Hydra v4.2

[Moderator]

12:29, 26th Mar 2015
IcemanBo: Creates hydra at location which shoots at enemies.
Code is written very professional. Can't see any wrong with it.

 

[BPower]

The projectile system is very basic, if I find time I'll improve it a little bit more (probably not ... )

 

[IcemanBo]

In method considerTarget your only factor for "best" target is the distance to unit. But it would make sense also to check for unit's life, to potentially attack a unit with low life.
Before you make a new group enumeration, you check if old target is still valid. In my opinion the check is not needed, because you might find a "better" target in a new enumeration.
And anyway, only checking for range and UnitAlive(node.aim) is not enough for old target. You again need to call your filter function, to ensure correct result. (the owner of target might have changed, for example)

if best == null then
    set node.aim = new
else
    set node.aim = best
endif
set best  = null
set new  = null

^You don't need to null best, if it's null already. ^^

 

[Kazeon]

^You don't need to null best, if it's null already. ^^

:0

Man, those kind of optimizations are really really unnecessary. I would rather ignore those in my reviews (but not in my codes tho).

@BPow

I can't compile your code, you got some errors in summon method

        method summon takes nothing returns nothing

 

[mogulkhan1Axe]

i thought u retired? :O

 

[BPower]

I can't compile your code, you got some errors in summon method

Thank you. I forgot an endif when debug mode is disabled.

But it would make sense also to check for unit's life, to potentially attack a unit with low life.

Yes, no ... there are endless possibilities for higher priority targets (stunned, disabled, percentlife, life, distance, ...)
Now when you play the test map, you will see that it's more and more unlikely to hit an moving enemy which is more far away.
Whereas close units can still be hit, even when they are moving.

Before you make a new group enumeration, you check if old target is still valid. In my opinion the check is not needed, because you might find a "better" target in a new enumeration.

I think if an head aquired a target, it should focus it as long as possible.

And anyway, only checking for range and UnitAlive(node.aim) is not enough for old target.

Yes that's true.

You don't need to null best, if it's null already. ^^

Yes


Missiles now consider terrain height aswell.
- Missiles will be destroyed on hills, assuming that this hill has a higher z offset than the missiles offset.
- Missiles keep their fly height, when there is a canyon in their path. (visual improvement)
- Missiles will not collide with unit in canyons, assuming their z + DEFAULT_UNIT_Z offset is lower than the z offset of the missile.

Flying units can be hit, but missiles do not fly in a curve or alter their fly height
In conclusion Hydra can't shoot uphill or downhill, just on the same terrain level as it is summoned on

 

[Kazeon]

• Trigger/Code:
  • I hope you're not forget about the convention for constant variable naming: CONSTANT_VARIABLE. No exception for constant struct members. JPAG
    .

  • set d    = SquareRoot((x - originX)*(x - originX) + (y - originY)*(y - originY))

    It's not a right way to calculate total traveled distance of a missile. Should be something like this instead:
    current dist = current dist + velocity
    
    If you want to keep it that way (because the missile trajectory is just simply linear), then variable d is unneeded.
    .

  •         @If you have BoundSentinel@ in your map you don't need a periodic 
            check if the protectile is within bounds.

    You can't list WorldBounds as requirement then? If user has BoundSentinel already, WorldBounds will be no use at all whereas it's required => burden.
    .

  • readonly string modelpath

    It shouldn't be there. For what iydm? Can we read/use it for something?
    .
  • MAXIMUM_COLLISION_SIZE should be removed. That makes the missile's AoE (hit radius) calculation become very vague for user, for me either.
    

    call GroupEnumUnitsInRange(thistype.enu, x, y, aoe@+MAXIMUM_COLLISION_SIZE???@, null)

    .
  • Well, your code is very fine as always.
  .
  .
• Object data & Misc:
  • Just right
  .
  .
• Rating:
  
  I can't expect the code to be any faster. You can spam it massively without any meaningful lag or fps drop. Concept is simple, neat, and the most important: original. Object data is fine. Visual effect is just right and fitting. The spell code itself looks very modular, allowing you to implement it in very various style of spell casting.
  
  Overall 4/5 and vote for Approved​

 

[BPower]

I don't see it so strict for struct members, never did.

It's not a right way to calculate total traveled distance of a missile. Should be something like this instead:

Perfectly returns the distance between launching point and the current position and that's excacly what is needed.

