Inserting Randomly Selected Records SQL

Question

I'm using Microsoft SQL Server 2005.

I'm creating a random record generator that will insert 10 records randomly into a temporary table. The records in the temporary table will then be used to update records in the table in memory.

Here is the statement that is giving me some troubles (assume the temp table is already created).

    insert into #tempTable
    select top (10 - @totalOverShort)
      d.depositid, d.location, d.amount, d.count, d.user_add, 
      d.date_add, d.status, d.comments, d.subtotal_difference, d.count_difference, d.user_checked, 
      d.date_updated, d.ip_checked, newID() as randomColumn
   from
      closing_balance..cb_depositbag as d
      left join
      #tempTable as t on d.depositid <> t.depositid and d.location <> t.location
   where
      d.date_add between @weekPrior and @weekPriorNight and 
      d.status = 'R' and d.depositID <> t.depositID
   order by
      randomColumn

This statement does give me randomly generated columns, but sometimes (about 1/10) I get 1 or more duplicates in the temp table. The table in memory has a 2-part key, DepositID and Location. I want to say that if there does not exit a randomly selected (depositID,location) pair, then insert that record into the temp table. Is there another way to do this? I'm thinking the reason that there are duplicates is because it's evaluating the select statement upwards of 10 times, which, since it's being randomly ordered, can cause a duplicate. I'm not sure though.

Thanks in advance.

Answer 1

Try adding DISTINCT . This should remove the duplicate depositID,location pairs.

However, you'd also need to put NEWID() only in the ORDER BY and remove "randomColumn"

Edit: remove the 1=1. It's pointless.

After comment: However, this may not give you 10 rows if the inner query gives dupes...

select DISTINCT  *
FROM
(
select top (10 - @totalOverShort)
  d.depositid, d.location, d.amount, d.count, d.user_add, 
  d.date_add, d.status, d.comments, d.subtotal_difference, d.count_difference, d.user_checked, 
  d.date_updated, d.ip_checked
from
  closing_balance..cb_depositbag as d
  left join
  #tempTable as t on d.depositid <> t.depositid and d.location <> t.location
where
  d.date_add between @weekPrior and @weekPriorNight and 
  d.status = 'R' and d.depositID <> t.depositID
order by
  newid()
) foo

Answer 2

Use WHERE NOT EXISTS:

 insert into #tempTable
    select top (10 - @totalOverShort)
      d.depositid, d.location, d.amount, d.count, d.user_add,
      d.date_add, d.status, d.comments, d.subtotal_difference,
      d.count_difference, d.user_checked, 
      d.date_updated, d.ip_checked, newID() as randomColumn
   from
      closing_balance..cb_depositbag as d
      left join
      #tempTable as t on d.depositid <> t.depositid and d.location <> t.location
   where
      d.date_add between @weekPrior and @weekPriorNight and
      d.status = 'R' and d.depositID <> t.depositID   order by      randomColumn

WHERE NOT EXISTS
(SELECT 1
FROM #tempTable
WHERE depositid = d.depositid);