[8i] Best way to replace "similar" values with one value...?

Question

Below is some sample data:
Sorry for the large set / complex set of sample data, but there should be no need to go through it in detail =).
2 things to keep in mind:
1) This is an 8i db
2) I cannot create any new tables within this db, though I can create views
CREATE TABLE     ord_master
(     part_nbr          CHAR(25)
,     ord_nbr          CHAR(10)
,     sub_ord_nbr     CHAR(3)
,     ord_type          CHAR(2)
,     ord_stat          CHAR(2)
,     date_closed     DATE
-- please note, though none of the columns in ord_master are explicitly defined as NOT NULL,
-- none of them are NULL in the db, and given how things are input into the table, I don't
-- even think it's possible
-- a unique order is described by ord_nbr, sub_ord_nbr.  there should only be one record in
-- this table for a given unique combination of those columns, though they are not the p.k.
-- NOTE: dates default to 12/31/1900, so in the table above, if an order is still open (thus,
-- not closed), it's date_closed will be 12/31/1900.  It will only change once the order closes.
INSERT INTO     ord_master
VALUES     ('ABC-123','1','1','AA','CL',TO_DATE('05/01/2009','mm/dd/yyyy'));
INSERT INTO     ord_master
VALUES     ('ABC-123','1','2','AA','CL',TO_DATE('9/14/2009','mm/dd/yyyy'));
INSERT INTO     ord_master
VALUES     ('ABC-123','2','1','AA','CL',TO_DATE('12/03/2009','mm/dd/yyyy'));
INSERT INTO     ord_master
VALUES     ('ABC-123','3','1','AA','OP',TO_DATE('12/31/1900','mm/dd/yyyy'));
INSERT INTO     ord_master
VALUES     ('ABC-123','4','1','AA','CL',TO_DATE('03/12/2004','mm/dd/yyyy'));
INSERT INTO     ord_master
VALUES     ('XYZ-123','5','1','AA','CL',TO_DATE('03/12/2004','mm/dd/yyyy'));
CREATE TABLE     op_detail
(     ord_nbr          CHAR(10)
,     sub_ord_nbr     CHAR(3)
,     major_seq_nbr     CHAR(4)
,     wctr_id          CHAR(5)
,     oper_stat          CHAR(2)
,     plan_su_rsrc     NUMBER(5,2)
,     plan_run_rsrc     NUMBER(11,8)
,     plan_subc_amt     NUMBER(15,5)
,     act_su_rsrc     NUMBER(5,2)
,     act_run_rsrc     NUMBER(7,2)
,     act_subc_amt     NUMBER(15,2)
,     act_start_dt     DATE
,     dt_completed     DATE
,     qty_planned     NUMBER(13,4)
,     qty_recvd          NUMBER(13,4)
,     qty_complete     NUMBER(13,4)
,     qty_scrap          NUMBER(13,4)
,     qty_on_hold     NUMBER(13,4)
-- similar to ord_master, thought none of the columns are explicitly defined as NOT NULL,
-- none of the records have NULL values for any of these columns, and I don't think it's
-- really possible with the way the system is
INSERT INTO     op_detail
VALUES     ('1', '1', '0081', 'X516 ', 'CP', 0, 0, 0, 0, .2, 0, TO_DATE('02/03/2010', 'mm/dd/yyyy'), TO_DATE('02/03/2010', 'mm/dd/yyyy'), 3, 3, 3, 0, 0);
INSERT INTO     op_detail
VALUES     ('1', '1', '0324', 'Y000 ', 'CP', .1, .2, 0, 0, .87, 0, TO_DATE('04/06/2010', 'mm/dd/yyyy'), TO_DATE('04/06/2010', 'mm/dd/yyyy'), 3, 3, 3, 0, 0);
INSERT INTO     op_detail
VALUES     ('1', '1', '0170', 'X928 ', 'CP', 0, 0, 9.6, 0, 0, 159, TO_DATE('03/05/2010', 'mm/dd/yyyy'), TO_DATE('03/05/2010', 'mm/dd/yyyy'), 3, 3, 3, 0, 0);
INSERT INTO     op_detail
VALUES     ('1', '1', '0130', 'H211 ', 'CP', .5, .1, 0, .08, .63, 0, TO_DATE('02/10/2010', 'mm/dd/yyyy'), TO_DATE('02/10/2010', 'mm/dd/yyyy'), 3, 3, 3, 0, 0);
INSERT INTO     op_detail
VALUES     ('1', '2', '0120', 'H214 ', 'CP', .3, .05, 0, 0, .68, 0, TO_DATE('05/22/2009', 'mm/dd/yyyy'), TO_DATE('05/22/2009', 'mm/dd/yyyy'), 3, 3, 3, 0, 0);
INSERT INTO     op_detail
VALUES     ('1', '2', '0122', 'T203 ', 'CP', 2.5, 1, 0, .67, .3, 0, TO_DATE('05/28/2009', 'mm/dd/yyyy'), TO_DATE('05/28/2009', 'mm/dd/yyyy'), 3, 3, 3, 0, 0);
INSERT INTO     op_detail
