[Proj] cs2cs the fast way

[Gerald I. Evenden]

On Saturday 30 May 2009 5:25:52 pm support.mn at elisanet.fi wrote:
> "Gerald I. Evenden" [geraldi.evenden at gmail.com] kirjoitti:
> > The datum shift may add a few coefficients but we do not have to
> > evaluated a possibly complex datum routine.
> >
> > Obviously, this is not yet ready for prime time.
>
> Ok. If the results are exactly (to adjustable degree) the same and you can
> do it faster, then you have really improved something.
>
> BTW: I am sure that there must be several ways to make routines faster for
> example by moving calculations to the initialization phase. Since the
> current computers have huge fast memories that might be a nice resource to
> be used to gain speed (using tables to speed up for example).
>
> Yes, there is always need for speed. That makes everything run smoother.
>
> Janne. / MNS Support

There is nothing new with this Chebychev stuff.  Back in the '80s proj 
generated these Chebychev files for use in the program system called MAPGEN.  
Because MAPGEN only made large scale maps (down to 1:250,000) in the mid 
latitudes this system worked fine.  Also, because we were working on mini 
computers where 1Mb was a big thing, the graphic programs did not have to 
have all the projection software aboard---only the routine to evaluate the 
the two polynomials.  Speed and size of the software were both 
considerations.

As for more ancient history: circa the mid 60's there was a popular routine 
for doing Transverse Mercator.  The interesting bit was the use of bivariate 
approximating polynomials.  The region was divided into about several zones 
in the N-S direction with a set of polynomials for each segment.  I do recall 
if I ever knew where this version came from but I never saw it again.  Of 
course this routine was hard-wired for Clarke '66.

-- 
The whole religious complexion of the modern world is due
to the absence from Jerusalem of a lunatic asylum.
-- Havelock Ellis (1859-1939) British psychologist