[JASS] Which Blizz functions leak?

[Darkholme]

I just found out that GetUnitsOfPlayerAndTypeId and CountUnitsInGroup leak every time they're called. Can someone give a list of all the blizzard functions that leak? is there already one?

 

[Hatred]

Every function that returns a handle is a possible source for a leak.

 

[PurplePoot]

No BJs leak any more than nulls to my knowledge.

(CountUnitsInGroup does not leak at all, GetUnitsOfPlayerAndTypeId (and all other Get...) leak one null).

They usually return stuff you have to clean up to avoid leaks, though.

Solution: Avoid using BJs.

 

[Darkholme]

No BJs leak any more than nulls to my knowledge.

(CountUnitsInGroup does not leak at all, GetUnitsOfPlayerAndTypeId (and all other Get...) leak one null).

They usually return stuff you have to clean up to avoid leaks, though.

Solution: Avoid using BJs.

do you just mean the functions preceded with bj?

 

[PurplePoot]

I mean functions in the Blizzard.j file, pretty much non-natives.

 

[Darkholme]

I mean functions in the Blizzard.j file, pretty much non-natives.

so all of the blizzard non-natives suck then? Thats alot of stuff.

 

[PurplePoot]

so all of the blizzard non-natives suck then? Thats alot of stuff.

The majority of them...

 

[Drone]

TriggerRegisterAnyUnitEventBJ can be use full since it doesn't do useless stuff like lots of BJ's do. In short they just blow and its better to avoid them totally.

 

[Darkholme]

So what I need to learn to do is directly access the native functions all the time?

 

[Justify]

As much as possible.
Some functions like TriggerRegisterAnyUnitEventBJ can be rewritten (I think ReplaceUnitBJ is also a small codeblock), but imo THESE BJs are okay.
Stuff like GetUnitAbilityLevelSwapped (ALL swapped functions just change the order) 'has' to be replaced by natives

 

[Darkholme]

Damned.

More work for me. Alright. Premade functions almost universally=suck.

Anyone care to do the opposite? tell me a comprehensive list of the functions that Don't suck? I imagine since theyre as bad as you guys say, there are less than 30 of them?

 

[MindWorX]

One set of blizzard.j functions you don't have to avoid, since they don't have a native equivelant, unless you want to redo their system, is the TriggerQueue natives. Also, best way to find the native version of bj's, is to use TESH that comes with NewGen, since it highlights BJ's, and if you ctrl+left click them, it'll show you the full BJ function, and then you can usually copy out the parts you need.

 

[Silvenon]

Damned.

More work for me.

Not necessarily, some natives are easier and faster to use than BJs. You'll see once you get used to using natives that it's the right way to go.

Also, cinematic BJs are useful, moreover, they have to many function calls inside them so it isn't worth replacing them (a worse thing than an ordinary BJ is a BJ calling another BJ).

 

[HINDYhat]

You guys seems to have forgotten about the PolledWait function, which happens to be a BJ. Pretty useful function there.

 

[MindWorX]

Not too useful tho, since it leaks a handle every time. Vexorian made a PolledWait2 function, i think it's at wc3jass.com, which fixes this.

 

[PurplePoot]

And then there are timers <3

Also, cinematic BJs are useful, moreover, they have to many function calls inside them so it isn't worth replacing them (a worse thing than an ordinary BJ is a BJ calling another BJ).

However, they are horribly un-customizable (for example, you have to recode them to get rid of pressing Esc to skip a cinematic sequence)...

The only BJs I tend to use are:

TriggerRegisterAnyUnitEventBJ: Shortens code by a good deal
RAbsBJ/RMinBJ/RMaxBJ: There is no better way to do this without variables and an if-statement, which are sometimes undesirable.

That's about all I can think of.

 

[HINDYhat]

And BJDebugMsg, of course

 

[MindWorX]

If you're using NewGen i'd suggest something like this instead of BJDebugMsg

function Echo takes string s returns nothing
    call Cheat("DebugMsg: " + s)
    call DisplayTimedTextToPlayer(GetLocalPlayer(), 0, 0, 60, s)
endfunction

It's 10 function calls faster, and it writes the debug message to "logs//war3erruser.txt". Very useful.

 

[PurplePoot]

And BJDebugMsg, of course

Only for debugging, never in an actual game (And who cares about efficiency when you're debugging)

Mindy, who cares about the efficiency of your debug messages -.-

 

[Drone]

Only for debugging, never in an actual game (And who cares about efficiency when you're debugging)

Mindy, who cares about the efficiency of your debug messages -.-

Well maybe you don't, but I do. He's simply right about efficiency...

 

[Preskooldude]

But you don't need to be efficient in a non-released map. BJDebugMsg can primarily be used to make sure that a piece of code is actually running, or a number is what it's supposed to be. Since I'm pretty sure you want neither in your final release, efficiency doesn't matter.

 

[MindWorX]

You miss the the point, I know optimization isn't exactly the biggest issue, but the really nice part is the logging function that works with NewGen.

 

[PurplePoot]

The logging part is nice, yeah, but the rest of the complaint is just silly >_>

 

[Dr Super Good]

What do you mean by a null leaking. I heard they do that some times but in what situations and why.

 

[PurplePoot]

What do you mean by a null leaking. I heard they do that some times but in what situations and why.

Apparently if a handle's reference counter is >0 its index can't be recycled.

 

[~Void~]

I'd like to say the solution is to use vJass.