[R-sig-Geo] DELDIR

[Rolf Turner]

This question should have been, in first instance, directed to me (the 
maintainer of the deldir package), rather than to the r-sig-geo list.

On 24/07/19 1:47 AM, PONS Frederic - CEREMA/DTerMed/DREC/SRILH wrote:

> Dear all
> 
> Sometines we are using the deldir function for squeletisation.

Had to Google this word. According to what I see it should be spelled
"squelettisation" (with a double "t").  The English word is 
"skeletonisation", and this list is (for better or for worse) an English 
language list.

> Since the last review, deldir stops in some shapefile because polygons
> are too narrow. In the version berfore, there was no problems.
> 
> We have readed the " Notes on error messages" and the problem of
> anticlockwise order of triangle is listed.
> 
> In the trifnd R function , the code is
> # Check that the vertices of the triangle listed in tau are
> # in anticlockwise order.  (If they aren't then alles upgefucken
> # ist; throw an error.)
> call acchk(tau(1),tau(2),tau(3),anticl,x,y,ntot,eps)
> if(!anticl) {
>       call acchk(tau(3),tau(2),tau(1),anticl,x,y,ntot,eps)
>       if(!anticl) {
>           call fexit("Both vertex orderings are clockwise. See help for
> deldir.")
>       } else {
>           ivtmp  = tau(3)
>           tau(3) = tau(1)
>           tau(1) = ivtmp
>       }
> }
> 
> We don't understand why do not order the bad triangles into the good
> order. Perhaps, if this problem appears, the beginning of the deldir
> function is not good.
> 
> If someone can explain us.

The error message you quote indicates that *both* orderings are bad so 
something is toadally out of whack.  It would be rash (and arrogant) for 
the software to fool around with the structure.  In such situations the 
user should diagnose what is going wrong.

If you cannot diagnose the problem, please send me a reprex and I will 
look into it.

cheers,

Rolf Turner

-- 
Honorary Research Fellow
Department of Statistics
University of Auckland
Phone: +64-9-373-7599 ext. 88276