[ghemical-devel] about gnome2, and moving to a single "project" class.

Tommi Hassinen thassine@messi.uku.fi
Wed, 8 May 2002 17:55:50 +0300 (WET)


Hello all.

I have studied the gnome2 situation a bit, and found that at the moment
the only way to get a gnome2 environment is to build it from sources.
Therefore I haven't seen the gnome2 work that important yet, and I
probably wait until it appears to the debian unstable distribution before
I more seriously start looking at it.

So instead of gnome2, I have been working with an another idea. In my own
work with the simplified forcefield, soon I would like to make some mixed
models, where one part of the system has the simplified presentation and
the rest have the traditional all-atoms presentation. In a similar way,
mixed molecular mechanics/quantum mechanics models have been developed
and used for some time.

So, when planning this, it soon became evident to me that the best way of
implementing the mixed models would be to combine all our current 3
"projects" to a single "project". This single project would be like our
current molecular mechanics project (containing atom and bond objects).
This would cover the MM and QM all atoms models pretty well, but later I
need to modify my simplified models so that they also fit into this
environment (I plan to mark them as special dummy atoms and bonds). Then a
quite big rearrangement to the "engine" classes would be needed. The
properties of current MM and QM "engine" classes need to be combined, and
a new set of "setup" classes are needed to contain the information about
how the model is divided in the submodels (say, QM/MM and MM/simplified
forcefield). The "setup" class would become at the middle of our old
"model" and "engine" classes, so that the "setup" objects would finally
create the "engine" objects.

Yep, this might sound a bit messy, but I have already done the most of the
dirty work of this modification, and obtained a first version that
actually compiles and starts.  :)  It basically just compiles and starts
really, so for example you can't compute anything with it yet.

Many many modifications were needed, and it seems to be better to not
include the modifications to CVS yet. Instead, I have made a "preview"
version of the new code available at the download page. Please see:

http:/www.uku.fi/~thassine/ghemical/download/archive/ghemical-ONLY_ONE_PROJECT-preview-2002-0508.tgz
http:/www.bioinformatics.org/ghemical/download/archive/ghemical-ONLY_ONE_PROJECT-preview-2002-0508.tgz

if you wish to take a look at the new design. But please note: at the
moment it's quite messy, undocumented and still subject to many
changes. Also many features are now broken and/or disabled.

I think that at the moment this new design seems quite
promising:

	+ it would allow the mixed model described above in future.
	+ the user interface becomes simpler : only a single project type.
	+ in future, there would be only a single file format.
	+ in the code, for many classes the multiple implementations are
		gone and only one implementation remains, for example:
			- model, docview (now renamed to "project")
			- geomopt, moldyn
			- ribbon

So, what would be the future of these modifications? As I decribed above,
it seems to me that this approach might lead into simpler user interface
and code in the future, but it will also take some time for me to clean
things up and bring this alternative system to the same level of
functionality as the current v0.83 has (it takes the whole summer, at
least).

So, I think nothing changes at short term. I will keep on working with
the new design, and I will release more preview versions, perhaps weekly
or so. Let's see where this new desing leads to.

Regards,

	Tommi