[Pipet Devel] [Fwd: Loci]

[J.W. Bizzaro]

>  > I just took a look at it, and it looks good.  Thomas Sicheritz from Sweeden (one
>  > of the early volunteers) was going to make an editor in Python-GTK but has now
>  > switch to a phylogenetics project, PyPhy, which you probably saw mentioned on
>  > the list.
> I'd be open to providing this functionality.  Would it be a core component
> or a plug-in? Plug-in I would assume.

I'd like to carry the plug-in model to an extreme.  In fact, it is hard to
identify one program that represents Loci and performs the bulk of the work. 
This is not so with most applications that use plug-ins.

>  > We probably won't be able to avoid C altogether, so your experience will be
>  > helpful.  Have you looked into Python?  Most people who have, including myself,
>  > have been amazed at just how nice it is to work with.
> I know of it, but have no experience with it.  A question for those of you
> who have experience, how portable is it from one UNIX system to another?

With the exception of the GTK bindings, Python is as portable as Perl.  It may
be more portable than Tcl/Tk, but I'm not the best judge of that language.  Of
course I'm talking about the non-GUI Python scripts.  And to my knowledge,
Python exists on nearly every UNIX variant.

> In general, I have had much greater success in recompiling C source code
> than source from other languages.  (Of course this may be due to my
> greater familiarity with C or that the C code my be written better?)

I have found compiling C, even between Linux variants, to be unpredictable with
about 1 in 4 apps compiling on the first attempt.  And C is better at this than
C++.

With Python, you can make changes to the code at runtime :-)  And of course no
compiling.

> What is the status of developing a LocusML? In reexamining my own work, I
> think a well defined DTD would greatly aid in the development of widgets.
> IE for the multiseq ed it would be great if the widget itself understood
> xml info for the sequence as well as the various markup. Is the plan to
> modify BioML or start from scratch?

We have had some lengthy discussions about LocusML.  The people interested in
its development and most qualified to do so are Justin, Humberto, Tim, and
Rahul.  I may be forgetting someone.

We will actually be defining several DTD's, which may be derived from other
XML's:

    Sequence definition:  BioML + BSML
    Structure def:        mmCIF/XML + some CML
    Phylogeny def:        ??? probably make our own
    Database query def:   maybe from BLAST/XML      
    Workflow def:         make our own
    GUI def:              from GLADE/XML
    Graphics def:         maybe from some KDE programs

As far as graphics are concerned, the idea is to define vector (and OpenGL?)
graphics so that custom made schematics can be generated by the plug-in, and the
plug-in developer can choose some high-level representations...sort of like
graphical widgets.

> It would seem to me that a quality set of bioinformatic widgets written in
> C would be a tremendous resource to several bioinformatics projects going
> on.  With python wrappers, loci plugins could use them, with perl
> wrappers, bioperl users could use them, etc ...

Yes.  Agreed.  And they would be considerable faster in C.  Maximum reuse for
the components of Loci is an important consideration.  It would be nice if the
widgets could be compiled into unrelated applications.

>  > We are using PAOS, which holds some advantages for us.  But we are considering a
>  > CORBA interface that connects the whole of Loci to any CORBA client.
> What is PAOS? How is it simmilar/different from CORBA?

Check out the local home page for PAOS:

    http://bioinformatics.org/paos/

How it compares to CORBA is not an easy thing for me to answer, since I have not
worked with either for very long.  But I do have some thoughts about the
advantages of PAOS, which I will put in the next e-mail.

> Yeah, I saw the post.  I think it is a great idea. I would think this
> might be another reason for writting the widgets in C.  This would allow
> use of the glade lib modified to accept xml info for the bioinformatic
> widgets.

The trick is connecting the GTK signals to handlers in the plug-in itself.  And
yes to add bioinformatics widgets to the collection.  You may be the best man
for that job too.

I think it is important to develop an SDK concurrently with Loci.

> I also have some experience using PHP and mysql for developing web
> interfaces. Of course, once commandline tools are in place, this shouldn't
> be too difficult.  I have used phplib for doing session management.

That's good.  Rahul volunteered to help with the Web interface.  We may create a
separate CVS module and maillist in a little while.

> At this point, working on my own machine for development would be easier.
> If/When my widgets/programs become useful I would certainly consider
> distribution through TOL. Does your web server allow for php scripts?

Sure.  It's a PentiumI running RedHat Linux 5.2.

> Sounds like I should jump right in.  How can I get CVS access (The only
> cvs experience I have is downloading the current wine code from CVS.)

I can create an account for you.  There are a few important CVS commands, which
I can send to you with your account info.


Cheerios.
Jeff
-- 
J.W. Bizzaro                  mailto:bizzaro at bc.edu
Boston College Chemistry      http://www.uml.edu/Dept/Chem/Bizzaro/
--