You can't list WorldBounds as requirement then? If user has BoundSentinel already, WorldBounds will be no use at all whereas it's required => burden.

Hydra uses UnitIndexer, which uses WorldBounds. There is no way you don't have WorldBounds in your map if you want to have Hydra.

readonly string modelpath It shouldn't be there. For what iydm? Can we read/use it for something?

I'm not sure how to perfectly solve this, yet. Since the spell is highly configurable you can shoot different missiles e.g. fireballs or acid missiles.
If that missile collides with an enemy, you can read the used modelpath and apply for example a buff or change the damage option.
Another solution would be to assign fixed indexes to missiles. 1 for fireball, 2 for frost and read that index onCollide.

MAXIMUM_COLLISION_SIZE should be removed. That makes the missile's AoE (hit radius) calculation become very vague for user, for me either.

No!
While a missile can have a collision size of 40, a nearby townhall has a collision size of 196. It would not be within the groupenum, even though it should be.
It's also not vague because of the next line: IsUnitInRange(u, missile.unit, aoe)

 

[Kazeon]

I don't see it so strict for struct members, never did.

Said so because JavaScripting has the same convention, they use CAPITALIZED for constant struct/class/whatever members.

I'm not sure how to perfectly solve this, yet. Since the spell is highly configurable you can shoot different missiles e.g. fireballs or acid missiles.
If that missile collides with an enemy, you can read the used modelpath and apply for example a buff or change the damage option.
Another solution would be to assign fixed indexes to missiles. 1 for fireball, 2 for frost and read that index onCollide.

Well, you have returned thistype in create method. I believe user can use it instead of checking for missile's model. So removing it is your solution.

 static method create takes unit who, real x, real y @returns thistype@

No!
While a missile can have a collision size of 40, a nearby townhall has a collision size of 196. It would not be within the groupenum, even though it should be.

Is that a trick to get unit's collision size? I don't get it, each unit has different collision size.
Ah! I get it. ^^

But soo, I can conclude IsUnitInRange(u, missile.unit, aoe) will always return false for units outside the aoe? No matter if their collision collide with the aoe circle. You should make a function called IsUnitCollideCircle instead. Well it needs another function called GetUnitCollisionSize :/

 

[BPower]

Small update ...

But soo, I can conclude IsUnitInRange(u, missile.unit, aoe) will always return false for units outside the aoe ?

I though about that too, but didn't take the time to test it.

Btw I don't think the concept is neither simple nor original, but of course neat.

 

[Meatmuffin]

@Dalvengyr

Concept is simple, neat, and the most important: original.

You probably never played diablo 2.

Me being a big fan of diablo 2, this spell is awesome to me. It's fun to cast, and to remember good ol' d2.

 

[BPower]

Thank you. Feedback is very welcome

 

[Kazeon]

Btw I don't think the concept is neither simple nor original, but of course neat.

The outcome looks simple: summons 3 units which are periodically fires missile. Although it has leviathan code.
And yes it's original, afaik, since I haven't seen similliar spell around spell section so far. Seriously, you can't make something very original today, at least it's close to impossible.

 

[BPower]

Yes !

Although it has leviathan code.

The best solution would be to outsource the missile system, but xe(which is good imo) or Missile(dirac) are not my first choice.

Nice to hear you like the spell, because I also do so

 

[BPower]

Huge update.
Hydra:
- can now shoot with any attackspeed
- properly adjusts attack aninmation speed.
- improved target evaluation
- better debug options
Projectiles:
- can fly uphill/downhill
- collide with cliffs
- support onCollision/onDestructable/onExpire
- Take z axis into consideration, when it comes to collisions
- No longer uses CTL, but just one normal timer and a linked list

 

[BPower]

Fixed documentation.

 

[Kazeon]

How about just write your own missile system? I believe Dirac's will be moved to gy soon for it's current buggy state.

 

[BPower]

I thought about it for some time. Creating a system shouldn't be so difficult,
but I don't have the time to figure out a good algortihm for arcing missiles.

Dirac's needs a few obligatory fixes, but overall it is an acceptable system.

 

[BPower]

Finally I'm very satisfied with the final outcome of the Hydra spell.
If you have any criticism or ideas how to improve the current system, please feel free to tell me.

