ios - CALayer zPosition ordering with extremely large values -

I have attended an issue where the core animation appears to be zPosition for my values Ignoring the values ​​become larger.

example

It can be easily replicated

  CGFloat largeZPosition = 1000000000; CALayer * topLayer = ... CALayer * bottomLayer = ... belowLeather.frame = CGFFetset (toplinerframe, 50, 50); TopLayer.zPosition = Big ZPosition + 1; Downlayer ZPosition = Large ZPosition; // Sort arithmetic array in wrong order RootLayer.sublayers = @ [header, bottom letter];  

With small values ​​for z-position, this function, as expected, even if the sub-column array is not sorted correctly, the bottom side is displayed on the top side Such as:

However, the examples shown above , The result is as follows:

Obviously this is not right.

Question

I'm unable to classify manual inventory for my needs, so I'm relying on zPosition to perform as expected So my question is: How do I work around this problem? Thank you.

That's because Quartzcore has been composed of zPosition to float , therefore it loses the accuracy when CGFloat is double . IEEE single-precision floating-point numbers use 24 bits meaning more accurate, so more than 2 ^ 24 integral numbers can not be properly represented. For example,

  float biggeonation = 1000000000; Float bigZPositionPlus1 = 1000000001; NSLog (@ "% F% F% D", Largegeonation, LargeGroup Plus 1, Biggerency == LargeJustnessPlus1);  

Output

  1000000000.000000 1000000000.000000 1  

The next accurate representational integer is 1000000064.

Note that CALayer.zPosition is defined as CGFloat and it is capable of having an integral value greater than 2 ^ 24 , As long as the underlying CGFloat is able to do this

  layer.zPosition = 1000000001; NSLog (@ "% F", Layer .ZPosition); 64-bit OS X  

Where CGFloat is double , output

  1000000001.000000  

Therefore, on CALayer public purpose - at C class level, it has not been lost. Wherever there is a conversion for single-precision float anywhere in Quartzcor.

I suggest using integral numbers (including the result of extra) which is less than 2 ^ 24. Given that there is a floating-point type in zPosition , you can use small partial values ​​instead of large integral values.