>>>Correct! But you can generate random numbers to pick records from the table. As soon as a record is picked, you flag it or delete it... :)
>>>
>>>Don't take me seriously here! I'm only joking! You don't need to use the long way if you know the short one. :)
>>>
>>>Vlad
>>>
>>>>A quick test of SYS(2015) gave me:
>>>>
>>>>_RQ91824UH
>>>>_RQ91824UI
>>>>_RQ91824UJ
>>>>_RQ91824UK
>>>>_RQ91824UL
>>>>_RQ91824UM
>>>>_RQ91824UN
>>>>_RQ91824UO
>>>>
>>>>This appears to be unique but sequential, so if I then sort on this number, the result is not randomized. In fact the order may not change at all...
>>>>
>>>>thx
>>>>
>>>>>>No... not at all...
>>>>>>
>>>>>>RAND() works fine, even with the additional test for uniqueness. I was just wondering if there was an internal function to achieve the same thing.
>>>>>
>>>>>Well, you can use the sys(2015) to get unique keys. Of course, you can convert chars to digits. This will give you unique numbers, but not random numbers. I wouldn't use this!
>>>>>
>>>>>Vlad
>>>>>
>>>>>>>As I said: if it's random, then it's not unique. So, it doesn't matter what method you use to get random numbers, you must test them to be sure you choose each number only once.
>>>>>>>
>>>>>>>Why? Why is RAND() not good? Too slow?
>>>>>>>
>>>>>>>Vlad
>>>>>>>
>>hi,
>>
>>>>>>>>Is there another way to shuffle a list of phone numbers so that they are random?
>>>>>>>>
>>>>>>>>thx
>>>>>>>>
>>>>>>>>>If it's unique, it's not random! And if it's random than it's not necessary unique! :)
>>>>>>>>>
>>>>>>>>>Vlad
>>>>>>>>>
>>>>>>>>>>Yeah, I was just hoping for something like RAND(,U) <-- a unique parameter.
>>>>>>>>>>
>>>>>>>>>>I'm told that C++'s RAND() function returns unique numbers...
>>>>>>>>>>
>>>>>>>>>>Oh well, I did the SEEK to verify uniqueness and it works fine.
>>>>>>>>>>
>>>>>>>>>>Thanks
>>>>>>>>>>
>>>>>>>>>>>This is normal. Random numbers are not unique.
>>>>>>>>>>>
>>>>>>>>>>>Each time you generate a new number you must do a SEEK and add it only if it is not already in your table.
>>>>>>>>>>>
>>>>>>>>>>>Vlad
>>>>>>>>>>>
>>>>>>>>>>>>I fill a file having 12000 records with 'random' numbers. Index on number unique
>>>>>>>>>>>>shows only 7550 unique numbers
>>>>>>>>>>>>
>>>>>>>>>>>>>What's the problem? Ie: What doesn't work?
>>>>>>>>>>>>>
>>>>>>>>>>>>>Vlad
>>>>>>>>>>>>>
>>>>>>>>>>>>>>We are using RAND() to create a random sampling of a phone number file so
>>>>>>>>>>>>>>we need unique numbers so we don't re-use phone numbers
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>Any way?
>>
>>hi,
>>the way is to make unique or primary index in the table,
>>generate random number and replace it into table. If you move from this
>>record, unique type of index causes error event, so you shouldn't seek for
>>this random (but potentially non-unique) number. This way is better, especially
>>if you want to share table with another users (or using buffering)!.
>>
>>Peter
>
>Why is this better? I assume that you must change&restore the on error behavior to do this, isn't it?
>
>Vlad

This is better, if you use buffering in multiuser environment, because
another user can generate the same number as you if you have buffered
current record - it is "invisible" for him till you update your buffer.
Causing Error events in primary and unique indexes is especially
intended to cases like that.
I think, the best technique is buffer a record, generate number, try
to update record, if error event comes generate new number etc.
(Updating record causes error event in index without changing recordpointer).

Peter