Selecting rows based on monthly history

Question

Hi All
I have a table that looks like -
CREATE TABLE orders(store NUMBER, mnth VARCHAR2(12), orders NUMBER);
store     mnth     orders
243272     May-10     17
226798     May-10     20
631638     May-10     14
225779     May-10     10
226675     May-10     9
245593     May-10     13
534299     May-10     9
534440     May-10     12
539313     May-10     14
510617     May-10     11
248196     May-10     9
250472     May-10     9
235854     May-10     16
232899     May-10     12
233840     Apr-10     13
538425     Apr-10     11
338415     Apr-10     12
534336     Apr-10     15
536788     Apr-10     11
441471     Apr-10     11
509317     Apr-10     10
612752     Apr-10     9
338837     Apr-10     11
337869     Apr-10     11
335505     Apr-10     10
333358     Mar-10     11
315854     Mar-10     15
272751     Mar-10     18
270158     Mar-10     10
267979     Mar-10     13
248292     Mar-10     19
244136     Mar-10     14
231837     Mar-10     18
226894     Mar-10     53
The table has records of the number of orders per month for a number of stores, each store has a unique 6 digit id number. Each store can appear multipe times in the report, as they all order a number of times per month.
What I need to do, is apply a number of rules to create four reports -
If a store has more than 8 orders in the most recent month AND none in the next most recent month, include it in a report (eg 9 in March but none in February)
If a store has more than 8 orders for the most recent month AND more than 8 orders in either of the next two most recent months, include it in a report (eg 9+ in March AND 9+ in Jan OR Feb)
If a store has more than 8 orders for the most recent month AND more than 8 orders in either of the next two most recent months AND more than 8 orders in any of the previous 3 months, include it in a report. (eg 9+ in March AND 9+ in Jan or Feb AND 9+ in Oct. OR Nov. OR Dec.)
If a store has more than 8 orders per month, for the 3 most recent months, include it in a report (eg 9+ in Jan, 9+ in Feb, 9+ in March)
I would need to create 4 tables, one for each report, and the only column I would require in each table would be the store code.
I hope this makes sense, if you need any further info please message me. I am using Application Express, something as a pl/sql example woud be great as I am fairly new to all this and am trying to pick it up as I go along..
Thanks!

Answer 1

Hi,
billlagr wrote:
Hi All
I have a table that looks like -
CREATE TABLE orders(store NUMBER, mnth VARCHAR2(12), orders NUMBER);Thanks for posting the CREATE TABLE statement; that's very helpful.
store     mnth     orders
243272     May-10     17
226798     May-10     20
631638     May-10     14
...It would also be helpful if you posted the data as INSERT statements, for example:
INSERT INTO orders (store, mnth, orders) VALUES (243272, 'May-10', 17);Also, post the results you want from that data.
Data that represents a point in time, such as your mnth column, should be stored in a DATE column.
The table has records of the number of orders per month for a number of stores, each store has a unique 6 digit id number. Each store can appear multipe times in the report, as they all order a number of times per month.
What I need to do, is apply a number of rules to create four reports -
If a store has more than 8 orders in the most recent month AND none in the next most recent month, include it in a report (eg 9 in March but none in February)What is "the most recent month"? I assume you mean the last full month before the report is run. For example, if you create the report on June 2, 2010, then the "most recent month" is may, 2010.
Here's one way to do the first report:
WITH     all_months     AS
     SELECT     LEVEL               AS mnth_num
     ,     ADD_MONTHS ( TRUNC (SYSDATE, 'MONTH')
                  , -LEVEL
                  )          AS mnth_begin_date
     FROM     dual
     CONNECT BY     LEVEL     <= 2
SELECT       o.store
,       NVL (SUM (CASE WHEN m.mnth_num = 1 THEN o.orders END), 0)     AS mnth_1_total
,       NVL (SUM (CASE WHEN m.mnth_num = 2 THEN o.orders END), 0)     AS mnth_2_total
FROM       orders     o
JOIN       all_months     m     ON     TO_DATE (o.mnth, 'Mon-RR')     = mnth_begin_date
GROUP BY  o.store
HAVING       NVL (SUM (CASE WHEN m.mnth_num = 1 THEN o.orders END), 0)     > 8
AND       NVL (SUM (CASE WHEN m.mnth_num = 2 THEN o.orders END), 0)     = 0
ORDER BY  o.store
I would need to create 4 tables, one for each report, and the only column I would require in each table would be the store code.You don't need to create any tables to get the results you want, but if you really want to make tables out of the result sets, use CREATE TABLE AS.
Unfortunately, you can't use a column alias (like mnth_2_total) in the GROUP BY clause of a query if it was defined in the SELECT clause of the same query. That's why I repeated expressions such as <tt>NVL (SUM (CASE WHEN m.mnth_num = 2 THEN o.orders END), 0)</tt>. You may want to define the aliases in a sub-query, so you can use them in a su[per-query, like this:
{code}
WITH     all_months     AS
     SELECT     LEVEL               AS mnth_num
     ,     ADD_MONTHS ( TRUNC (SYSDATE, 'MONTH')
               , -LEVEL
               )          AS mnth_begin_date
     FROM     dual
     CONNECT BY     LEVEL     <= 2
,     got_aliases     AS
     SELECT     o.store
     ,     NVL (SUM (CASE WHEN m.mnth_num = 1 THEN o.orders END), 0)     AS mnth_1_total
     ,     NVL (SUM (CASE WHEN m.mnth_num = 2 THEN o.orders END), 0)     AS mnth_2_total
     FROM     orders     o
     JOIN     all_months     m     ON     TO_DATE (o.mnth, 'Mon-RR')     = mnth_begin_date
     GROUP BY o.store
SELECT     store
,     mnth_1_total
,     mnth_2_total
FROM     got_aliases
WHERE     mnth_1_total     > 8
AND     mnth_2_total     = 0
ORDER BY store
You may want to create a view that is like got_alaises, but goes back 6 months, and base all four reports on that view.
Edited by: Frank Kulash on Jun 2, 2010 6:14 AM
Added version with got_aliases