VALUES     ('1', '2', '0130', 'H211 ', 'CL', .5, .1, 0, .67, .3, 0, TO_DATE('05/28/2009', 'mm/dd/yyyy'), TO_DATE('05/28/2009', 'mm/dd/yyyy'), 3, 3, 3, 0, 0);
INSERT INTO     op_detail
VALUES     ('1', '2', '0170', 'X928 ', 'CP', 0, 0, 9.6, 0, 0, 153.12, TO_DATE('06/10/2009', 'mm/dd/yyyy'), TO_DATE('06/10/2009', 'mm/dd/yyyy'), 3, 3, 3, 0, 0);
INSERT INTO     op_detail
VALUES     ('1', '2', '0324', 'Y000 ', 'CP', .1, .2, 0, 0, .08, 0, TO_DATE('06/17/2009', 'mm/dd/yyyy'), TO_DATE('06/17/2009', 'mm/dd/yyyy'), 3, 3, 3, 0, 0);
INSERT INTO     op_detail
VALUES     ('2', '1', '0081', 'X516 ', 'CL', 0, 0, 0, 0, 0, 0, TO_DATE('07/20/2009', 'mm/dd/yyyy'), TO_DATE('07/20/2009', 'mm/dd/yyyy'), 4, 4, 4, 0, 0);
INSERT INTO     op_detail
VALUES     ('2', '1', '0324', 'Y000 ', 'CL', .1, .2, 0, 0, 3.03, 0, TO_DATE('09/11/2009', 'mm/dd/yyyy'), TO_DATE('09/11/2009', 'mm/dd/yyyy'), 4, 4, 4, 0, 0);
INSERT INTO     op_detail
VALUES     ('2', '1', '0170', 'X928 ', 'CP', 0, 0, 9.6, 0, 0, 204.16, TO_DATE('08/11/2009', 'mm/dd/yyyy'), TO_DATE('08/11/2009', 'mm/dd/yyyy'), 4, 4, 4, 0, 0);
INSERT INTO     op_detail
VALUES     ('2', '1', '0130', 'H211 ','CP', .5, .1, 0, 1.97, 2.67, 0, TO_DATE('07/28/2009', 'mm/dd/yyyy'), TO_DATE('07/29/2009', 'mm/dd/yyyy'), 4, 4, 4, 0, 0);
INSERT INTO     op_detail
VALUES     ('3', '1', '0170', 'X928 ', 'CP', 0, 0, 9.6, 0, 0, 255.2, TO_DATE('11/06/2009', 'mm/dd/yyyy'), TO_DATE('11/06/2009', 'mm/dd/yyyy'), 5, 5, 5, 0, 0);
INSERT INTO     op_detail
VALUES     ('3', '1', '0130', 'H212 ', 'CP', .5, .1, 0, .6, .03, 0, TO_DATE('10/21/2009', 'mm/dd/yyyy'), TO_DATE('10/21/2009', 'mm/dd/yyyy'), 5, 5, 5, 0, 0);
INSERT INTO     op_detail
VALUES     ('3', '1', '0550', 'R501 ', 'WK', 1, 16, 0, 2.2, 12.4, 0, TO_DATE('05/19/2010', 'mm/dd/yyyy'), TO_DATE('12/31/1900', 'mm/dd/yyyy'), 5, 5, 5, 0, 0);
INSERT INTO     op_detail
VALUES     ('3', '1', '0324', 'Y000 ', 'CL', .1, .2, 0, 0, .07, 0, TO_DATE('02/04/2010', 'mm/dd/yyyy'), TO_DATE('02/04/2010', 'mm/dd/yyyy'), 5, 5, 5, 0, 0);
INSERT INTO     op_detail
VALUES     ('4', '1', '0324', 'Y000 ', 'CP', .1, .2, 0, 0, .59, 0, TO_DATE('02/06/2004', 'mm/dd/yyyy'), TO_DATE('02/06/2004', 'mm/dd/yyyy'), 4, 4, 4, 0, 0);This is the query I'm working with right now:
SELECT     o.part_nbr               
,          s.major_seq_nbr          AS op_nbr
,          CASE          -- CASE statement to replace similar-to wctr_id's with a single new wctr_id, so we can treat them as the same
               WHEN     s.wctr_id     IN ('H211','H212')
               THEN     'Z211'
               ELSE     s.wctr_id
          END                         AS wctr
,          MIN(s.plan_su_rsrc)          AS min_plan_setup
,          AVG(s.plan_su_rsrc)          AS avg_plan_setup
,          MAX(s.plan_su_rsrc)          AS max_plan_setup
,          STDDEV(s.plan_su_rsrc)     AS sdev_plan_setup
,          MIN(s.plan_run_rsrc)     AS min_plan_run
,          AVG(s.plan_run_rsrc)     AS avg_plan_run
,          MAX(s.plan_run_rsrc)     AS max_plan_run
,          STDDEV(s.plan_run_rsrc)     AS sdev_plan_run
,          MIN(s.plan_subc_amt)     AS min_plan_subc
,          AVG(s.plan_subc_amt)     AS avg_plan_subc
,          MAX(s.plan_subc_amt)     AS max_plan_subc
,          STDDEV(s.plan_subc_amt)     AS sdev_plan_subc
,          MIN(s.act_su_rsrc)          AS min_act_setup
,          AVG(s.act_su_rsrc)          AS avg_act_setup
,          MAX(s.act_su_rsrc)          AS max_act_setup
,          STDDEV(s.act_su_rsrc)     AS sdev_act_setup
,          MIN(s.act_run_rsrc)          AS min_act_run
,          AVG(s.act_run_rsrc)          AS avg_act_run
,          MAX(s.act_run_rsrc)          AS max_act_run
,          STDDEV(s.act_run_rsrc)     AS sdev_act_run
