Small Vectors Proposal

[Mikola Lysenko]

Frits van Bommel wrote:
> Mikola Lysenko wrote:
>> Hmm...  Fair enough.  We could define ordering on a per component 
>> level arbitrarily so associative arrays work.  I could see this being 
>> confusing in some situations, but it would be worth it in order to 
>> make the types consistent.  Therefore:
>>
>> float3(x, y, z) < float3(a, b, c)
>>
>> would translate too:
>>
>> (x < a) ? true : ((y < b) ? true : (z < c))
> 
> Your code is wrong when e.g. (x > a) and (y < b). It returns false true 
> while it should be false; x != a means ((x, y, z) < (a, b, c)) == (x < a).
> 
> Think of the order of words in a dictionary: first you compare the first 
> character, if it's equal you go on to the second, and so on. The first 
> character that differs determines the sort order.
> 
> It'd be implemented something like this:
> ---
> int compare (short3 a, short3 b)
>     if (a.x != b.x) return a.x - b.x;
>     if (a.y != b.y) return a.y - b.y;
>     if (a.z != b.z) return a.z - b.z;
>     return 0;   // all elements equal, so the vectors are equal
> }
> ---
> Note 1: I used shorts to produce clear code but avoid overflows.
> Note 2: This uses the opCmp/TypeInfo.compare return value convention: 
> negative means a < b, positive means a > b, zero means equality.
> 


Whoops!  I wasn't really thinking too carefully when I wrote that. 
You're correct, a lexicographic ordering is definitely the simplest in 
this case.