 

[Kazeon]

Good job! I'm a big fan of this spell ^^

 

[BPower]

Thank you !

 

[Deleted member 219079]

So you've optimized this to the point where you inline RMaxBJs?

Reminds me of making a program in TI-basic.

Edit: +rep for the effort

 

[BPower]

There are two small "mistakes" in the z collision calculation, depending on the missile's collision size.

1. Correct target z offset towards the ground, because currently the missile system assumes that units in collision range have the same ground level as the missile.
In rare cases this may be wrong (cliffs & missile collision size above a tile size).

2. The collision in z axis is measured from foot to head of the target and the center of the missile.
We already assume that missiles are squares in x/y collision. I will add a member for z collision, which leaves the option to make it spherical, but also egg-shaped or flat.

Update will come after the New Year.

EDIT: Updated and solved both aspects mentioned above.

 

[Nestharus]

HydraCollision review

Documentation typo
You can *savely* delete onCollide, onExpire and/or onDestructable,


I recommend making these variables private and putting them in another macro

Nothing's stopping you from using two macros.

globals
    // For demostration issues I set up unique missile strings. 
    // Do not privatize missile file variables! They must be readable from the Hydra library.
    // constant string MY_MISSILE = "path\\..."
    
    constant string CHIMAERA_ACID_MISSILE = "Abilities\\Weapons\\ChimaeraAcidMissile\\ChimaeraAcidMissile.mdl" 
    constant string FROST_WYRM_MISSILE    = "Abilities\\Weapons\\FrostWyrmMissile\\FrostWyrmMissile.mdl"
endglobals

This should be private

static method onCollide takes thistype this, unit picked returns boolean

The level of customization here is fantastic. However, if you are going to focus on nice customization like this, it might be be better to make it instanced customization rather than global static customization. For example, you can have a weaker form of the Hydra spell or a stronger form. Maybe do instanced global settings and when you create an instance of a hydra spell, you pass in what settings you want to use. I would then make the callback a part of those settings and then have general behavior.

Hydra review

These global settings are the same as mentioned above. These also may be based on level.

    globals
        // Initial missile offset in the shooting process. Use method operator missileoffset to change it individually.
        private constant real HYDRA_HEAD_DEFAULT_HEIGHT   = 55.
        // Attack animation duration.
        private constant real HEAD_ATTACK_ANIMATION_POINT = .5
        // The initial facing angle.
        private constant real FIRST_FACING_ANGLE          = 225.
        // The polar projection angle for the first head.
        private constant real FIRST_HEAD_ANGLE            = 90.*bj_DEGTORAD
        // Time to wait between spawning each head.
        private constant real HEAD_SPAWN_INTERVAL         = .23
        
        // Largest collision size in your map. Townhall e.g. has 196.
        private constant real MAXIMUM_COLLISION_SIZE      = 196.
        // Minimum fly height for projectiles, before they get destroyed.
        private constant real MINIMUM_MISSILE_OFFSET      = 0.
        // Projectiles consider the z axis aswell. Every unit uses this value as
        // default body height. In order to hit a target, it must be in 
        // DEFAULT_UNIT_Z of the missiles offset. High quality maps
        // may use a custom z size for each unit type and unit used.
        private constant real DEFAULT_UNIT_Z              = 96.
    endglobals

There is a location z library in the JASS section now -> http://www.hiveworkshop.com/forums/jass-resources-412/snippet-zlibrary-237821/

    globals
        private location LOC = Location(0., 0.)
    endglobals
    private function LocationZ takes real x, real y returns real
        call MoveLocation(LOC, x, y)
        return GetLocationZ(LOC)
    endfunction

AllocFast will do this. It is under the Alloc submission.

    // Hydra missile system. Allocation module
    // Pro: Fastest way of allocation.
    // Con: Big initial loop.
    private module PROJECTILE_INITIALIZER_CODE
        private static method onInit takes nothing returns nothing
            local integer i = 0
            set rc[8191] = 0
            loop
                set rc[i] = i + 1
                
                exitwhen i == 8190
                set i = i + 1
            endloop
        endmethod
    endmodule

There are several Linked List data structure modules that will do this

I would probably recommend this one for your purposes

https://github.com/nestharus/JASS/blob/master/jass/Data Structures/StaticUniqueList/script.j

The StaticUnique list will handle allocation for you and it does specialized allocation for lists. It'll save you some variables. It won't specifically use the AllocFast algorithm though. You can copy the module code and modify it to your needs if you want a custom solution. You can also drop the methods you don't need.

