On Fri, Nov 2, 2012 at 12:25 PM, Alan W. Irwin <irwin@beluga.phys.uvic.ca> wrote:
On 2012-11-02 07:45-0000 Andrew Ross wrote:

On Wed, Oct 31, 2012 at 02:50:53PM -0400, Hezekiah M. Carty wrote:
On Wed, Oct 31, 2012 at 11:57 AM, Alan W. Irwin
<irwin@beluga.phys.uvic.ca> wrote:
Hi Andrew:

On 2012-10-31 09:55-0000 Andrew Ross wrote:


You are right that setting the map transform for plmap should be
identical to setting the global transform here.

One obvious question - have you tried with a version before the recent
plmap changes to make sure it isn't a new bug we've introduced?

I haven't tried such a test with the old code, but such a test would
probably be worthwhile if someone decides to address this issue.


Alan and Andrew,

The difference in the latitude lines in example 19 has been there as
long as pltransform and the Baltimore example 19 page have been
included.  The difference in dash positions occurs because plmeridians
uses different rendering paths depending on whether it is given a
coordinate transform function or not.

If plmeridians receives a coordinate transformation function as it's
first argument (C API) they it manually segments the latitude and
longitude lines by repeatedly calling plline.  If no coordinate
transformation function is passed to plmeridians then it calls plpath
to draw the latitude and longitude lines.  plpath only calls plline
once.  I expect that this is where the difference in dash positions
comes from.


Thanks for the explanation. Now you mention it I seem to remember us
having the same conversation once before. Guess I should check the
archives first.

Hez and Andrew:

I still feel it is a good idea to obtain the same result in the two
cases if at all possible.  I assume manual segmentation of the latitude and longitude lines is necessary in the first case so
why not use that exact same method in the second case?


I think it would make more sense to do the reverse - replace the mapform case with a single call to plline.  That should give the best rendering results across devices since each device can do its own optimized rendering/handling of the single multi-segment line.

I'm not sure when I'll get a chance to make the code changes but I have added this to my PLplot TODO list.