Topic: Tiny topic

[MP]

Do you ever have need to normalize values that are approaching, but not quite zero, like -1.83691e-016, that is, have it evaluated as zero if it's smaller than an arbitrary threshold?

I had a colleague who needed something like this so I thought I'd share it.

(defun Normalize ( x )
    (cond
        ((zerop x) x)
        ((eq 'int (type x)) x)
        ((equal 0 x 1e-15) 0.0)
        (x)
    )
)   

(Normalize 0)

==> 0

(Normalize 0.0)

==> 0.0

(Normalize 1e-5)

==> 1.0e-005

(Normalize -1.83691e-016)

==> 0.0

(mapcar 'Normalize '(1.0 -1.83691e-016 0.0))

==> (1.0 0.0 0.0)[/color]

The idea is that it won't change ints to reals and vice versa, but it does return 0.0 for minuscule values.

Obviously the 1e-15 fuzz factor is subjective.

What would you use?

Any other thoughts?

[CAB]

Interesting, I have always used just the fuzz.

Would you ever have need of this?

Code: [Select]

((equal 0 (- x (fix x)) 1e-15) (float (fix x)))

[MP]

Interesting, I have always used just the fuzz.

Would you ever have need of this?

Code: [Select]

((equal 0 (- x (fix x)) 1e-15) (float (fix x)))

Hmmm. let me play with that Allan. Thank you.

Tangent: My normalize function was lifted from this little thing I penned long ago:

(defun IsZero ( x )
    (or (zerop x)
        (equal x 0 1e-15)
    )
)

[MP]

Interesting func Allan: Round a number to its integer value if it's fractional portion is below the threshold.