>>>Hi everybody,
>>>
>>>How can I randomly select n (110) codes from some huge table, so these codes would be unique? I'm trying to create a test case now...
>>
>>

>>*   Assumes V. 7 for RW cursor, but can emulate in prev versions.
>>
>>SELECT cCodes, .F. AS lSelected ;
>>   FROM BigTable ;
>>   INTO CURSOR UniqueCodes readwrite ;
>>   GROUP BY cCodes
>>
>>lnCodes = _Tally
>>
>>SELECT UniqueCodes
>>
>>*  This approach to using RAND() will give randomized numbers, but
>>*  the same sequence each time it is run (this is helpful for testing
>>*  to determine that changes in execution speed are due to algorithm
>>*  changes rather than the distribution of the values).
>>
>>*  If you want a different series between runs, you can issue a
>>*  preliminary call to RAND (-1) which will seed with a value
>>*  from the system time clock. For that, uncomment the line below
>>
>>*  = RAND (-1)
>>
>>lnSelected = 0
>>DO WHILE lnSelected < 110
>>   GO lnCodes * RAND () + 1
>>   IF NOT lSelected
>>       REPLACE lSelected WITH .T.
>>       lnSelected = lnSelected + 1
>>   ENDIF
>>ENDFOR
>>
>>SELECT cCodes ;
>>   FROM UniqueCodes ;
>>   WHERE lSelected ;
>>   INTO CURSOR SelectedCodes
>>
>>

>
>I'd rather create cursor SelectedCodes and insert records into it from within the Do While lnSelected<110 loop, so you wouldn't have to run a select against a long table in the end. But other than that, this seems to be the ticket.

Dragan --

Thanks for your thoughts.

From Nadya's comments, it's clear that the original table was large. But, it wasn't clear the number of codes. Even ZIPs only number about 50,000 at last count.

So, if the codes are a relatively small number (and I'd put ZIPs in that category), I think either approach would work well. If that assumption is incorrect, then I'd agree, that populating a new cursor in the loop would be the way to go.

Jay