CGAL users discussion list

["Fernando Cacciola" <>]

Hi Alejandro,

> the point is supposed to lie over the line within let's
> say tol = 10e-14.
This is an arbitrary interpretation of what lying over a line means.
If this interpretation works in you particular case, you can implement it easily 
via the squared_distance() function, but have in mind that your sample point 
miught not actually lie on the line in spite of the number type used.
That is, you will get the wrong answer for using plain double, but you could 
also be just expecting the opposite answer because you asummed closed enough is 
the same as exactly on the line.  CGAL predicates are strictly exact: they 
return the answer they would if you used an exact number type.

> Why would I use an exact number type?

You wouldn't.
Please read the FAQ much more carefully, in particular de paragraph starting 
with:

"If your program does not involve the construction of new objects from the 
input data..."

CGAL uses a technique knows as "Floating-point filtering" which exploits the 
precision of 'double' as much as possible yet it guarrantees correct results 
even when that precision is detected as not being enough for a particular 
operation.

HTH

Fernando Cacciola
www.geometryfactory.com