Calling cancel on drag gets called twice (on strong drags) about use-gesture

[morgs32]

In this line here, why do we have to use a requestAnimationFrame. Some other race condition?

[morgs32]

What the... I just "updated" codesandbox and this ceases to happen. Sorry for the noise.

[morgs32]

Ok... it only happens when dev tools is open?

[dbismut]

Hey @morgs32, thanks for reporting the issue, nice catch :) Yes, requestAnimationFrame is used to let the gesture logic finish before terminating it with cancel.

onDrag = (canceled, cancel) => {
  if(!canceled) { // <-- logic that depends on calling cancel()
    console.log('calling cancel() with canceled equal to', canceled)
    cancel()
  }
  console.log('canceled is equal to', canceled)
}

/* output using RAF */
// calling cancel() with canceled equal to false
// canceled is equal to false
// canceled is equal to true

/* output without using RAF */
// calling cancel() with canceled equal to false
// canceled is equal to true
// canceled is equal to false <-- not exactly what we want

What seems to happen with the devtools opened in the example is that RAF debounces the call to terminate the gesture and the browser sometimes has time to execute the gesture twice, hence the logic (that depends on cancel being called) to increase the index occurring twice. I'll investigate further why this might happen, but one solution would be to remove RAF and it would be up to the user to use cancel as the final statement to its gesture logic:

onDrag = (canceled, cancel) => {
 if(!canceled) {
   console.log('calling cancel() with canceled equal to', canceled)
   cancel()
 }
 else // <-- added else
   console.log('canceled is equal to', canceled)
}

/* output without using RAF */
// calling cancel() with canceled equal to false
// canceled is equal to true

[morgs32]

I figured as much! Here's what resolved it for me:
#53

Of course, it was only really an issue because in development. Interesting!

[dbismut]

I think it's actually an issue every time the browser would drop frames so I believe this needs a fix! I'll review your PR now :)

[dbismut]

Should be fixed in 5.0.1!