[MATRIX-SIG] [long] A NO-Proposal for a plot-package

[David Ascher]

On Mon, 21 Jul 1997, Andrew P. Mullhaupt wrote:

> I've been watching the discussion here, and it seems fine to split
> plotting away as a discussion from arrays, except for one thing.
> 
> Maybe some kinds of plots are pretty hard to do without the array
> objects.

Definitely -- it's just that we can define those operations using NumPy
arrays without having to modify the architecture of NumPy, hence it need
not be discussed on the same SIG, or so I take Geoffrey's point.  But yes,
NumPy should definitely be involved for contour plots, surface plots, 3d
projections onto 2d, line-fitting & interpolation, etc.

On the other point, my forgetting to mention PIL was simply that, a
temporary lapse of comprehensiveness -- PIL is definitely something which
should be taken into consideration, along with PLPlot, Narcisse, Gist,
Gnuplot, OpenGL, Tk, GLUT, etc. etc.  All of these have problems and
features.  How much we can use existing work and how much we have to do
from scratch is up for discussion, and will depend on too many factors to
list. Arrays I'm pretty sure we'll need, for the reasons mentioned by
Andrew among others, and at least we know what code to use for that part.

I've looked a bit at the architecture behind a few packages (gnuplot, gle,
gist, etc.), and some design decisions are clear winners, while others
clearly aren't.  I feel it would behoove us to learn from history, should
a redesign be deemed necessary.  The distinction between graphics engine
and plot driver allows maximum portability with potentially very
distributed workloads (you do Xterms, I do postscript).  That driver
definition can be extremely limiting if it's not done carefully, and with
enough extensability in mind (more a problem in static languages). And so
on (look at how few packages handle interacting with graphics so as to
allow exploratory data analysis (see S+ for an example, DataDesk for
another, maybe StatLisp as well)). But all this should probably wait until
the SIG creation decision has been reached.

can-you-tell-i'm-unhappy-with-every-single-plotting-package-i've-tried-
and-i've-tried-a-lot-ly yours,   --da






_______________
MATRIX-SIG  - SIG on Matrix Math for Python

send messages to: matrix-sig@python.org
administrivia to: matrix-sig-request@python.org
_______________