Next: Formulas as Closures
Up: New Directions
Previous: Multi-threading
A final area of major change will be in the graphics system. Some of
the changes will be simple in concept. These include adding
hierarchical menus, hierarchical and nested window structures,
geometry management, and support for using native widgets.
At a higher level, there is a need to redesign the basic model for
statistical graphs. The current model represented a compromise of
expressiveness and performance. Performance is still an issue but for
different reasons than five years ago. Design changes need to take
account of the fact that techniques may now be applied to data sets
ranging in size up to hundreds of thousands of observations.
The primary reason for a need to redesign the graphical model is best
illustrated by an example. The plot in Figure 1 can be
enhanced by adding a smooth through the highlighted points in the
scatter plot as shown in Figure 5.
Figure 5:
Plot linking enhanced by a smoother.
|
This is a simple and useful idea that could in principle be made
available as an option for any plot. But there are many similar
enhancements that could be made available, and doing so would create
an unmanageable collection of options. A better approach is to provide
a flexible set of primitive operations that allows an analyst to
assemble this form of enhancement with a few operations taking a few
seconds. Currently it is possible to add this enhancement with about
a page of code taking an experienced user about an hour to write.
This is adequate for a researcher in dynamic graphical methods to try
out a new idea, but it is prohibitive for an analyst exploring a data
set.
By collecting and analyzing a number of similar examples where
enhancements should be possible with a few simple operations it is
hoped that it will be possible to identify an effective set of
primitives and a graphical model to support them.
Next: Formulas as Closures
Up: New Directions
Previous: Multi-threading
Luke Tierney
5/27/1998