Is this a bug?

08-17-2007, 04:51 PM

Am i blind or is this a bug?:

JASS:


function l2r takes location l returns real
    return l
    return 0.
endfunction

function i2r takes integer i returns real
    return i
    return 0.
endfunction    

function r2i takes real r returns integer
    return r
    return 0
endfunction    

function Trig_a_Actions takes nothing returns nothing
    local location l = Location(0,0)
    local integer i = 0
    call BJDebugMsg("addr: "+I2S(r2i(l2r(l))))
    set i = r2i(l2r(l))
    if i==0 then
        call BJDebugMsg("error 1: "+I2S(i)+"; addr: "+I2S(r2i(l2r(l))))
    elseif r2i(l2r(l))==0 then
        call BJDebugMsg("error 2: "+I2S(i)+"; addr: "+I2S(r2i(l2r(l))))
    endif
endfunction

//===========================================================================
function InitTrig_a takes nothing returns nothing
    set gg_trg_a = CreateTrigger(  )
    call TriggerRegisterTimerEventSingle( gg_trg_a, 0.20 )
    call TriggerAddAction( gg_trg_a, function Trig_a_Actions )
endfunction

The local i is set to the value r2i(l2r(l)) returns. The expression returns the address of the location created at the start of the function.
So i expected both tests to show the same result: false

But:
call BJDebugMsg("addr: "+I2S(r2i(l2r(l)))) shows a value not equal to 0
r2i(l2r(l))==0 it returns true
i==0 returns false

Did i make a mistake? If so i must be blind atm.

08-17-2007, 05:10 PM	#2
cohadar	r2i and i2r are soooooo bad idea Not all real numbers are valid.

08-17-2007, 05:34 PM

Deckname

Quote:

Originally Posted by cohadar

r2i and i2r are soooooo bad idea

Not all real numbers are valid.

What does this have to do with the code?
How does it explaint the behaviour of the script?

08-17-2007, 05:49 PM

cohadar

Ok a lesson in floating-point arithmetic.

integers: 45, -5, 78, 456
real numbers: 1.0, -0.0001, 0.45e-3

See that last one?
All real are made from 2 parts: mantissa and exponent
0.45e-3 == 0.45 * 10^-3 (0.45 mantissa, -3 exponent)

in reals -0e-4 <> 0e-4 (negative zero exists)

real numbers also have special values that
represent -infinity and +infinity

The problem is that mantissa and exponent can have invalid values
and that kind of number is NaN (not a number)

This is a messy stuff, you can learn more about it if you type "IEEE floating" in google

Bottom line: DON'T use I2R and R2I

08-17-2007, 06:30 PM

Deckname

@cohadar:

And how does it affect this (i know floating point representation btw):

JASS:

set i = r2i(l2r(l))

    if i==0 then
        call BJDebugMsg("i==0")
    else
        call BJDebugMsg("i!=0")
    endif

    if r2i(l2r(l))==0 then
        call BJDebugMsg("r2i(l2r(l))==0")
    else
        call BJDebugMsg("r2i(l2r(l))!=0")
    endif

    if i==r2i(l2r(l)) then
        call BJDebugMsg("i==r2i(l2r(l))")
    else
        call BJDebugMsg("r2i(l2r(l))!=0")
    endif

    if r2i(l2r(l))==r2i(l2r(l)) then
        call BJDebugMsg("r2i(l2r(l))==r2i(l2r(l))")
    else
        call BJDebugMsg("r2i(l2r(l))!=r2i(l2r(l))")
    endif

It prints:
"i!=0"
"r2i(l2r(l))==0"
"i==r2i(l2r(l))"
"r2i(l2r(l))==r2i(l2r(l))"

So i is not 0. But r2i(l2r(l)) returns 0.
But both should have the same value since I set i = r2i(l2r(l))
And in addition do i and r2i(l2r(l)) have the same value when i compare them.

And i is an integer and r2i(l2r(l)) returns an integer.

08-17-2007, 06:47 PM	#6
cohadar	Look I don't want to bother why exactly it does not work in your trigger, it simply does not because handle casting ALWAYS makes problems and especially when you do i2r and r2i. So here is what: You tell me what is exactly that you want to do and I will tell you how to do it without i2r and r2i.

08-17-2007, 07:12 PM

Deckname

Quote:

Originally Posted by cohadar

Look I don't want to bother why exactly it does not work in your trigger,
it simply does not because handle casting ALWAYS makes problems
and especially when you do i2r and r2i.

Well, they seem to work.
But there is a problem when using the values in the way i do it in the example.
(But when you store the value in the integer it seems to work correctly)

Quote:

So here is what:
You tell me what is exactly that you want to do
and I will tell you how to do it without i2r and r2i.

I want to get the address of a location.
Then i want to assign the address of that location to the x coordinate of a different location. But i have a solution for that.

08-17-2007, 07:23 PM	#8
cohadar	If you are trying to make linked lists than DON'T DO IT Vexorian already made linked list system with locations like 1000 years ago, but it is deprecated now, there are much better systems now. (like the ones from my signature if I may notice...)

08-17-2007, 07:46 PM	#9
Ammorth	@ Deckname, there is a bug with type-casting reals/integers while working with handles. Storing the value to a global fixes this (at least thats what I remember). @ Cohadar, it really depends on what you need the list for.

08-17-2007, 08:03 PM

#10

cohadar

Quote:

Originally Posted by Ammorth

@ Cohadar, it really depends on what you need the list for.

Well people usually need lists when they have variable number of data
and they want to add/remove data arbitrarily from the list.

Collections do that so I don't see why he should not use them.

08-19-2007, 01:30 PM	#11
Pyrogasm	iNfraNe made the Location Linked Lists, not Vexorian.

08-19-2007, 01:58 PM	#12
cohadar	Oh pardon me, I saw them in an earlier version of CS_ system so I assumed...