Re: new graph plotting package

[Ed Frank]

> 
> I've just completed a new Lout package for plotting graphs
> (the statistical kind, not the graph-theoretic kind).  It
> will be distributed as part of the forthcoming release of
> Basser Lout, but meanwhile I would appreciate it if those
> of you who are into graphs have a look at it with a view to
> suggesting improvements before the public release.
> 

Jeff,
   Thank you for Lout and Graph.  Unfortunately, I haven't time right
now to test it extensively, but I have a few comments based on reading
the documentation carefully.  With a few exceptions (comments 1&2),
the documentation was quite clear.  Some of the comments below pertain
to you current work, others are relevent to extensions that you or
others may want to write.  I list them in case their possibility
affects the design.

Comments:

1.  Section 4:  It is not clear if text and symbols (point markers on
    a curve) can be given separate sizes/scaling factors.  
    This is needed.

2.  It was not clear what the arguments to the various captions
    commands were.  The appendix makes it clear that an arbitrary
    Lout object is allowed.  The point here is that it is necessary
    to allow abritrary text in the captions, e.g., formulae via Eq,
    without resorting to the Object command. It looks like you 
    have this covered.

3.  Encapsulation:  I have had my eye on Lout since first hearing
    of it for use in graph generation: most graphing
    programs are very weak in their ability to handle text.  To
    serve this purpose the final output must be usable inside another
    postscript document, e.g., a figure included in TeX. Sorry.  This
    is a statement that my group uses TeX, and I provide the
    results of my work as graphs which go into our TeX documents,
    regardless of what I personally use for typesetting.  A less
    upsetting example would be importing one of your graphs into
    a drawing program like Fig where it could be doctored by hand.

4.  Size of ordinate axis:  You mention the problem of not knowing
    the size of the y-axis text and ticks and provide leftgap to
    help work around the problem.  An alternative is have a "box"
    of some size abutting the axis within which the ticks and text
    are placed.  Material falling outside the box are clipped. If
    the box size is inconsistent with the user's input, it will be
    clear in the output, BUT, you, the programmer, will always know
    the size of this component of the graph.  The point here is
    that I think it is very important that the programmer/user be
    able to know the sizes of each "piece" of the graph.  Unpredictable
    portions could be wrapped in a clipping region to achieve this.
    I feel this will become important as people start extending your work.
    This might also help with my next querry:

5.  Multiple axes:  will the current design allow an extension that
    provides multiple x or y axes?  Example:  a graph may have the
    abscissa labeled both in degrees Kelvin and milli-electron-volts.

6.  Will the current design allow the addition of filled (shaded) 
    regions?  The region would be defined by the user's data, the
    axes, or even intersections, unions, etc. of regions.

7.  Histograms:  Personally, I would need a style that did not
    draw vertical lines for each bin, but just drew the top and
    as much of the vertical lines as needed to connect it to
    the adjoining bins.  Variable bin widths might be needed too.
    Here it would probably be best for me to just modify the 
    Graph code for my tastes rather, but I just wanted you to know
    what others might like.


Thanks again for all your work.  By the way, add me to the list of
people who would prefer not to think about white space: I would like a
mode in which the number of blanks between words in the input did not
affect the output.


        -Ed

-------------------------------------------------------------------------------
Ed Frank                                             Department of Physics     
Office:  (215)898-5979                               University of Pennsylvania
address@hidden                          Philadelphia, PA  USA