[Plplot-devel] Coordinate wrapping problems for the python, tcl, and java front ends

[Alan W. I. <ir...@be...>]

This is mostly addressed to Geoffrey since the coordinate wrapping code is
due to his efforts I believe. I would appreciate his comments in case I have
misinterpreted something.

I believe the code that wraps angular (or other coordinate) values for
plcont, plshade, etc. in plmodules.c, tclAPI.c, and javabind.c has a nasty
bug that I am currently trying to fix.

Let's take the python definition of plshade in plmodules.c as an example.
The problem is the wrap code extends xg2 and yg2 appropriately and makes the
corresponding changes to cgrid2, but no corresponding change is made to grid
(which carries the z information).  Thus, the current implementation of the
python wrap code has inconsistent z, xg2, and yg2 values, and the same is
true of the plcont code. From my quick check I believe exactly the same
problems occur for tcl as implemented in tclAPI.c.

I believe the reason why this problem has not been seen before is the
current x09.tcl, x16.tcl, and xw09.py, xw16.py, and x09.java (x16.java is
not written yet) don't actually test coordinate wrapping that well. The
problem is that the example theta values range from 0 to 2 pi, the last
point is already essentially equal to the first point, and the wrapping has
already occurred before you ask for additional wrapping that is implemented
(incorrectly I believe) in the plmodules.c, tclAPI.c, and javabind.c front
ends.

For my local copy of xw16.py I reduced the maximum index by one so that the
last t (the theta coordinate) value was 2 pi - delta t.  Setting wrap = 2
should fill in the equivalent of t=0 (or 2 pi) for the last value.  However,
the result did not look good, and certainly was not equivalent to the C
front end result which uses the full t range from 0 to 2 pi without any
additional wrapping.  I also tried wrap = 1 (just in case I misinterpreted
the order of the x and y coordinates) and that produced a segfault. Further
investigation showed the problem in plmodules.c, tclAPI.c, and javabind.c
that I have discussed.

Alan

email: ir...@be...
phone: 250-727-2902	FAX: 250-721-7715
snail-mail:
Dr. Alan W. Irwin
Department of Physics and Astronomy,
University of Victoria, P.O. Box 3055,
Victoria, British Columbia, Canada, V8W 3P6
__________________________

Linux-powered astrophysics
__________________________