Question : Problems with stored procedure and SELECT INTO

In a stored procedure I try to use two conditional SELECT INTO statements which select into the same temporary table, but they create an error.

.
.
.
if @inst_type_id = '1'
begin
  select d.ORG_ID, d.REPORTING_TYPE_ID, d.REPORTING_PERIOD, d.FINANCIAL_PERIOD,
         #incremental.CREATION_ID, #incremental.CREATION_DATE, EXTRACT = 0
    into #dvc
    from MAIN..DVC d, MAIN..INSTALL_ID_ORG io, MAIN..INSTALLATION_ID i, #auth, #incremental
    where i.ME_IND = 'Y'
      and i.INSTALLATION_ID = io.INSTALLATION_ID
      and io.ORG_ID = d.ORG_ID
      and io.ORG_ID = #auth.ORG_ID
      and d.ORG_ID = #incremental.ORG_ID
      and d.REPORTING_PERIOD = #incremental.REPORTING_PERIOD
      and d.FINANCIAL_PERIOD = #incremental.FINANCIAL_PERIOD
end

if @inst_type_id = '2'
begin
  select ORG_ID, REPORTING_TYPE_ID, REPORTING_PERIOD, FINANCIAL_PERIOD,
         CREATION_ID = null, CREATION_DATE = null, EXTRACT = 1
    into #dvc
    from MAIN..ORDER_ITEM_DVC
    where ORACLE_STATUS = -1
end
..
.


When I create the procedure I get the error message:
There is already an object named '#dvc' in the database.

I know I can use a create table #dvc ... first and then insert the records, but I understand from a performance point of view, it is better to use SELECT  .. INTO

Can this be achieved in a stored procedure?

We use Sybase ASE 12.5

Answer : Problems with stored procedure and SELECT INTO

The problem is that a "select into" is an implicit CREATE TABLE command.  The compiler sees that there are two commands to create the same table and unfortunately doesn't recognize that only one of those statements will actually be executed.  This is actually intentional I believe, and I wouldn't expect it to change.

A possible way around this might be to select all this with a null set, which should create the table.  Sort of like Select a, b, c from TableX where 1-2.  I believe this will create the table for you and set up all your columns.  Then you can do a simple insert, and you should be fine.

If not, you'll have to create the table upfront and specify all the columns and have standard insert statements within the if-end sections.

I know this seems like extra BS to go through, but this does actually help the code be more maintanable.  Having two possible "select into"  branches raises the possibility that the tables won't be identical, and it's a pain tracking down a bug like that when it shows up.

Greg