Help with LAG () OVER () and dates

Question

I have the following query right now
  SELECT "DATE",
          "CELL_SITE",
          "LASTV_ATTCNT",
          "LASTV_ATTCNT2",
          "LASTV_BLKCNT",
          "LASTV_DRPCNT",
          "V_ATT_CNT",
          "V_CUST_BLK_CNT",
          "V_DRP_CALL_CNT",
          "LASTD_ATTCNT",
          "LASTD_BLKCNT",
          "LASTD_DRPCNT",
          "D_ATT_CNT",
          "D_CUST_BLK_CNT",
          "D_DRP_CALL_CNT"         
     FROM (  SELECT DATE,
                   CELL_SITE,
                    LAG (SUM (V_ATT_CNT), 24)
                       OVER (PARTITION BY BSM_NM ORDER BY DATE)
                       AS "LASTV_ATTCNT",
                    LAG (SUM (V_CUST_BLK_CNT), 24)
                       OVER (PARTITION BY BSM_NM ORDER BY DATE)
                       AS "LASTV_BLKCNT",
                    LAG (SUM (V_DRP_CALL_CNT), 24)
                       OVER (PARTITION BY BSM_NM ORDER BY DATE)
                       AS "LASTV_DRPCNT",
                    LAG (SUM (D_ATT_CNT), 24)
                       OVER (PARTITION BY BSM_NM ORDER BY DATE)
                       AS "LASTD_ATTCNT",
                    LAG (SUM (D_CUST_BLK_CNT), 24)
                       OVER (PARTITION BY BSM_NM ORDER BY DATE)
                       AS "LASTD_BLKCNT",
                    LAG (SUM (D_DRP_CALL_CNT), 24)
                       OVER (PARTITION BY BSM_NM ORDER BY DATE)
                       AS "LASTD_DRPCNT",
                    SUM (V_ATT_CNT) AS "V_ATT_CNT",
                    SUM (V_CUST_BLK_CNT) AS "V_CUST_BLK_CNT",
                    SUM (V_DRP_CALL_CNT) AS "V_DRP_CALL_CNT",
                    SUM (D_ATT_CNT) AS "D_ATT_CNT",
                    SUM (D_CUST_BLK_CNT) AS "D_CUST_BLK_CNT",
                    SUM (D_DRP_CALL_CNT) AS "D_DRP_CALL_CNT"
               FROM DMSN.DS3R_FH_1XRTT_FA_LVL_KPI
              WHERE DATE >
                         (SELECT MAX (DATE) FROM DMSN.DS3R_FH_1XRTT_FA_LVL_KPI)
                       - 2
           GROUP BY DATE, CELL_SITE)
    WHERE DATE >=
             (SELECT MAX (DATE) - NUMTODSINTERVAL (12, 'HOUR')
                FROM DMSN.DS3R_FH_1XRTT_FA_LVL_KPI)What I've noticed is that the LAG function is doing kind of what I want but not exactly. Lets say I have data for hours 12AM, 1AM, 2AM, 3AM, 4AM, 5AM, 6AM, 7AM, 8AM, 9AM, 10AM, 11AM, 12PM, and that the current hour is 12PM, the LAG function goes back to 12AM, which is what it should be doing.
but lets say I have data for 12AM, 1AM, 2AM, 3AM, 4AM, 5AM, 7AM, 8AM, 9AM, 10AM, 11AM, 12PM, but am MISSING 6AM (no data for 6AM). The lag function will now go back to 11PM instead of 12AM since the 6AM hour is missing.
if I have data for 12AM, 1AM, 2AM, 3AM, 4AM, 7AM, 8AM, 9AM, 10AM, 11AM, 12PM and am MISSING 5AM and 6AM, the data will go back to 10PM then and not just 12AM.
Can I prevent this from happening and always just go back 12 hours?
Edited by: k1ng87 on Apr 25, 2013 1:27 PM

Answer 1

Hi,
k1ng87 wrote:
LAG (SUM (V_ATT_CNT), 24)
OVER (PARTITION BY BSM_NM ORDER BY DATE)
AS "LASTV_ATTCNT",
LAG (SUM (V_CUST_BLK_CNT), 24)
OVER (PARTITION BY BSM_NM ORDER BY DATE)
AS "LASTV_BLKCNT",
LAG (SUM (V_DRP_CALL_CNT), 24)
OVER (PARTITION BY BSM_NM ORDER BY DATE)
AS "LASTV_DRPCNT",
Since you want so many columns from the same previous row, it would probably be simpler and more efficient to do a self-join, rather than so many LAG functions.
What I've noticed is that the LAG function is doing kind of what I want but not exactly. Lets say I have data for hours 12AM, 1AM, 2AM, 3AM, 4AM, 5AM, 6AM, 7AM, 8AM, 9AM, 10AM, 11AM, 12PM, and that the current hour is 12PM, the LAG function goes back to 12AM, which is what it should be doing.You're always passing 24 as the 2nd argument to LAG; that means "Return the value from the 24th precding row." If there's a row for every hour, and the current row is 12PM, then wouldn't 24 rows back would be 12PM yesterday, not 12AM? Maybe you meant to pass 12, not 24, as the second argument to LAG. I'll assume you really want the 24th row rom now on, but the technique is the same whether that number is 24 or 12.
but lets say I have data for 12AM, 1AM, 2AM, 3AM, 4AM, 5AM, 7AM, 8AM, 9AM, 10AM, 11AM, 12PM, but am MISSING 6AM (no data for 6AM). The lag function will now go back to 11PM instead of 12AM since the 6AM hour is missing.
if I have data for 12AM, 1AM, 2AM, 3AM, 4AM, 7AM, 8AM, 9AM, 10AM, 11AM, 12PM and am MISSING 5AM and 6AM, the data will go back to 10PM then and not just 12AM.
Can I prevent this from happening and always just go back 12 hours?A self-join would handle this automatically: it will look for the matching row, without regard for which other rows might be present or absent.
If you really had to do this with analytic functions, you could use MIN or FIRST_VALUE with a range window:
RANGE BETWEEN  1 + (.5 / 24)  PRECEDING
      AND      1 - (.5 / 24)  PRECEDINGWhen you ORDER BY a DATE column (don't call it "DATE"), the units are 1 day, so 1+(.5/24) days ago is 24.5 hours ago; that is, the window only includes DATES between 24.5 and 23.5 hours ago.
If, for some reason, you really wanted to call LAG, then you could do an outer join to guarantee that a row was present for each of the preceding 24 hours.
I hope this answers your question.
If not, post a little sample data (CREATE TABLE and INSERT statements, relevant columns only), and also post the results you want from that data.
Explain, using specific examples, how you get those results from that data.
Always say which version of Oracle you're using (e.g., 11.2.0.2.0).
See the forum FAQ {message:id=9360002}