Hi Fabio,

I got completely different result from yours in VFP9 RTM(2412). The time for both queries is roughly the same

Rushmore optimization level for intermediate result: none
Rushmore optimization level for intermediate result: none
Joining intermediate result and intermediate result using temp index
0.177 98200
Rushmore optimization level for intermediate result: none
Rushmore optimization level for intermediate result: none
Joining intermediate result and intermediate result using temp index
0.179 98200

>This have a hard impact for a WHERE filtering,
>but on a join ( or multijoin !) this put VFP into this class:
>"Customer use task manager for kill me please"
>
>Try this on VFP8 and VFP9
>

>CLEAR
>SET TALK OFF
>SET ANSI OFF
>SYS(3054,11)
>CLOSE TABLES all
>SYS(1104)
>
>#DEFINE lnMaxKeys  10
>#DEFINE lnMaxRec   10000
>CREATE CURSOR lossitem ( Lit_LosFK C(10))
>FOR i=1 TO lnMaxRec
>	INSERT INTO lossitem  VALUES (STR(RECCOUNT()))
>ENDFOR
>FOR K=1 TO 10000000 && BURNIN CPU
>NEXT
>
>T1=SECONDS()
>select COUNT(*) NR;
>   from LOSSITEM X JOIN LOSSITEM Y ON RIGHT(X.Lit_LosFK,3) == RIGHT(Y.Lit_LosFK,3);
>   into cursor RESULT nofilter
>? SECONDS()-T1,NR  && VFP8 = 0,200 VFP9b = 0.110 s  <== !!!!
>
>T1=SECONDS()
>select COUNT(*) NR;
>   from LOSSITEM X JOIN LOSSITEM Y ON RIGHT(X.Lit_LosFK,3) = RIGHT(Y.Lit_LosFK,3);
>   into cursor RESULT nofilter
>? SECONDS()-T1,NR && VFP8 = 0,180 VFP9b = 86.634 s
>

>
>If this is not written in the VFP9's documentation,
>with Font Bold 42,
>I assume many future curses.
>
>Surprise, i see VFP9b faster with == comparison ( COLLATE optimization ? );
>with another test i see a 8X speed up.
>
>But with = i leave any comment to you.
>
>Fabio