Floating point increment bugs out. Incr should take float

[TERRENCE MONROE BRANNON]

[1] incr should take a float argument -- i dont see why not

[2] the following bugs out:

for {set i 0.0} {$i != 1.0} {set i [expr $i+0.1]} {
echo $i
}

resulting in:

0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.7999999999999999
0.8999999999999999
0.9999999999999999
1.1
1.2

.. and so forth until infinity
--
Terry Brannon tbra...@lion.eecs.lehigh.edu
medical biology via acupunctural bioelectrics
primitive reigns supreme over nearly everyone

[Keith Steiger]

In article <TB06.93Ap...@CS2.CC.Lehigh.EDU> tb...@CS2.CC.Lehigh.EDU (TERRENCE MONROE BRANNON) writes:
>[1] incr should take a float argument -- i dont see why not

Because then the result would have to be a float, and some people don't wamt
their numbers gratuitously changed to floats when they're not looking. Or use

proc incrf {v {i 1.0}} {upvar 1 $v x; set x [expr $x+$i]}

>[2] the following bugs out:
> for {set i 0.0} {$i != 1.0} {set i [expr $i+0.1]} {
> echo $i
> }
>resulting in:
>0.0
>0.1
>0.2
>0.3
>0.4
>0.5
>0.6
>0.7
>0.7999999999999999
>0.8999999999999999
>0.9999999999999999
>1.1
>1.2
>.. and so forth until infinity

Two options:
for {set i 0.0} {$i <= 0.99} {incrf i 0.1} {echo $i}
for {set i 0} {$i != 10} {incr i} {echo [expr $i/10.0]}

The second of these gets much better accuracy on its results by holding off
on the switch to floating point as long as possible, and I recommend it.

Keith Steiger
(who is going through a crash course on Tcl and loving it)
"Integers are GOOD! Floating-point numbers are BAD!"