Sorry I should have said that.
(Mythic might remember those stress tests.)
I have ran some stress tests with a complete empty Damage Detection System where there are 2 units that attack each other and a third that gains a number of "this unit takes damage" events every second.
So after a while it has 70k events. Then it runs slow.
Of course that is only the events.
Then you also have to think about the other stuff that is in your map.
So with numbers and stress tests like these you can find out how long the re-initialize interval may be so it will never run slow.
There is a second performance thing that is when a unit is actualy attacked.
In the map I posted I have 50 - 250 units attacking each other every second.
It also displays how many (maybe that trigger is disabled and you have to enable it)
But I will run stress test on it as soon as it is finished...
I don't assume that my system will have the best performance but I at least want 2k events and 800 unit attacks per second.
If that is met then I can say it is not too cpu consuming to not use it.
not really good stress test, the lag you notice on all systems is because the numbers that you print want to be drawn even tho they are off screen, and so if you want to print quite a few K lines of text, it will start to lag immensly.
Return type declarations are broken anyways in JNGP. Its perfectly valid to write something like:
JASS:function foo takes nothing returns nothing array endfunction // or function foo takes nothing returns real integer damagetype return 1.0 endfunction
So the parser is quite broken here^^
Lol that is silly, I didnt know that is possible. But it is runnable, because JassHelper only keeps the first one