StringHash() absurdity

09-19-2009, 02:10 AM	#1
ToukoAozaki	StringHash("BLIZZARD IS RETARDED") == StringHash("blizzard is retarded") is true. Apparantly Blizzard did a retarded implementation of StringHash.

09-19-2009, 02:14 AM	#2
Vexorian	Sounds more like they wanted to make it easier to move on from gamecache to them .

09-19-2009, 02:29 AM

ToukoAozaki

Quote:

Originally Posted by Vexorian

Sounds more like they wanted to make it easier to move on from gamecache to them .

Was gamecache case-insensitive? Whoa.

BTW I've written a function to deal with this. While this is quite slower than a single native call, it will prove useful when case sensitivity is important. This will be included in StringUtils library.

JASS:

    function StringHashCS takes string s returns integer
        local integer ri = 0
        local integer li = 0
        local integer length = StringLength(s)
        local string result = ""
        local string ss
        
        loop
            exitwhen ri >= length
            set ss = SubString(s, ri, ri+1)
            if ss == "\\" then
                set result = result + SubString(s, li, ri) + "\\\\"
                
                set ri = ri + 1
                set li = ri
            elseif StringCase(ss, false) != ss then
                // this is a uppercase character; make it different.
                set result = result + SubString(s, li, ri) + "\\" + ss
                
                set ri = ri + 1
                set li = ri
            else
                set ri = ri + 1
            endif
        endloop
        
        set result = result + SubString(s, li, length)
        
        return StringHash(result)
    endfunction

09-19-2009, 03:41 PM	#4
Troll-Brain	Good to know.

09-20-2009, 06:30 PM	#5
Strilanc	I was expecting a post about the fact that the string is hashed before passing it to the function, meaning there are non-equivalent strings which will overwrite each other. Who cares about case insensitivity, when that bug-waiting-to-happen exists?

09-20-2009, 07:08 PM	#6
TheDamien	StringHash also features "slash insensitivity."

09-21-2009, 01:16 PM

ToukoAozaki

Quote:

Originally Posted by Strilanc

I was expecting a post about the fact that the string is hashed *before* passing it to the function, meaning there are non-equivalent strings which will overwrite each other.

Who cares about case insensitivity, when that bug-waiting-to-happen exists?

What do you mean?

Edit: Oh, I see. You mean the hash collision issue. Yes, it is a problem, but there are standard ways to circumvent that problem. However, since Jass doesn't support polymorphism pretty much there was no way to incorporate that inside hashtable implementation. It has to be done manually. Should I make a library for that?

BTW StringLib doesn't have any problem from this since hash is only used to determine each bytes, and confirmed no hash collision among them.

Quote:

Originally Posted by TheDamien

StringHash also features "slash insensitivity."

While I didn't know that in the first place, the nature of my implementation addressed that problem as well.

09-21-2009, 03:16 PM	#8
Troll-Brain	Well, i guess if we really need strings to save datas, game cache is the way.

09-21-2009, 05:28 PM

Vexorian

Quote:

Originally Posted by Troll-Brain

Well, i guess if we really need strings to save datas, game cache is the way.

How would that be better in anyway? I bet gc is slash insensitive as well.

09-21-2009, 05:43 PM

#10

Troll-Brain

Quote:

Originally Posted by Vexorian

How would that be better in anyway?

game cache is case sensitive ?

Quote:

I bet gc is slash insensitive as well.

I will test it later.

09-21-2009, 05:56 PM

#11

Strilanc

Quote:

Originally Posted by Troll-Brain

game cache is case sensitive ?

No it's not.

The main thing Game Cache has going for it is ... uh ... nothing. I guess it doesn't have the string collision issue, but it's vulnerable to malicious clients injecting values.

09-21-2009, 06:38 PM	#12
Troll-Brain	Hmm ok, so just forgot what i've said ...

09-21-2009, 07:02 PM	#13
Anitarf	People use StringHash? I thought that was mainly for backwards compatibility.

09-21-2009, 07:07 PM	#14
Troll-Brain	Personally, i don't use it. Currently i can't figure a case where strings are absolutely needed to link a data, where you can't do it with an integer instead. But again, maybe i'm wrong ...

09-21-2009, 07:31 PM	#15
Strilanc	I use it for storing and retrieving actions associated with mode commands. I have asserts to ensure there are no collisions (and the commands don't change, so it's safe).