        // Linked List
        private static integer array n
        private static integer array p

Replae with Alloc

private static integer array rc

Just give the variables proper names. You shouldn't have to comment on them with their actual names.

        private real cos    // Cosinus
        private real sin    // Sinus
        private real dist   // Maximum travel distance
        private real ox     // Origin x
        private real oy     // Origin y
        private real oz     // Origin z
        private real of     // Origin fly height
        private real ag     // Angle
        private real vel    // Velocity
        private real zAdd   // Z increment per loop 
        private real zInc   // Total z increment so far

You are using ErrorMessage, so why don't you do ThrowWarning or whatever the thing is called ; ).

It seems silly to write your own.

            debug if (sfx != null) then
            debug call BJDebugMsg("WARNING: Library Hydra, struct projectile, method operator model=, sfx is not null for instance " + I2S(this)) 
            debug endif
            debug if (file == null) then
            debug call BJDebugMsg("WARNING: Library Hydra, struct projectile, method operator model=, file is null for instance " + I2S(this)) 
            debug endif

Right now you are embedding a timer system into the resource, which muddles your code with extra things like this. This is kind of sloppy.

            if (0 == n[0]) then
                call PauseTimer(thistype.t)
            endif

You don't need the DEBUG_MODE there. Also, and/or don't work with static ifs.

            static if LIBRARY_ErrorMessage and DEBUG_MODE then
                debug call ThrowError(rc[this] != -1, "Projectile", "terminate", "thistype", this, "Attempted To Deallocate Null Instance.")
            endif

Improper name
private static method cb takes nothing returns nothing

This depends on the object movement type. Be sure to let people know not to change the movement type of the object or to use a specific movement type.

                set f    = -(z - oz) + of + zInc// Keep the correct fly height,
                                                // in case the terrain elevates or flattens.

Static if doesn't support and/or
static if not LIBRARY_BoundSentinel and LIBRARY_WorldBounds then

Since you are using UnitIndexer, I recommend you pass around unit indexes instead of units. This will reduce handle ref count management. Furthermore, users will either be depending on unit-specific data, meaning that they will be using GetUnitUserData, or they will rely on the unit type id, which is a simple array read with Unit Indexer. This is a choice though, but I have personally found that passing around a UnitIndex is more beneficial than passing around a unit. Although, this largely depends on the function. Some of your functions, like unit default height, would be best with a UnitIndex.

This should probably be a ThrowWarning

            debug if (z < MINIMUM_MISSILE_OFFSET) then
            debug call BJDebugMsg(SCOPE_PREFIX + " WARNING: struct projectile, static method create, passed in z below MINIMUM_MISSILE_OFFSET")
            debug endif

A lot of extra logic that doesn't need to be in the code

            set n[this] = 0
            set p[this] = p[0]
            set n[p[0]] = this
            set p[0]    = this

There is a new resource called OnInit or something. Feel free to use this instead of writing special modular initializers from now on.

implement PROJECTILE_INITIALIZER_CODE

The regular Alloc uses this allocation method

    /*
    *   Simple allocation module.
    *   Pro: Less overhead on init.
    *   Con: Slightly slower than in Projectile. 
    */

This method exists outside of debug mode for Alloc

You can also check for memory leaks

        method operator exists takes nothing returns boolean
            return (recycler[this] == -1)
        endmethod

This, and the subsequence methods, should probably be attackSpeed, attackPeriod, attackRange, and so on. This isn't Lua : P.

There also don't seem to be getters for this stuff, only setters.

method operator attackspeed= takes real value returns nothing
method operator attackperiod= takes real value returns nothing
method operator attackrange= takes real value returns nothing
method operator missilemodel= takes string path returns nothing
method operator scale= takes real value returns nothing
method operator lowerDamageBound= takes real value returns nothing
method operator upperDamageBound= takes real value returns nothing
method operator missileoffset= takes real value returns nothing

I would use an array here

        private static method headAnimationSuffix takes  integer index returns string
            if     (0 == index) then
                return thistype.FIRST
            elseif (1 == index) then
                return thistype.SECOND
            else
                return thistype.THIRD
            endif
        endmethod

This seems to be double free detection. Alloc handles this. Also should be a ThrowError.

