UUID>>hash should be removed

> UUID [7.7.1 - 2] included in "preview" for VisualWorks 7.8.1 should
> probably have UUID>>hash removed. The method considers only a few of
> the bytes in the UUID, which under "normal" circumstances is OK.
> However, Oracle-generated UUIDs happen to mostly differ on bytes
> outside the bytes used in the function.
>
> The best solution would be to remove UUID>>hash and rely on the super
> class' (ByteArray) implementation. In ByteArray, primitive 1700 is
> used. My limited tests shows that it performs faster than UUID>>hash.
>
> The following code can be used to test performance:
>
> |id|
> id := UUID version4.
> [10000000 timesRepeat: [id hash] ] millisecondsToRun
>
> Kind regards,
> Runar Jordahl
> _______________________________________________
> vwnc mailing list
> [hidden email]
> http://lists.cs.uiuc.edu/mailman/listinfo/vwnc
>

> UUID [7.7.1 - 2] included in "preview" for VisualWorks 7.8.1 should
> probably have UUID>>hash removed. The method considers only a few of
> the bytes in the UUID, which under "normal" circumstances is OK.
> However, Oracle-generated UUIDs happen to mostly differ on bytes
> outside the bytes used in the function.
>
> The best solution would be to remove UUID>>hash and rely on the super
> class' (ByteArray) implementation. In ByteArray, primitive 1700 is
> used. My limited tests shows that it performs faster than UUID>>hash.
>
> The following code can be used to test performance:
>
> |id|
> id := UUID version4.
> [10000000 timesRepeat: [id hash] ] millisecondsToRun
>
> Kind regards,
> Runar Jordahl
> _______________________________________________
> vwnc mailing list
> [hidden email]
> http://lists.cs.uiuc.edu/mailman/listinfo/vwnc
>

> I also would like to point out that sometimes hash function evaluation
> speed is not the most important factor that drives hashed collection
> performance.  As an extreme example,
>
> hash
>
> ^1
>
>
> is super fast but useless.

> On 6/28/2012 8:46 AM, Andres Valloud wrote:
>> I also would like to point out that sometimes hash function evaluation
>> speed is not the most important factor that drives hashed collection
>> performance.  As an extreme example,
>>
>> hash
>>
>> ^1
>>
>>
>> is super fast but useless.
>
> Oh, perhaps you should switch to using random numbers with this technique:
>
>       http://imgs.xkcd.com/comics/random_number.png
>
>
> ;-)
>
>
>
>
> _______________________________________________
> vwnc mailing list
> [hidden email]
> http://lists.cs.uiuc.edu/mailman/listinfo/vwnc
>