CGAL users discussion list

["Fernando Cacciola" <>]

Alejandro Aragon wrote:
> Oh, and for your information, the precision(20) that you suggested is
> stupid because you're way beyond the precision for that type

numeric_limits<double>::digits10 is 15 in a typical implementation, so I 
wouldn't say 20 is way beyond that.

Also, according to this paper:

http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2006/n2005.pdf

17 would be a reasonable precision for outputing double values in decimal 
format

> , so you're basically outputting garbage.
What makes you think that using more decimal digits than minimally required 
results in garbage??

Try this for yourself:

double d = ( 1.0 / 3.0 ) + numeric_limits<double>::epsilon() ;

cout.precision(numeric_limits<double>::digits10); cout << d << endl ;
cout.precision(20); cout << d << endl ;
cout.precision(40); cout << d << endl ;
cout.precision(200); cout << d << endl ;

Finally: did you understand that your point IS NOT really on the line because 
the actual coordinates are not exactly the same as the decimal literals you used 
to initialized them?

Best

Fernando Cacciola
www.geometryfactory.com