            debug if not this.exists then
            debug call BJDebugMsg(SCOPE_PREFIX + " ERROR: method destroy, thistype " + I2S(this) + " Is Not Allocated!")
            debug return
            debug endif

Should probably rename node to head.

            loop
                exitwhen 0 == node
                if (UnitAlive(node.unit)) then
                    call KillUnit(node.unit)
                endif
                set node.unit = null
                set node.aim  = null
                set node = node.next
            endloop

This is checked for twice in destroy for some reason. You have it once towards the top and once towards the bottom.

            static if LIBRARY_ErrorMessage and DEBUG_MODE then
                debug call ThrowError(recycler[this] != -1, "Alloc", "deallocate", "thistype", this, "Attempted To Deallocate Null Instance.")
            endif

You used ThrowWarning here but nowhere else?

            static if LIBRARY_ErrorMessage and DEBUG_MODE then
                call ThrowWarning(this.mind > this.maxd, "Hydra", "fire", "mininum - maximum damage", this, "mindamage is higher than max damage.")
            endif

p isn't a very good name

local Projectile p = Projectile.create(ox, oy, this.height*this.size, this.angle)

This should just be a repeating timer. Under the animation attack speed setter is where you want to restart a timer. A repeating timer is quite a bit faster than a timer that is started over and over again.

call TimerStart(t, a, false, function thistype.attemptShoot)

For determining the next head to fire, I think that you should store the heads inside of a queue, not a stack. Even more, I think you should store the heads in a list. When a head dies, it is removed from the list. This way you don't have to check if a head is alive during your timer loop. In fact, you should use a circular list. You have an active head in the list. The active head is always the best one to fire. If the node being removed is the active head, advance the active head forward by 1. Whenever the active head fires, advance it by 1. This will remove your loop entirely and make the entire thing O(1) instead of O(n).

            loop
                exitwhen 0 == node
                if UnitAlive(node.unit) then
                    if (node.lastShot + this.timeout <= elapsed) then
                        if (0 == picked) or (picked.lastShot > node.lastShot) then 
                            set picked    = node
                        endif
                    endif
                elseif (node.unit != null) then
                    set node.unit  = null
                    set this.count = this.count - 1
                endif
                set node = node.next
            endloop

It is ok that the heads do not have a unit index. They can't damage and they can't be damaged, meaning that they are purely visual.

I recommend that, since you aren't doing a unit index for the head, you start a timer for each one based on the lifetime. When the timer expires, remove the head from the list. This will remove your periodic check.

I recommend that this stuff turn into warnings and one error at the end.

            debug local integer error = 0
            debug if (period < 0.) then
            debug call BJDebugMsg(SCOPE_PREFIX + " ERROR: method summon, attackperiod is less than 0!")
            debug set error = error + 1
            debug endif
            
            debug if (life < 0.) then
            debug call BJDebugMsg(SCOPE_PREFIX +" ERROR: method summon, duration is less than 0!")
            debug set error = error + 1
            debug endif
                
            debug if (range < 0.) then
            debug call BJDebugMsg(SCOPE_PREFIX + " ERROR: method summon, range is less than 0!")
            debug set error = error + 1
            debug endif
                
            debug if (file == null) then
            debug call BJDebugMsg(SCOPE_PREFIX + "ERROR: method summon, modelfile is null!")
            debug set error = error + 1
            debug endif
                
            debug if (error > 0) then
            debug call BJDebugMsg(SCOPE_PREFIX + " " + I2S(error) + " errors occured within Hydra instance " + I2S(this) + ". Shutting down Hydra. Check your settings.")
            debug set error = 1/0
            debug endif

I give this a rating of 4/5. Some improvements can be made, but it should be approved.

 

[Kazeon]

There is a location z library in the JASS section now

I wouldn't recommend that library but this instead. But heck, it makes no difference.

 

[Nestharus]

Why are two of them approved?

 

[BPower]

Thank you for the extensive review. I will comment and consider everything you mentioned.
Probably after the weekend, because currently I have only little time.

 

[Deleted member 219079]

Static if doesn't support and/or

>>>>

static ifs are like normal ifs, except that a) the condition must contain only constant booleans, the and operator and the not operator and b) They are evaluated during compile time. Which means that the code that is not matched to its condition is simply ignored.

review

He only highlighted errors and advertises his own resources, how is that a review?

