>>John,
>>
>>This is the idea I got from Al Doman quite recently (search "Comparing two arrays" thread):
>>
>>You can do it this way:
>>pseudo-code
>>

  local lcUniqueTableName, lcGroupBy, i, lnFields
>>lcUniqueTableName=sys(2015) && could not remember this sys
>>select FreeTable
>>copy structure to (m.lcUniqueTableName)
>>select (m.lcUniqueTableName)
>>append from FreeTable
>>append from DBCTable
>>lcGroupBy=""
>>lnFields=fcount()
>>for i=1 to m.lnFields
>>    lcGroupBy=m.lcGroupBy+field(m.i)+iif(m.i=m.lnFields,'',',')
>>endif
>>
>>select *, count(*) as cntDups from (m.lcUniqueTableName) ;
>>   group by &lcGroupBy ;
>>   having cntDups>=2 ;
>>   into cursor curDups
>>
>>select *, count(*) as cntUnique from (m.lcUniqueTableName) ;
>>   group by &lcGroupBy ;
>>   having cntUnique=1 ;
>>   into cursor curUnique
>>
>>use in select(m.lcUniqueTableName)
>>delete file (m.lcUniqueTableName+".*")
>>

>>
>
>Nadya, that is the one I was thinking of, thanks!
>John

I can not guarantee the speed, though. Can you test other solutions too (that Brien gave you) and make benchmarking? BTW, I made few minor corrections in the code.

Thanks in advance.