delayTime rounding about flanimatedimage

[raphaelschaad]

Great catch. Curious that NSNumber doesn't do "the right thing" here.

We could use a small enough epsilon to let the minimum slide (but not much beyond that) e.g. by defining const CGFloat kEpsilon = 0.00001; or using the constant FLT_EPSILON. But I never liked mocking "manually" with float comparisons.

Any suggestions?

[pronebird]

What if we clamp kDelayTimeIntervalMinimum to float so both types match when comparing? I just tested this and it works fine.

if ([delayTime compare:[NSNumber numberWithFloat:kDelayTimeIntervalMinimum]] == NSOrderedAscending) {
    // ...
}

Fun stuff to read: https://github.com/opensource-apple/CF/blob/master/CFNumber.c#L1347

[raphaelschaad]

Interesting, so if I interpret the Core Foundation code correctly, all NSNumber does is upcasting floats to doubles when comparing.

If we go the route of "normalizing" by down casting to float -- which is what I tried to avoid by just letting NSNumber do "the right thing" --, it seems more straight-forward to just do it by (float) kDelayTimeIntervalMinimum and then comparing with less than.

Does that capture the case with your test GIF?

[pronebird]

That should work.

[raphaelschaad]

Could you verify it addresses the case with your specific GIF? If yes, would you like to open a pull request with the change? I'm happy to merge it in.

Let me know if you'd prefer me doing it.

[pronebird]

Sure, let me check. I can prepare PR, it's not a problem.

On Friday, July 18, 2014, Raphael Schaad [email protected] wrote:

Could you verify it addresses the case with your specific GIF? If yes,
would you like to open a pull request with the change? I'm happy to merge
it in.

Let me know if you'd prefer me doing it.

—
Reply to this email directly or view it on GitHub
#20 (comment)
.