 

[BPower]

I started to change a few things you mentioned. However I do not agree on every point you made.

For example I'm listing Alloc as dependency, because the module initializer I wrote does the same.
For those who really wish to use Alloc/AllocFast can easily replace this small peace of code.
My personal opinion is that many users will not import a resource, if it has too many external
dependencies listed. I also wouldn't.

You were criticising the timer system I used for the Projectile. I guess you recommend to use CTL instead.
I will consider using it.

All in all it is going slowly, because I don't have much time at the moment.

 

[BPower]

Switched from Stack to List.

Replaced the global constant setup, through functions or setters.

I also tried to work on a better read-ability. Dunno if that turned out to be successfull :/.

There are still no getters implemented, because I don't see a reason for them.

 

[Almia]

@BPower
Use tabs bro. My screen is stretched. Literally

 

[BPower]

On my screen it looks nice .

I will change it soon.

 

[Nestharus]

Stuff like this will always be false because vJASS does not evaluate anything except for single boolean values that may have a not in front of them

static if LIBRARY_ErrorMessage and DEBUG_MODE then

 

[BPower]

I see, thanks. I will go throw the code again very soon.

Is there already a circular list for public use submitted somewhere (collection)?
I didn't find anything, however I have to admit, spending not much time looking for it.

 

[Nestharus]

There is. The API is changing though.

 

[BPower]

Stuff like this will always be false because vJASS does not evaluate anything except for single boolean values that may have a not in front of them

static if LIBRARY_ErrorMessage and DEBUG_MODE then

Incorrect.

      private function Init takes nothing returns nothing
            static if DEBUG_MODE and not HELLO and HELLO2 and LIBRARY_Table then
                call BJDebugMsg("HELLO WORLD")
            endif
        endfunction

Will Print HELLO WORLD.

Change one argument, so it compiles to false and it won't print anything.

 

[Nestharus]

Oops

static ifs are like normal ifs, except that a) the condition must contain only constant booleans, the and operator and the not operator and b) They are evaluated during compile time. Which means that the code that is not matched to its condition is simply ignored.

They don't do the or operator

 

[IcemanBo]

Really great spell. You made a nice code structure and it has good modularity.

//static method create takes unit who, real x, real y returns thistype 
set hydra = Hydra.create(GetTriggerUnit(), 'o002', x , y, 225.)

^The comment is not very correct, if you know what I mean.

No... I only see one thing I would add. Rest looks all fine to me. You sometimes null things that don't need to be nulled,
but I know you know it yourself and just do it because of habit. We could discuss about how to make prioroties
for the nextTarget filter, but it's okay as it is now and would probably be just more personal.

So, could you add a safety check for if hydra got removed from game? In a case the user removes units by a player
this could happen and then the hydra heads will bug. Once it's removed it should not attack anymore,
but just being destroyed properly.

 

[BPower]

Yes I can apply some code improvements and safety checks.
- The data structure should be optimized.
- I should to reconsider the current API

I'm always open for discussions, especially towards the target filter.
But I think the heads don't have to be too clever, in my opinion it's important to evaluate if
they actually can hit the target, they are aiming at.

I will make it on the weekend.

 

[BPower]

I ran in some unexpected issues. I will update the map this week.

UPDATE:
- removed setters which seemed not useful.
- removed constant functions in favour of setters, mainly applies to missile data.
- add method optimize()
- attack period is now working properly with attack speed and attack animation.
- fixed a bug: When one head couldn't attack the whole hydra was not attacking.
- heads can now be set in range of 1 to x, and are no longer constant.
- better debug features
- added 4 "labs" to explore how the system is working.

What I didn't do:
- Replaced List with SinglyCircularList which would be the best data structure.

 

[IcemanBo]

Removed hydras still attack, though with no damage. But some code still runs, and there are still the visual missles effects.

By removed I mean when the user remove them explicitly, for example with help of "PickAllUnitsByPlayer" operation.

 

[BPower]

There is a a warning message, if this case is happening. It's in attemptShoot()
I would recommend to use destroy and re-create it the way you want.
If someone really wants something like addHead() or removeHead(), then it could be easily implemented. But I don't think it's useful to have by default.

 

[IcemanBo]

How it's useful the hydra keeps shooting with no damage effect and no model anymore?
Though, yes having the possibility is not bad, too. Approved.

