Is there a point in creating a "point" struct?

[WereElf]

struct point
    real x
    real y
endstruct

Is there a point to create such struct, since WE already provides location.
Both this struct, and the location needs creating and destroying, so I don't really see a reason to use point instead of location.

Anyone has a different point of view?

 

[Flux]

You might use something like that if you want those points to have a data structure (e.g. stack, list or tree). For example, if you want to pick all "point" created, you can easily enumerate them all via stack or list.
Another reason would be syntax sugar, if you prefer myPoint.x over GetLocationX(myPoint).

 

[DracoL1ch]

can you simply attach produced JASS before speaking of comparing?

 

[Dr Super Good]

You could add vector and matrix related functionality such as transforms.

 

[WereElf]

You could add vector and matrix related functionality such as transforms.

Can't I do that with location too? Only the syntax will be different

 

[Emm-A-]

I would prefer the faster and lighter way. Would be interesting, whether GetLocationX() is faster than location.x ...

 

[Flux]

I think the speed difference would be negligible. In terms of compiled JASS script, it's not that bad having your own point struct, just a few functions and array variables added. If it would speed up your map development significantly (due to syntax), then I say go. Plus instead of leaking handle locations, you would leak struct indices instead.
I'm just pointing out the two sides of the coin.

 

[DracoL1ch]

guys guys guys
why are you talking like "struct" is a jass native or something?
can you simply compile the function using 1st and 2nd type and provide FINAL jass code? I would compare them with each other in 5 secs

 

[Aniki]

I would compare them with each other in 5 secs

Array reads (x // p.x in vJass ) can't possible be slower than native function calls (GetLocationX(p)... so no point in doing a comparison =).

Also if one uses a native location for representing points they would be restricted to having only x and y coordinates, because there's no arbitrary SetLocationZ(p, z) (only MoveLocation(p, x, y)).

 

[Flux]

guys guys guys
why are you talking like "struct" is a jass native or something?
can you simply compile the function using 1st and 2nd type and provide FINAL jass code? I would compare them with each other in 5 secs

The compiled JASS script is not the only factor. Besides like I said, it will only add a few lines and array variables in wc3map.j (whatever the script name file is) .If you're too hurt by that, you can always use a custom Alloc (like the one in Spells section by Nestharus) to lessen the generated script by default allocator of JassHelper.

 

[WereElf]

I read somewhere, that a native function is faster than a BJ function call (even if it does nothing).
I would imagine that custom functions will be as slow as BJ functions, thus adding methods to this struct would not appeal to the "speed demons", who care a lot about code's speed efficiency.

 

[DracoL1ch]

I read somewhere, that a native function is faster than a BJ function call (even if it does nothing).
I would imagine that custom functions will be as slow as BJ functions, thus adding methods to this struct would not appeal to the "speed demons", who care a lot about code's speed efficiency.

function mySetUnitVertexColor takes unit u, integer r, integer g, integer b, integer a returns nothing
    call SetUnitVertexColor(u,r,g,b,a)
endfunction

call SetUnitVertexColor(udg_Hero[1],100,60,200,200)//x500
call mySetUnitVertexColor(udg_Hero[1],100,60,200,200)//x500

400-500 % difference, yet not even touches 1ms (~200 for native, ~800-900 for custom)

as I just said at speed test topic, depends on amount of arguments passed

function mySetUnitVertexColor takes nothing returns nothing
    call SetUnitVertexColor(udg_Hero[1],100,60,200,200)
endfunction

call SetUnitVertexColor(udg_Hero[1],100,60,200,200)//x500
call mySetUnitVertexColor()//x500

diff ~40%

 

[Flux]

Then again, single line custom functions can be inlined. Btw, what does custom functions have to do with a point struct? You access the point struct attribute through an array read not custom function.

 

[IcemanBo]

You have direct access and can directly manipulate x/y over using funtions GetLocationX/Y and MoveLocation(x/y).
It doesn't really bring new functionality, but your're closer to wanted data.

 

[Wietlol]

I read somewhere, that a native function is faster than a BJ function call (even if it does nothing).
I would imagine that custom functions will be as slow as BJ functions, thus adding methods to this struct would not appeal to the "speed demons", who care a lot about code's speed efficiency.

Code efficiency is not top priority though.
The top priority is maintainability, readability, etceterability of the code.
The simple BJ functions are just redirects with sometimes swapped arguments.
Some BJ functions like ModuloReal/Integer, RMinBJ, RMaxBJ, RAbsBJ, RSignBJ, (integer versions), TriggerRegisterPlayerUnitEvents, etc. are still useful as you would want to keep most of those things inside functions anyway.

Ofcourse, you could inline those functions but that goes at the cost of the readability of the code.