[mlview-list] MlView activity News

[Dodji Seketeli <dodji seketeli org>]

Hello folks,

I have been away from emails for a long time now, unfortunately.
A lot of activity happened in the GNU Arch repository of MlView though.

We have done a lot since the last 0.7.1 release that happened in
November last year.


Here is a summary of what happened in the mlview--main--0.8 branch (that will give birth to the coming 0.8 release) :

* a new "create document" dialog 

* a new source oriented editing view. 
  MlView is no more a tree oriented (only) editor. Okay, we need
  to add a lot of editing feature to this new source view so that
  people writing docs can use it efficiently. 

* added some DBus stuffs so that application can be controlled remotely.
  this is an initial work to add a non regression tests framework.
  In the future, after compiling MlView, 'make tests" should run the
  editor, and launch series of tests to catch possible regressions.

*  A new firefox like search UI. This only works in the tree oriented
   editing view. The source editing view doesn't have any search feature yet.
   Any volunteer ?

* tons of bug fixes

All this changes are now committed into CVS HEAD.

In the mean time, we have been pondering switching the code base to C++.
Yeah, I know that will hurt some people. But I am tired of writing gobjects in
C, really. It's bug prone, long, and tedious. What makes the process tedious is
not the use of the C libraries. It more the need of concepts like
inheritance, polymorphism that makes the use of gobjects tedious in C.

Basically, we needed a language that 

	*let us write object oriented code easily, 
	 without re-writing the project. 

	* let us leverage easily on
	 the existing C code base around in GNOME. 

	* has a rather complete set of bindings in GNOME today. 

	* enforces type safety at compile time, 
	  to help catch more bugs.
	
To me, the best language that matches these requirements *today* is C++. 
Okay it has drawbacks. But which language doesn't ? 
All in all this is a matter of trade off.

So we started to port the entire MlView code base to C++.
The first step of this process was the compile the project using g++, keeping
the code unchanged, except for the parts that needed compilation errors fixing do
to the stronger type strictness of C++.
This first step is completed now.

The next step is to re-factor the former C classes (that derive from gobject)
and make them be good c++ citizens. This implies starting from the upper level
classes and going downward. I'd like this work to be done progressively so
that the code compiles and runs after the porting of each class. I think that
we should not enter a long "tunnel" where nothing works until all the classes
are ported. Moving forward through incremental changes is usually a good thing. 

All this c++ related changes are taking place in the mlview--cpp--0.8
branch, in GNU Arch. I haven't committed anything to CVS yet, 
and I don't think I will, unless some of you want it.

Furthermore, this work won't be merged back in the 0.8 line, but rather in a
further code line. Maybe, maybe 0.10, I don't know for now.

Only bug fixes and changes that affect classes that haven't been
ported to C++ will be merged in the mlview--main--0.8 code line.

Oh, and I would like to thank Philippe Mechaï, Stephane Wirtel 
and Benjamin Dauvergne for their invaluable help.

Best wishes,

Dodji