Re: [guppi-list] New features in CVS, feature-plan ?

[Jon Trowbridge <trow emccta com>]

On Mon, Feb 07, 2000 at 07:05:47PM +0000, Neil J Pilgrim wrote:
> How is the CVS version coming along ? Are there any new features not in
> 0.33.2 ?
> ie. Is it worth having a look in CVS for added features ?

Right now, CVS is basically a union of all of the old code and
a bunch of unintegrated new code --- if you build it, you basically
just get 0.33.2.

The infamous "new data model" code is in there in a semi-complete
form... it isn't totally complete, and it isn't debugged, and parts
are badly in need of optimization, but it is complete enough to merge
into Guppi proper.

I've got a boatload of changes that I've done locally, performing that
giant merge.  This has led to lots of breakage, and I'm waiting until
I get it all patched together, and then I'm going to commit it all in
one giant lump.

The other thing I'm hacking on is language binding issues.
Previously, all of the wrappers for the scheme bindings were
hand-coded.  (Which was very tedious, trust me.)  I've now switched to
a scheme like that used by gnome-guile, where there is a file
containing information on the functions and types, and a script
processes that file into C code.

The nice thing about doing it this way (besides removing the giant
PITA of coding guile bindings by hand)is that it lowers the entry
barrier for new scripting languages.  When we add Python bindings, we
just need to write a new script to auto-generate all of the C wrappers
needed to expose Guppi's API to Python.  It takes almost all of the
work out of creating and maintaining language bindings.

(Personally, I don't want to settle for just Guile and Python.  I'd
like to add Perl as well, and I should think that Java would also be
doable.  Something like Haskell would be tricker, just because there
is a fair amount of state floating around...)

The other issue is that the new data model has broken data importing
pretty badly.  This is OK, because I'm going to move data import and
export into plug-ins, which would have broken the existing stuff
anyway.:-)  It will all need to be set right.

> I might have a gander at the source at some point, but I've not used GTK
> before (mainly use C++). Is there a 'OO' documentation tool for GTK+ - I
> understand it is effectively OO in C, so is there an equivalent to doc++
> (or similar) for it ?
Hmm... I don't know anything about this.  Anyone?

> Is there a plan of which features are projected in which order/version ?
> Or just a plan of features (apart from those in other similar tools) or
> TODO ?

Right now, the TODO is intentionally left blank.  There is so much to
do that I find it depressing to try to enumerate it.

I have a Guppi "roadmap" written out in a notebook somewhere, but it
is a bit too speculative to post right now.  The short-term plan is
this: I'm hoping to release Guppi 0.34.1 in a week or two.  There
probably won't be any major user-visible changes, just a lot of "under
the hood"-type stuff: It will have the new data model integrated (with
data implementations in plug-ins), totally auto-generated language
bindings, and data import/export plug-ins.

Then for 0.35, I'll get back to work on the plotting.  The scatter
plot needs to be finished (it still lacks little details like axes),
and I'll add a few other simple plot types: line graphs, the dreaded
pie chart, etc.  I'm also going to be sitting back and waiting for the
user-contributed language bindings and import/export plug-ins. :-)

-JT

-- 
COMINT secret | Jam Echelon! http://www.echelon.wiretapped.net | NSA explosives
assassin terrorist car bomb tactical bin laden waco white house nuclear clinton