Ok, Pablo, I will check the newsgroup.

José

>>>I have a two table join, one table (table1) holds only a record an the other (table2) holds 400.000 records, the match give me 400 records. The index are three numerical on both tables and the DELETED().
>>>
>>>Why if I index with a SET COLLATE TO "GENERAL" is faster (very far away) than if I do a SET COLLATE TO "MACHINE". I thought that MACHINE would be always faster than GENERAL?
>>
>>Hola Pablo,
>>
>>You don't tell us the whole story. If you want to change the collating sequence of an index, you first have to destroy it, then recreate if with the sequence you want. What happens when you do this?
>>
>I know this, :) I issued a DELETE TAG ALL and a INDEX ON , all this is done in the command window.
>
>Don't mind this is a Carlos Yohn question and he solved it, the answer is in the newsgroup.
>
>The original prg is:
>
>** con este programa parece que se cae un poco la teoría comunmente
>** aceptada de que la secuencia de ordenación 'machine' es siempre mas rápida
>** que 'general'
>** cuanto mas aumentemos el valor de 'tope' mas veremos la diferencia
>** Hay que fijarse también que rushmore dice que está usando exclusivamente
>** los índices en 'deleted()' y sin embargo afirma que el nivel de optimización es completo...
>
>** ¿Hay algo que he pasado por alto ??
>
>m.tope = 50
>
>clear
>close data all
>set safety off
>set talk on
>set escape on
>=sys(3054, 1)
>select 0
>create table muchos (campo1 n(3), campo2 n(3), campo3 n(2))
>for c1 = 1 to m.tope
> for c2 = 1 to m.tope
> for c3 = 1 to tope
> insert into muchos values (m.c1, m.c2, m.c3)
> next m.c3
> next m.c2
>next m.c1
>select 0
>create table uno (campo1 n(3), campo2 n(3), campo3 n(2))
>insert into uno values (15, 15, 15)
>*
>set collate to "general"
>=consulta()
>set collate to "spanish"
>=consulta()
>set collate to "machine"
>=consulta()
>
>function consulta
>select muchos
>delete tag all
>index on campo1 tag campo1
>index on campo2 tag campo2
>index on campo3 tag campo3
>index on deleted() tag borrados
>select uno
>delete tag all
>index on campo1 tag campo1
>index on campo2 tag campo2
>index on campo3 tag campo3
>index on deleted() tag borrados
>
>m.inicio = seconds()
>select all * ;
> from muchos, uno ;
> where uno.campo1 = muchos.campo1 ;
> and uno.campo2 = muchos.campo2 ;
> and uno.campo3 = muchos.campo3 ;
> into cursor consulta
>? set("collate") + ": "
>?? seconds() - m.inicio
>?? " Segundos"
>return