JassHelper 0.A.2.B - A vJass and Zinc 2 Jass compiler

[Zoxc]

I was hoping for a second I would be able to do stuff like this:

Hidden information: (Example of a wierd language with Ruby like syntax) Code: class Handle public static constant MinId: Integer = 0x100000 method id: Integer GetHandleId(self) end end class Group < Agent private static flag: Boolean = false static refr: Group = nil # Assorted constants static constant MaxHandleCount: Integer = 408000 # Arrays and counter for the group stack static groups: Group[] static status: Integer[MaxHandleCount] static count: Integer # ForGroup wrapper static each_callback: ->(Unit) function each_forgroup each_callback(GetEnumUnit()) end function add_ex if flag GroupClear(refr) set flag = false end GroupAddUnit(refr, GetEnumUnit()) end public function allocate: Group if count == 0 then groups[0] = CreateGroup() else count =- 1 end status[groups[count].id - Handle::MinId] = 1 return groups[count] end method destroy stat: Integer = status[self.id - Handle::MinId] case when self == nil debug("Error: Null groups cannot be released") when stat == 0 debug("Error: Group not part of stack") when stat == 2 debug("Error: Groups cannot be multiply released") when count == 8191 debug("Error: Max groups achieved, destroying group") DestroyGroup(self) end GroupClear(self) groups[count] = self count += 1 status[self.id - Handle::MinId] = 2 end method refresh flag = true refr = self ForGroup(refr, function add_ex) if flag GroupClear(self) end endfunction method each(callback: ->(Unit)) each_callback = callback ForGroup(self, function each_forgroup) end end function wierd_stuff evil_units: Group = Group.new evil_units.each ->(evil_unit: Unit) # Get rid of 'em evil units KillUnit(evil_unit) end evil_units.free end class Effect function allocate(path: String, x: Real, y: Real): Effect return AddSpecialEffect(path, x, y) end end class Timer private static callbacks: ->[] function one_shot_wrapper timer: Timer = GetExpiredTimer() id: Integer = timer.id - Handle::MinId callbacks[id]() callbacks[id] = nil PauseTimer(timer) DestroyTimer(timer) end public function one_shot(time: Real, callback: ->) timer: Timer = CreateTimer callbacks[timer.id - Handle::MinId] = callback TimerStart(Timer, time, false, function one_shot_wrapper) end end function timed_sfx(path: String, time: Real, x: Real, y: Real) sfx: Effect = Effect.new(path, x, y) Timer.one_shot(time) -> sfx.free end end

But I guess the overhead is too much to get people to use it. It could allow some nice generators if it was inlined however.

[Anachron]

Increasing numeric variables with $var++; syntax?
I guess you won't at ZINC, because it is, as the name suggests, not c.

[Dark_Dragon]

Originally Posted by Vexorian

Problem is that this solution only works with ForGroup, what if you use the anonymous function for a trigger or a timer?

What if you use it to just assign a code variable?

Zinc:

type meh extends function(real) -> real;
...
      meh fn;
      code c;
      ...
      c = function(){  BJDebugMsg(I2S(x) ); };
      TimerStart(t, 5.0, false, c);
      TriggerAddAction(tr , c);

      TimerStart(t, 65.0, false,  function(){
             BJDebugMsg("65 seconds later");
      });
      TriggerAddCondition(tr ,   Condition(   function()->boolean  { return GetSpellId()==SPELL_ID; }    )  );


      fn = function(real x) -> real { 
            return Sin(x);
      };

i see! i never thought that they will be so powerfull :)

then implement it as u think is the best and ill be happy!

Greets!
~DD

[weaaddar]

Actually Vex,
I think in your example that you can do, you should do it in C# fashion and create basically a static variable x and assign it before usage.
dream code below::

Zinc:

int x = 0x10034;
ForGroup(g, function() { if x < GetHandleId(GetEnumUnit()) BJDebugMsg("Less than x"))};

could translate to...

JASS:

globals
integer AnonyFunc_0001_x
endglobals
function AnonyFunc_0001 takes nothing returns nothing
if(AnonyFunc_0001_x < GetHandleId(GetEnumUnit()) then
 call BJDebugMsg("Less than x"));
endif
enfunction

// in that caller
local integer x = 5;
set AnonyFunc_0001_x = x;
call ForGroup(g, function AnonyFunc_0001)

Or whatever as I haven't written pure Jass in a long time. I actually don't think its that hard and anonymous functions can then become passable entities that encapsulate their definitions.

[Vexorian]

that only happens when the function is going to be called right away, that's only true for ForGroup mostly. For timers and triggers, it would involve attaching and a lot of incredibly annoying things. Would have to decide between allowing it for ForGroup and dissallowing it for the rest, or doing it for the whole deal and then turn the task incredibly complicated (yet still impossible when you want it for function interfaces or assigning code variables (or using local arrays)

So, the question is, I either implement anonymous funcs as a way to implement functions in line without further usage of locals or things like that (I guess you would still be able to use private globals and static members) Or I just don't add anonymous functions.

ForGroup with static locals could turn into another construct , reminds me of this: http://www.wc3c.net/showthread.php?t=100417&page=3

[Vexorian]

The closest thing to a correctly working multi-purpose anonymous function would be... an anonymous trigger.