,          MIN(s.act_subc_amt)          AS min_act_subc
,          AVG(s.act_subc_amt)          AS avg_act_subc
,          MAX(s.act_subc_amt)          AS max_act_subc
,          STDDEV(s.act_subc_amt)     AS sdev_act_subc     
,          MIN(s.act_start_dt)          AS min_start
,          MAX(s.act_start_dt)          AS max_start
,          MIN(s.dt_completed)          AS min_comp
,          MAX(s.dt_completed)          AS max_comp
,          MIN(s.qty_planned)          AS min_qty_plan
,          AVG(s.qty_planned)          AS avg_qty_plan
,          MAX(s.qty_planned)          AS max_qty_plan
,          STDDEV(s.qty_planned)     AS sdev_qty_plan
,          MIN(s.qty_recvd)          AS min_qty_recvd
,          AVG(s.qty_recvd)          AS avg_qty_recvd
,          MAX(s.qty_recvd)          AS max_qty_recvd
,          STDDEV(s.qty_recvd)          AS sdev_qty_recvd
,          MIN(s.qty_complete)          AS min_qty_comp
,          AVG(s.qty_complete)          AS avg_qty_comp
,          MAX(s.qty_complete)          AS max_qty_comp
,          STDDEV(s.qty_complete)     AS sdev_qty_comp
,          MIN(s.qty_scrap)          AS min_qty_scrap
,          AVG(s.qty_scrap)          AS avg_qty_scrap
,          MAX(s.qty_scrap)          AS max_qty_scrap
,          STDDEV(s.qty_scrap)          AS sdev_qty_scrap
,          MIN(s.qty_on_hold)          AS min_qty_hold
,          AVG(s.qty_on_hold)          AS avg_qty_hold
,          MAX(s.qty_on_hold)          AS max_qty_hold
,          STDDEV(s.qty_on_hold)     AS sdev_qty_hold
,          COUNT(*)                    AS rec_cnt
,          CASE
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-12)
               THEN     '12 mos. back'
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-24)
               THEN     '24 mos. back'
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-36)
               THEN     '36 mos. back'
               ELSE '37 mos. back +'
          END                         AS time_period
FROM          ord_master o
,          op_detail     s
,          (     -- begin in-line view a, to get the min act_start_dt for each order
          SELECT     s1.ord_nbr
          ,          s1.sub_ord_nbr
          ,          MIN     (
                         CASE
                              WHEN     s1.act_start_dt     > TO_DATE('01/01/2000','mm/dd/yyyy')
                              THEN     s1.act_start_dt
                         END
                         )          AS ord_min_start
          FROM          op_detail s1
          GROUP BY     s1.ord_nbr
          ,          s1.sub_ord_nbr
          ) a     -- end in-line view a          
WHERE     o.ord_nbr          = s.ord_nbr
AND          o.sub_ord_nbr     = s.sub_ord_nbr
AND          o.ord_nbr          = a.ord_nbr
AND          o.sub_ord_nbr     = a.sub_ord_nbr
AND          o.ord_type     = 'AA'
AND          o.ord_stat     IN ('CL','OP')
AND          s.oper_stat     IN ('CL','CP')
GROUP BY     o.part_nbr
,          s.major_seq_nbr
,          CASE          -- CASE statement to replace similar-to wctr_id's with a single new wctr_id, so we can treat them as the same
               WHEN     s.wctr_id     IN ('H211','H212')
               THEN     'Z211'
               ELSE     s.wctr_id
          END
,          CASE
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-12)
               THEN     '12 mos. back'
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-24)
               THEN     '24 mos. back'
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-36)
               THEN     '36 mos. back'
               ELSE '37 mos. back +'
          END
ORDER BY     o.part_nbr
,          CASE
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-12)
