RAD with Python

Ted Holden medved at fcc.net
Sun Sep 14 16:51:23 CEST 2003


John J. Lee wrote:


> I've never seen anybody give a convincing reason why GTk is a good
>  choice for *anything* except writing GNOME apps.


I've just been putting myself through a sort of a crash course in pygtk,
basically a text indexing and retrieval application which already existed
in several Windows and KDE versions as well as a Tcl/Tk version which
nobody had used in several years.  The idea is to have some sort of a
version of the thing which integraters could work with, i.e. written mainly
in a scripting language, and python appears to be several levels of
advancement above Tcl.

Tk appears to be more functional at this point, nonetheless there was no way
to avoid having Tk code twisted and wound in and around app code in the Tcl
version of the thing whereas the glade/gtk version of the thing produces a
gui in an XML file and an absolute minimum of anything related to guis in
my own code.  The difference between the two is about 400 lines of very
readable code versus about 2500 lines of code which was marginally
readable.  I'd MUCH rather deal with pygtk, and assume anything I might
miss from Tk will be there in a year.

The LINUX community appears to have developed a sort of a magical nexus of
things including gcc, swig, python, pygtk, glade, and zope which, taken
together, amounts to a new sort of programming paradigm.  The really big
piece of magic in the picture is swig.  This says that you you can write
library code in c++ and call it from a python program with a gui produced
by glade sitting there in an XML file, and all you'd ever need to
distribute to customers would be the .so file which swig produced, one or
two .py files, and the xml file.  That also says that if a customer wasn't
totally happy with some aspect of the interface, he could screw around with
it until he was.

I don't see an easy way to beat all of that.






More information about the Python-list mailing list