Zinc:

        integer x= 3,y=5,z=8;
        trigger trig = trigger {
                   BJDebugMsg(I2S(x)+I2S(y)+I2S(z));
        };
        TriggerExecute(trig);

JASS:

function trig000 takes nothing returns nothing
 local integer t__  = GetHandleId( GetTriggeringTrigger() )
        call BJDebugMsg(   I2S(LoadHashTableInt(ht, t__, 0 ) )  +  I2S(LoadHashTableInt(ht, t__, 1 ))+I2S( LoadHashTableInt(ht, t__, 2 ) ));
endfunction

...
local integer x=3
local integer y=5
local integer z=8
local trigger trig = CreateTrigger()
    call TriggerAddAction(trig, function trig000)
    call SaveHashTableInt(ht, GetHandleId(trig), 0, x)
    call SaveHashTableInt(ht, GetHandleId(trig), 0, y)
    call SaveHashTableInt(ht, GetHandleId(trig), 0, z)

... But, DestroyTrigger is not the most reliable thing ever...

[weaaddar]

Which is fine really. Anyway there is no guarantee that the value set by the anony func should be ported forward after all, you've just now compiled a call-by-value entity. I'll verify the c# behavior later. And please the calling method for integer is like SaveInteger...


DestroyTrigger is fine, provided you destroy the triggeraction as well.

edit c# does forward port the value.
[zinc]
int x = 5;
Func<int> f =(() => x = x - 1);
Console.WriteLine(f());
Console.WriteLine(x);
Console.Read();
[/jass]

prints 4 newline 4.
I suppose the Jass could be handled as such...

Zinc:

global
integer Anony_Func_x;
integer Anony_Func_RetVal;
trigger Anony_Func_Trigger;
triggeraction Anony_Func_TriggerAction;
endglobals
function Anony_Func_X takes nothing returns nothing
      set Anony_Func_x = Anony_Func_x-1
      set Anony_Func_RetVal = Anony_Func_x
endfunction

//in the caller...
     local integer x = 5
     set Anony_Func_X = x
     Anony_Func_Trigger = CreateTrigger()
     Anony_Func_TriggerAction = TriggerAddAction(trig, function Anony_Func_x)
     call ExecuteTrigger(Anony_Func_Trigger)
     set x = Anony_Func_x
     call DestroyTriggerAction(Anony_Func_TriggerAction)
     call DestroyTrigger(Anony_Func_Trigger)
     call BJDeubgMsg(I2S(Anony_Func_RetVal))
     call BJDebugMsg(I2S(Anony_Func_x))

Although, you may not have to destroy these things right away. The user may want to keep his anonymous function/trigger and use it later. I'd be perfectly content if you made them a type, and that I can always just call Anony_Func.Dispose if I wanted to get rid of it.

I do see a need for both Anonymous functions (for Forgroup/ForForce/Boolexprs and all other code takers that tend to be quiet trivial), and anonymous triggers. The ideal would be both to be implemented, but I can see how its tricky.

[Vexorian]

I think I could just make it so if you use it a local inside an anonymous function it only works if you call it instantly... What I really like about the anonymous functions right now is to have a better code structure.

The user would have to destroy the trigger, some how we'll also need a hook to recycle the stuff. So a type is kind of necessary.
I really, really hate dynamic triggers ... Because I learned to distrust DestroyTrigger, the leak is one thing, but the fact that it can screw up the handle stack when you do not use it correctly (and lack of documentation means we'll never know we used it correctly...) Is very lame.

Anonymous triggers' flexibility also bounds us to triggers and hashtables, where there are really much faster ways to do that stuff.

So, how about I make anonymous functions, let use locals only when it is run right away, and be done with it?

[Vexorian]

z.5

[Dark_Dragon]

Originally Posted by Vexorian

I think I could just make it so if you use it a local inside an anonymous function it only works if you call it instantly... What I really like about the anonymous functions right now is to have a better code structure.

The user would have to destroy the trigger, some how we'll also need a hook to recycle the stuff. So a type is kind of necessary.
I really, really hate dynamic triggers ... Because I learned to distrust DestroyTrigger, the leak is one thing, but the fact that it can screw up the handle stack when you do not use it correctly (and lack of documentation means we'll never know we used it correctly...) Is very lame.

Anonymous triggers' flexibility also bound us to triggers and hashtables, where there are really much faster ways to do that stuff.

So, how about I make anonymous functions, let use locals only when it is run right away, and be done with it?

yeah i as well like that new code structure with anonymous functions!

i just want em even if u dont allow locals in them!

So, how about I make anonymous functions, let use locals only when it is run right away, and be done with it?

i agree with this!

~DD

[Troll-Brain]

thistype doesn't seem to be supported in static ifs :

JASS:

struct S
    static constant boolean B = true
    
    method m1 takes nothing returns nothing
    
        static if S.B then
            // true
            call BJDebugMsg("true")
        else
            // false
            call BJDebugMsg("false")
        endif
        
    endmethod
    
endstruct

JASS:

    function s__S_m1 takes integer this returns nothing
    
            // true
            call BJDebugMsg("true")
        
    
            // true
    endfunction

JASS:

struct S
    static constant boolean B = true
    
    method m1 takes nothing returns nothing
    
        static if thistype.B then
            // true
            call BJDebugMsg("true")
        else
            // false
            call BJDebugMsg("false")
        endif
        
    endmethod
    
endstruct

JASS:

    function s__S_m1 takes integer this returns nothing
    
            // false
            call BJDebugMsg("false")
        
    
            // false
    endfunction

Also, as you can see the 2 comments are not added but the right comment is still doubled after the end of the static if.

[grim001]

Originally Posted by Vexorian

I think I could just make it so if you use it a local inside an anonymous function it only works if you call it instantly...

Yeah, this way makes the most sense.

[Vexorian]

Troll-brain: Yes, that was expected. thistype support will get added later.

[Troll-Brain]

Originally Posted by Vexorian

Troll-brain: Yes, that was expected. thistype support will get added later.

Ok, but what about the double comment ?

[Vexorian]

I can say it is not caused by the static if phase, but something that happens later, will take a look when there's more time.