Justin,
I did it in a debug player - oops. I have just redone the tests in a
release player, and with objects in addition to integers. I am using
flashplayer 11 on Linux because that is the latest version available.
Timings for incrementing each integer in a million item array:
Array: 0.1 sec
old ArrayList (get/set Item): 0.4 sec
new ArrayList (get/set Item): 0.4 sec
new ArrayList ([]): 3.7 sec
Timings for initialising an array with either setItemAt({x: 0}, i) or
array[i] = {x: 0} and then getting each item and incrementing x:
Array: 0.5 sec
old ArrayList (get/set Item): 0.9 sec
new ArrayList (get/set Item): 1.0 sec
new ArrayList ([]): 12.7 sec
So the results are similar to before: actually using the sugar is much
slower than not doing so, but just extending Proxy makes little difference.
Gareth
On 26/05/15 05:47, Justin Mclean wrote:
Hi,
I have done some very unscientific tests: for each kind of list (Array, old
ArrayList, new ArrayList, new ArrayList - using sugar) the test code creates a
million-item array and then increments each value that it contains.
Was this done in a release player? And did you try testing it on array of
objects rather than integers?
Thanks,
Justin
