[v]Jass - Poly Slash v0.2

[Moderator]

9th Nov 2011
Bribe: It's just another knockback. There are hundreds of these. Yours uses lots of function interfaces, bad API and for what benefit over the others?

Your "onLoop" method is also called via function interface making this much slower and bulkier than it needs to be.

Rising_Dusk's knockback is better and I consider this redundant in its favor.

 

[Magtheridon96]

Reading the code.
First of all, good job on using a library

edit

Uhm..

private constant attacktype ATTACK = null
private constant damagetype DAMAGE = DAMAGE_TYPE_UNIVERSAL
private constant weapontype WEAPON = WEAPON_TYPE_WHOKNOWS

->

private constant attacktype ATTACK = ATTACK_TYPE_NORMAL
private constant damagetype DAMAGE = DAMAGE_TYPE_UNIVERSAL
private constant weapontype WEAPON = null

Another thing.. don't use onDestroy..

Also, to check if a unit is alive, you could simply use this: GetWidgetLife(u)>=0.405

Finally, don't use TriggerAddAction because TriggerAddCondition is better and faster
You just have to return false at the end ^^

Oh and:

call TriggerRegisterAnyUnitEventBJ( trig, EVENT_PLAYER_UNIT_SPELL_CAST )
->
call TriggerRegisterAnyUnitEventBJ( trig, EVENT_PLAYER_UNIT_SPELL_EFFECT )

Also, when it comes to locals, try to make the names as short as possible for efficieny

 

[Thanathos]

Thanks for your hints Magtheridon96

Regarding the damage options my memory didnt serve me very well, i might change that later.

Since i'm trying to exploit OOP i need the onDestroy in the polyslash_move struct but i might get rid of the onDestroy in polyslash_data.

If i recall correctly there was once an issue with GetWidgetLife(u)>=0.405 and thats why i prefer GetUnitState(). And i think technically there is not much different about those two natives.

In future i will prefer TriggerConditions over TriggerActions

Is there a big difference between EVENT_PLAYER_UNIT_SPELL_CAST and EVENT_PLAYER_UNIT_SPELL_EFFECT ? I will change it anyway.

Im going to update this as soon as a mod reviewd this.

 

[Adiktuz]

Spell cast is abusable... Since spell cast event is fired right when you start casting, you can actually fire the trigger without costing mana and cooldown if you hit the Stop button fast enough...

 

[baassee]

GetWidgetLife is faster than GetUnitState, that's the difference.

 

[Luorax]

No, you don't need onDestroy. Simply rename it to "destroy" and call "this.deallocate()" at the end of the method. Job well done, no more trigger evaluation, and it's just a simple line.

 

[Deuterium]

Nice to see you still active Thanathos
& I see your back baassee

Yeah the GetWidgetLife is preferable although GetUnitState would work... it's just better coding (speedwise it BARELY differs, but we tend to squeeze as much speed as possible ). But in my opinion, not much of a big deal.

But the spell cast is an issue you need to fix

EDIT:
& in my opinion, start fixing before the mod checks it out... why wait?

 

[Thanathos]

Updated the spell to v0.2

Fixed all issues mentioned by Magtheridon96, except the onDestroy problem (For more information see changelog).
I've tried to do it the way Luorax said, but i somehow got malfunctions and errors, which is probably because of the interface or the multiple struct inheritations. However i will repair this as fast as possible.

Further thanks and rep+ to you all for your hints

And its also nice to see you again seed

 

[Luorax]

I've tried to do it the way Luorax said, but i somehow got malfunctions and errors, which is probably because of the interface or the multiple struct inheritations. However i will repair this as fast as possible.

Exactly that's what causes errors I'll check the code again when the TSM vs Fnatic match ends and will come back with some suggestions.

EDIT: okay, so that's how I'd do this:

1, Add this to the basic interface (Ext): method doDestroy takes nothing returns nothing defaults nothing

2, rename onDestroy() in struct "Indexable" to destroy, call this.doDestroy() and then call this.deallocate()

3, rename onDestroy() in struct "TNTKnock" and in the move struct to doDestroy()

4, call super.doDestroy() at the end of the doDestroy() method in struct move.

I think that's all.

EDIT2: however I just realised that the "doDestroy" method in TNTKnockback is useless. You're initializing those members in the create method(s) and nulling them is useless. You should simply remove that method and you don't have to call super.doDestroy() in your move struct.
And a hotkey would be fine Like E or R

 

[Thanathos]

Well that would do it but its kinda pointless.

The idea of replacing onDestroy with destroy and calling deallocate instead is to boost the performance by getting rid of the trigger mechanism introduced Vex to manage the destruction of a struct instance. If this works successfully you avoid the slower triggers and have just some nested function calls instead.

If i now would use an interface method for destroying my structs JassHelper will add an trigger mechanism (similar to that one introduced by Vex) that takes care of the calls of that certain method. So i would just change the names of the methods while the desired performance boost will simply not occur.

Further i think that Vex's built in destructor would be still more efficient than a custom mechanism, since Vex's destructor is optimized while the custom mechanism i just an "all purpose" interface method.

All i know is that your way would be at least that efficient than Vex's method, to make a definite statement i need to compare the compiled version of your mechanism and Vex's destructor.

Besides, have you ever seen a vJass script which has been turned into normal jass by JassHelper ? It looks horrible xD

 

[Luorax]

I did not know that interfaces use triggers But you know, we learn something every day (However I should have known, it's just simple logic)
But there's still no need for the Knockback onDestroy method.
Yea, vJASS looks horrible in normal JASS. However array structs are okayish.

 

[Deuterium]

would have loved to join the discussion on the whole destroy issue but after 2 years of no-coding my jass is somehow rusty