nonunit.TestRubySubclassing::Foo\n" +

        "def initialize(a)\n" +
          "puts \"initializing\", a\n" +
          "@foo = a\n" +
        "end\n\n" +

        "def self.new(*args,&body)\n" +
          "obj = self.allocate\n" +
          "puts \"allocated\",obj\n" +
          "obj.send :initialize, args[0]\n" +
          "obj.instance_variable_set(:@bar,args[1])\n" +
          "obj\n" +
        "end\n" +

        "def f(a,b)\n" +
          "return a*a+b*b\n" +
        "end\n" +
      "end\n\n" +
      "A.new(3,4)";


    ScriptingContainer script = new ScriptingContainer();
    Foo obj = (Foo)script.runScriptlet(code1);

    System.err.println ("value: " + obj.f (3.0, 4.0));
  }
}

On Mon, Mar 1, 2010 at 9:43 PM, Jonathan Shore 
<jonathan.shore@gmail.com> wrote:
> Ok, there is a bug/feature in jruby such that the # of arguments to a ruby
> class ctor must equal the number of args for the java superclass, assuming
> one is extending a java super-class.   In:
> http://jira.codehaus.org/browse/JRUBY-2457
> it indicates that overriding self.new is a work-around.   The ruby side of
> this works, however when the interpreter tries to create a java proxy for
> the ruby obj it generates a null pointer exception.   Is there something
> special that needs to be done in the self.new initialization to make this
> work?   More generally, how can I subclass a java class where the ruby class
> has more ctor arguments than the java class?

I *think* if you extend the Java class and call "super" with the
appropriate number of arguments for the Java superclass, it should
work. Is that what you tried?

FWIW, the current class-extension code is a really complicated rat's
nest of logic...if you can write some specs (a la
spec/java_integration in JRuby source repo) for specific extension
cases like this, we can try to fix them now and ensure they work in
the future.

- Charlie

---------------------------------------------------------------------
To unsubscribe from this list, please visit:

    http://xircles.codehaus.org/manage_email

I had read the bug report and thought that one needed to use the 
self.new workaround.   The most direct approach (your suggestion) 
worked, simply explicitly calling super() with the correct number of 
arguments.  Thanks!


On Mar 2, 2010, at 1:36 PM, Charles Oliver Nutter wrote:

>> has more ctor arguments than the java class?
> 
> - Charlie
> 
> ---------------------------------------------------------------------
> To unsubscribe from this list, please visit:
> 
>    http://xircles.codehaus.org/manage_email
> 
> 


---------------------------------------------------------------------
To unsubscribe from this list, please visit:

    http://xircles.codehaus.org/manage_email

We probably should warn or error if you *don't* call super for a
superclass constructor that requires arguments (obviously, calling the
default no-arg version for you is fine if it works. Can you file a bug
for that?

On Tue, Mar 2, 2010 at 7:03 PM, Jonathan Shore 
<jonathan.shore@gmail.com> wrote:
>>> it indicates that overriding self.new is a work-around.   The ruby side of
>> FWIW, the current class-extension code is a really complicated rat's
>>    http://xircles.codehaus.org/manage_email
>
---------------------------------------------------------------------
To unsubscribe from this list, please visit:

    http://xircles.codehaus.org/manage_email