Re: [Axiom-developer] Spad and its object model

[Stephen Wilson]

Gabriel Dos Reis <address@hidden> writes:

> On Tue, 12 Jun 2007, Stephen Wilson wrote:
> 
> | > | Global change to a hash would certainly cost us in execution time.
> | > 
> | > Why does that cost more than what you have proposed so far?
> | 
> | Lookup via hash vs. direct indexing?  Over millions of iterations?  Im
> | confident the current implementation would win. 
> 
> Except that in the scheme you proposed, the result of the indexing is not
> directly the operation, but another step must be taken.

Just for the sake of completeness:

 I argued that integer indexes were one of several possible methods
available to interrorgate a domain vector.  I posited that the vast
majority of such interrorgations could be handled using integer keys,
the remainder could use alternative keys.  The cost of using
alternative keys could be made small, on par with a hash. Thus, the
ratio of lookups using alternative keys / integer keys would be small.
Performance would not be impacted, and you have all the flexibility
which you need in order to adress your issues with coherence.


Steve