I got smacked by this morning…..

I have a Sequence Blend blending between two SOPs.

If have an expression

fit(frac($FF),0,.4,0,1) set for the blend parameter. The neat part is, the Blend SOP fails when $FF = 1.4. (If $FF = 2.4 or 3.4 etc it works fine.)

So upon further testing..

fcur 1.4
echo `frac($FF)==0.4`
Returns 0.
Which is fair enough considering your never suppose to do direct comparisions, ==, between floats.

Even further testing….

Changing
fit(frac($FF),0,.4,0,1)
to
fit(frac($FF),0,.4,0,.999)

Fixes the problem.

So, back to the question… If the Sequence Blend works with any number why is the expression failing? NaN?

frac(x) is implemented as x - floor(x).

Due to floating point precision, frac(1.4) does not return 0.4 but a value slightly less than 0.4. Try
echo `1.4-floor(1.4)-0.4`
or better yet
echo `1.4-1.0-0.4`
now try
echo `2.4-2.0-0.4`
and
echo `3.4-3.0-0.4`
Notice that for 2.4 and 3.4, you get a value slightly positive while for 1.4 you get a value slightly negative. When you feed these frame values into your expression, you get values equal to 1 and slightly less than one respectively for the blend factor. The Sequence Blend SOP did not handle the slightly less than 1 case. This is now fixed in 8.0.452 (Bug ID=20924).

Cool, thanks Glenn.

I figured it precision related, but couldn't make heads or tails of why the Sequence Blend SOP was struggling.

Thanks again.