Transverse cylidrical map projection.

[James Kuyper <kuyper(at)wizard.net>]

I want to plot data using a transverse cylindrical map projection. An
equal-area one would be best, but equidistant or mercator would be
almost as good, just so as long as it's transverse version of one of the
cylindrical projections. MAP_SET accepts a tilt angle, which doesn't do
what I want for most of the cylindrical projections. The user's guide
contains an example command:

        map_set,0,0,45,londel=20,latdel=20,/grid,$
        /continent,/cyl,title='Oblique Cylindrical Equidistant'

which is shown in the book as producing a map with the projection axis
tilted by 45 degrees: the lines of constant latitude and longitude are
curved. When I try it, I get a map tilted by 45 degrees, which is a very
different thing: The lines of constant latitude and longitude are
strait, tilted by 45 degrees. This suggests that the book was printed
using a different (hopefully later) version of IDL than I'm using. I saw
the problem first in version 5.0.3, but I've recently discovered where
they've hidden version 5.2 on our machine, and I still see the same
results using it.

Luckily, I've found that the transverse mercator projection does
implement the tilt properly. However, in large maps it often considers
one or more of my limit points unmappable, for reasons that escape me.
For example,

map_set,-15.7970,-90.4190,260.1820, limit=[78.548,-31.494, $
-27.66,-64.441, -64.066,103.55, -0.792,-114.296],$
/continents,/grid,/label,/isotropic,/transverse_mercator

Produces the complaints:
% MAP_SET_LIMITS: Unmappable limit point:      -31.4940      78.5480
% MAP_SET_LIMITS: Unmappable limit point:       103.550     -64.0660

The only points that should be unmappable in a mercator projection are
points near the projection axis, and of the four points I gave, those
two are the ones farthest from the axis.

If you use /lambert instead of /transverse_mercator, you'll get a feel
for the kind of plot I want. /stereo and /azimuthal also work, each with
their own distortion properties. However, I want a cylindrical
projection, not an azimuthal one - the projection properties aren't
quite what I want.

Can anyone suggest a solution?