T-SQL equivalent of Oracle's collection methods (first, last, next)? - sql

Question

We're moving from Oracle to SQL Server and I'm converting a query from a table variable populated with a BULK COLLECT INTO query. I'm thinking of using a cursor (definitely open to other suggestions), but in the Oracle code that processes the query it's using Table_var.FIRST .NEXT and .LAST. Here's some sample code of how it's using these. It appears that the first/next/last are giving indexes into the table var's records.
TYPE Pers_DOB_LastInitial IS RECORD (
Person_ID Person.Person_ID%TYPE,
DOB Person.Birthdate%TYPE,
LastInitial VARCHAR2(1)
);
TYPE Dup_Table IS TABLE OF Pers_DOB_LastInitial INDEX BY BINARY_INTEGER;
Dup_Tab Dup_Table;
and a function that uses these types:
FUNCTION Last_In_Group( pStart NUMBER, pDOB Person.Birthdate%TYPE, pLastInitial VARCHAR2 )
RETURN NUMBER IS
vResult NUMBER;
BEGIN
IF pStart = Dup_Tab.LAST THEN
RETURN pStart;
END IF;
vResult := pStart;
FOR vIndex IN pStart .. Dup_Tab.LAST LOOP
IF Dup_Tab.EXISTS( vIndex ) THEN
IF Dup_Tab( vIndex ).DOB = pDOB AND Dup_Tab( vIndex ).LastInitial = pLastInitial THEN
vResult := vIndex;
ELSE
EXIT;
END IF;
END IF;
END LOOP;
RETURN vResult;
END Last_In_Group;
I don't need the coding done for me, just need to be pointed in the right direction. I'm thinking of using a cursor, but the only thing I'm familiar with is simply fetching the next record from a cursor in T-SQL and want to see if there are equivalent ways to reference row indexes for cursors (or temp tables).
EDIT: I've just discovered the following and am looking into it. Definitely open to hints on whether this is a good path to pursue or if cursors are still better.
http://www.sql-server-performance.com/2004/operations-no-cursors/2/
DECLARE #dupTab TABLE (
person_id numeric(8,0),
DOB date,
LastInitial char(1)
)
INSERT #dupTab
SELECT ...

Answer 1

Not 100% sure what you're asking, but have you looked at the new LEAD, LAG, FIRST_VALUE, and LAST_VALUE keywords in Sql Server 2012?

Answer 2

Assuming your table is called Dup_Tab, and #DOB and #LastInitial are variables you passed into your function, might this work?
select PersonID from(
select
rank = row_number() over (order by DOB DESC, LastInitial DESC),
PersonID
from dup_tab where DOB=#DOB and LastInitial=#LastInitial
) subgroup
where rank=1
It's just using the windowed Rank() function to assign each record a row number based on DOB/LastInitial ordered last-first and then you select the row with rank = 1, giving you the last record in the set of persons with a specific DOB and LastInitial.

Answer 3

I translated your sample code into it's SQL-Server counterpart. If you want to do row-by-row processing the best performing way is imo to use a FAST_FORWARD cursor.
code:
IF EXISTS (SELECT * FROM dbo.sysobjects WHERE id = OBJECT_ID(N'[Last_In_Group]') AND xtype in (N'FN', N'IF', N'TF'))
drop function Last_In_Group
go
create function Last_In_Group(#pStart int, #pDOB date, #pLastInitial varchar(1))
returns int
AS
BEGIN
DECLARE #Dup_Tab TABLE(
Person_ID int,
DOB date,
LastInitial varchar(1)
)
-- populate the table with some data
insert #Dup_Tab values
(1, '31.12.2013', 'P'), (2, '24.12.2013', 'C'), (3, '24.12.2013', 'C')
declare #vResult int = 0,
#cDOB date,
#cLastInitial varchar(1)
if(#vResult = #pStart) return #pStart
set #vResult = #pStart
-- loop over your data with a fast cursor
declare cur cursor FAST_FORWARD for select DOB, LastInitial from #Dup_Tab
open cur
fetch next from cur
into #cDOB, #cLastInitial
while ##FETCH_STATUS = 0
begin
if(#cDOB = #pDOB and #cLastInitial = #pLastInitial)
set #vResult += 1
fetch next from cur
into #cDOB, #cLastInitial
end
return #vResult;
end
go
print dbo.Last_In_Group(1, '24.12.2013', 'C')
go
It might not work as your sample above but I hope this gives you some hint where to go further.