PR#751

Peter Dalgaard BSA <p.dalgaard@biostat.ku.dk> writes:

[1]   0+0i   0+0i -10+0i

which aren't all roots of x^3.

That one is way out, however. Even more "fun":

polyroot(c(-100,0,1))

[1]  10+0i -10+0i

x<- polyroot(c(0,0,0,1))
x

[1]   0+0i   0+0i -10+0i

x<-polyroot(c(2,1,1))
x<- polyroot(c(0,0,0,1))
x

[1]  0.0+0.000000i  0.0+0.000000i -0.5-1.322876i

I.e. the result of polyroot(c(0,0,0,1)) depends on the previous
polyroot calculation! So there's more to the problem than numerical
instability.

Thanks for pointing it out.

Found and fixed. Unfortunately the code had just forgotten to test for
the case where you end with a zero-degree polynomial after zero-root
removal, so the fix has zero effect on the accuracy issue. 

The fix is simply 

--- src/appl/cpoly.c    2000/11/24 21:34:31     1.20
+++ src/appl/cpoly.c    2001/01/17 19:45:24
@@ -148,6 +148,8 @@
     nn++;
     /*-- Now, global var.  nn := #{coefficients} = (relevant degree)+1 */
 
+    if (nn == 1) return;
+
     /* make a copy of the coefficients and shr[] = | p[] | */
     for (i = 0; i < nn; i++) {
        pr[i] = opr[i];

O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard@biostat.ku.dk)             FAX: (+45) 35327907
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._