$Id: TODO,v 2.3 1999/10/14 03:16:14 mhagger Exp $

Some ideas for future work.  I don't know whether I'll get around to
any of these items; some of them are probably better left undone
anyway.  Of course if anybody else wants to take one up...

* Implement missing values.  Gnuplot allows you to specify that values
  are missing from data files via the `set missing' command.  I'm
  thinking about implementing this as an optional 'mask' argument to
  the Data and GridData PlotItems.  (Comments?)  The real benefit of
  missing values is that (I think) gnuplot can still put a mesh on a
  surface even if the surface has holes.

* Add a more systematic way of setting arbitrary gnuplot options.  One
  cute possibility that occurred to me (maybe too cute?) would be to
  implement a magic Gnuplot.set object that could be used like:

      g = Gnuplot.Gnuplot()
      g.set.data.style('linespoints', 1, 2)
      g.set.pointsize(2.0)

  which I think could be accomplished with __setattr__ and
  __getattr__.  But this is not a high priority; it's easy enough to
  use the __call__ method to set arbitrary options.

* Figure out how to suck gnuplot error messages back into Python and
  turn them into exceptions.  This would be tricky: for one thing, not
  all gnuplot output is indicative of an error.  Moreover,
  implementation on Windows and Mac would require additional effort to
  change their GnuplotProcess classes.  The output from the 'show'
  command could be processed in a similar manner.

* Implement multiple graphs on a single page.  I haven't used this
  feature of gnuplot much myself, so I'm not sure how difficult it
  would be to build it into the Gnuplot.py framework.  Ideally one
  would like to be able to spawn multiple GnuplotWindow's from a
  single Gnuplot instance, each one behaving much like a Gnuplot
  object.  Or maybe treat sub-windows as a kind of meta-PlotItem that
  can itself contain PlotItems.

