[reportlab-users] Widgets as Flowables...

Andy Robinson reportlab-users@reportlab.com
Sun, 29 Sep 2002 22:46:22 +0100

> I forgot to say that preliminary research indicates that people are
> using linear inequality solvers for the sizing problem. The approach is
> indicated here
> http://www.cs.washington.edu/research/constraints/cassowary/

I was going to suggest that the drawing problem was simple -
we just needed the drawing to be big enough or the contents
but then I remembered lots of other knotty areas we have
not really tackled, or have hard coded limited rules for:
 - optimal selection of tick intervals on chart axes
 - text labels that don't collide around a pie
 - choosing sensible colors when they have not been set

And then there's the long-document-formatting problem,
which we've now concluded needs measures of goodness
and badness where objects fit together, and multiple
passes to resolve indices and tables and xrefs.

It would be good at least to start enumerating constraints,
leaving open the question of whether they are best solved
by linear programming, genetic algorithms or (as the drummer
said in Spinal Tap) "best left unsolved"....

- Andy