Re: [Java-gnome-developer] Which libraries are used?

[Clemens E. <Lin...@we...>]

Hello Jeff!

> 1) Simplifying the public interface of the library.  In the past
> the java-gnome public interface provided a one-for-one mapping to
> the methods of the native libraries.  This was a bad design decision
> for two reasons.  First, it made the libraries very large and
> difficult for a new developer to grasp.  Second, the public interface
> did not feel very java-like.  It did not leverage the existing classes
> from the JDK and did not follow the patterns that are common to
> java class libraries.  In the simplified version currently under
> development the classes will still have access to all of the
> methods of the native libraries via protected methods in the classes.
> A public interface will be built on top of that to provide a robust
> class library.
>
Hmm. I think the public interface of java-gtk0.7.1 is very clear and 
very simple. I started to learn swing these days, because I decided tow 
rite my vocab-programm in SWING (I'm really sorry, but thats the best 
solution I think), and I really wondered how complex swing is. I worked 
4 hours on an JTable and there arent any methods for asking the value...
So, I liked it very much that all java-gtk widgets are static and I dont 
think that they need to be more java-like.

> 2) Simplify the bindings themselves to make it easier for developers
> to contribute the project.  Time and time again developers have joined
> the project only to find out how complex the project is and eventually
> leave the project altogether.  If this project is to be a success I
> am going to need help.  Most people that want to contribute come from
> a Java background and know little about C or even JNI.  As a result I
> am restructuring the bindings by moving more of the code from the
> C/JNI layer up into the Java layer.  In fact, the C/JNI layer and the
> Java methods that will provide access to them will be fixed and in
> place.  All that will be necessary for most developers is to help
> design the public interface and write Java code that calls the native
> methods to implement that interface.  Unless the native access layer
> changes (which I hope will not happen often) there will be no need
> for the majority of developers to even compile the native libs more
> than once.  I have also added the files necessary to work in the
> eclipse IDE to cvs.  Eclipse will work on Linux and Windows.  If
> there is a Windows port of the bindings developers should be able
> to contribute from either platform.
>
Hmm, this semms to be a pretty solution. To make more things in Java 
will make the bindings smaller.

> 3) Simplify the build process.  When I first got started with java-gnome
> about two years ago I knew nothing about the GNU build tools.  As the
> project grew I continued to hack my way through the build process and
> managed to create a "beast of burden".  In this new version the generator
> will not be part of the build process.  The new build will provide
> four targets.  They will be jar, native, test, and all (the default).
> By default it will build all of the wrappers (libglade, libgladegnome,
> gtk, libgnome, and libgnomeui) but there will be configure options that
> will allow you to exclude parts of the bindings.  The test target needs
> a little more explanation.  I intend to build up a testing framework
> that will be based on JUnit to provide tests for all methods in the
> public interface.  This will help us ensure that we are producing a
> quality product.
>
Good idea! I had to build my libjavagtk.so myself because the 
build-scripts where broken.

> So the next question is "Where am I in this process?".  I have completed
> and checked into cvs the java and native code for glib, pango, atk,
> and gdk.  I have not provided the public interface for these libs yet.
> I hope to check in the java and native code for gtk today and the same
> for gnome and glade tomorrow.  Over the weekend I will complete the new
> build files.  At that point I need to start writing the public interface
> and tests for the bindings.  This will go much faster if I have help. 
> I have started an object model that is in cvs.  I would like to find
> somebody that would be willing to maintain this model.
>
> I think it would be very simple to compile the new version of the
> bindings on Windows.  I have already downloaded the Windows version
> of GTK2 and the associated libraries with the intention of trying
> to get this to work.  If you could do this that would be great.  Once
> you complete the compilation of the native layer you could help with
> the definition and coding of the public interface.
>
Of course I'm willed to help you!
But there are several problems:
1. I dont understand much C, nothing JNI, and my java isnt the best.
2. Time: I've to produce moch other apps for my school, and everybody I 
asked said to me, that I should do it with swing ;-((
3. Not sure about the future of java-gtk:
I came to the java-gtk project because I liked to produce compiled apps 
with gcj. But now gcj supports awt and I'm sure that a 
swing-implemetaion will follow in the next 1-2 years. The swing-part 
will be also built on the top of gtk.
So if I use swing I'll be compatible to all VMs and maybe soon the gcj, 
which would be also native.

What do you think about a swing-implementationon the top of gtk and how 
complex would that be?
What do you think of two possible java-gtk interfaces? the original and 
the swing....?
I also belive that that solution will strongly increase the number of 
java-gtk programms, because now many coders dont want to learn a new api.
I hope you are not angry because I said my thoughts, you're still the 
best developer I've ever seen.