 

[jonbon29]

Wow I like this spell, very unique for me. I'll use this on my map!

 

[BPower]

Nice to hear. If you realise something that should be changed or you need help setting up the spell, then just make a post here

 

[BPower]

I'm planning a huge update on Hydra. What I want to change:

1.) Use Missile for projectile movement.
2.) Demo includes two types of Hydra libraries.

The first has the same structure as it is now ( just integrates Missile )
That one should be used if you just use 1-2 Hydra spells in your map.

The second has the following convention: struct MyStruct extends Hydra
That one should be used if you want to create multiple Hydra type spells in your map (3 - x).
It will allow you to controll the shooting process and collision evaluation directly in your struct.
Basically it's more flexible, with little extra overhead.

 

[BPower]

It's done. The syntax for Hydras is now.

struct FireHydra extends Hydra
struct IceHydra extends Hydra

Everything can be customized in each respective struct,
so Hydra is no longe running with long if the else condition blocks.

Example:

scope FireHydra initializer Init

    globals
        private constant integer HYDRA_ABILITY = 'A002'
        private constant integer HYDRA_UNIT_ID = 'o002'
    endglobals

    struct FireHydra extends Hydra

        private static method onRemove takes Missile missile returns boolean
            return true
        endmethod
    
        private static method onTerrain takes Missile missile returns boolean
            return true
        endmethod
    
        private static method onCollide takes Missile missile, unit hit returns boolean
            if UnitAlive(hit) and IsUnitEnemy(hit, missile.owner) then
                return UnitDamageTarget(missile.source, hit, missile.damage, false, false, ATTACK_TYPE_NORMAL, DAMAGE_TYPE_NORMAL, null)
            endif
            return false
        endmethod
    
        implement MissileStruct
        
        method onFire takes unit shooter, Missile missile returns nothing
            set missile.model      = "Abilities\\Weapons\\FireBallMissile\\FireBallMissile.mdl"
            set missile.scale      = 1.
            set missile.damage     = GetRandomReal(50, 150)
            set missile.collision  = 32.
            set missile.collisionZ = 15.
            set missile.speed      = 27.        
            call thistype.launch(missile)
        endmethod
        
        method onTargetFilter takes unit target, player owner returns boolean
            return UnitAlive(target) and IsUnitEnemy(target, owner) 
        endmethod
    
        method onRemoveHead takes unit head returns nothing
        endmethod
    
        method onSummonHead takes unit head returns nothing
        endmethod
        
    endstruct

    private function OnCast takes nothing returns boolean
        local FireHydra hydra   
        local real x    = GetSpellTargetX()
        local real y    = GetSpellTargetY() 
        local real z    = 0.
        local real face = GetRandomReal(0., 2*bj_PI)*bj_RADTODEG
        local real interval = .23

        if (GetSpellAbilityId() == HYDRA_ABILITY) then
            set hydra                     = FireHydra.create(GetTriggerUnit(), HYDRA_UNIT_ID, x, y, z, face, interval)
            set hydra.heads               = 3   
            set hydra.attackAnimationTime = .5 
            set hydra.attackSpeed         = 3.  
            set hydra.attackPeriod        = .25
            set hydra.duration            = 12
            set hydra.attackRange         = 950.
            set hydra.headHeight          = 55. 
            set hydra.scale               = 1.  
            set hydra.hasArcOrCurve      = false
            call hydra.summon()
        endif
        return false
    endfunction
    
    private function Init takes nothing returns nothing
        local trigger t = CreateTrigger()
        if UnitAddAbility(DemoMap_HERO, HYDRA_ABILITY) then
            call TriggerRegisterAnyUnitEventBJ(t, EVENT_PLAYER_UNIT_SPELL_EFFECT)
            call TriggerAddCondition(t, Condition(function OnCast))
        endif
    endfunction
    
endscope

I could do a second version with delegates and the same API for those who need a extrem performace.
This will finally generate much more code in your map file. Dunno if it's simething we need.

 

[Blightsower]

hi, i came accross this while trying to have fun with your skill

 

[Nestharus]

Oh, the urls are based on the filenames : |. The filenames changed from script.j to main.j, heh >.<.

Anyways, uh oh o-o. That comes from reading a deallocated node. List is still compliant with the collection testing suite, so it's definitely not a problem with the lib : ).