               THEN     '12 mos. back'
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-24)
               THEN     '24 mos. back'
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-36)
               THEN     '36 mos. back'
               ELSE '37 mos. back +'
          END
,          s.major_seq_nbrThis gives me a bunch of descriptive statistics on some data stored in the database, broken into groups by the time period.
I have 7 sets of possible values for the column wctr_id that are groups of similar/same work centers, and should be treated as the same, assuming the ord_nbr, sub_ord_nbr, and major_seq_nbr are the same. In my example case statement above, anytime I find 'H211' or 'H212', I am replacing it with 'Z211', which is my new, unique identifier for any wctr in the group ('H211', 'H212'). I have 7 of these sorts of groups that need to be replaced by a new, unique identifier. Is my CASE statement (noted above) the best way to do this, or is there a better way? These groups of similar-to wctr_id's are not listed in a table anywhere.

Answer 1

Hi,
As always, thanks for posting the CREATE TABLE and INSERT statements; that's very helpful, and leads to faster, better solutions.
Another thing that is helpful, and promotes faster, better solutions, is to simplify your problem.
For example, you have to compute a lot of aggregates, but it looks like you already know how to do that. This problem only involves the CASE expressions, and how they interact with the GROUP BY clause. So just cut out most of the aggregate functions.
Also, it looks like the intricacies of in-line view a have nothing to do with your current problem. For purposes of posting this problem, pretend that a is a regular view, so it doesn't clutter up the main query.
I think your problem could be restated like this:
You already have a query like this:
CREATE OR REPLACE VIEW a
AS
          SELECT     s1.ord_nbr
          ,          s1.sub_ord_nbr
          ,          MIN     (
                         CASE
                              WHEN     s1.act_start_dt     > TO_DATE('01/01/2000','mm/dd/yyyy')
                              THEN     s1.act_start_dt
                         END
                         )          AS ord_min_start
          FROM          op_detail s1
          GROUP BY     s1.ord_nbr
          ,          s1.sub_ord_nbr
SELECT     o.part_nbr               
,          s.major_seq_nbr          AS op_nbr
,          CASE          -- CASE statement to replace similar-to wctr_id's with a single new wctr_id, so we can treat them as the same
               WHEN     s.wctr_id     IN ('H211','H212')
               THEN     'Z211'
               ELSE     s.wctr_id
          END                         AS wctr
,          MIN(s.plan_su_rsrc)          AS min_plan_setup
          -- lots of other agtgregates go here in real query
,          CASE
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-12)
               THEN     '12 mos. back'
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-24)
               THEN     '24 mos. back'
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-36)
               THEN     '36 mos. back'
               ELSE '37 mos. back +'
          END                         AS time_period
FROM          ord_master o
,          op_detail     s
,          a          -- in real query, a is an in-line view
WHERE     o.ord_nbr          = s.ord_nbr
AND          o.sub_ord_nbr     = s.sub_ord_nbr
AND          o.ord_nbr          = a.ord_nbr
AND          o.sub_ord_nbr     = a.sub_ord_nbr
AND          o.ord_type     = 'AA'
AND          o.ord_stat     IN ('CL','OP')
AND          s.oper_stat     IN ('CL','CP')
GROUP BY     o.part_nbr
,          s.major_seq_nbr
,          CASE          -- CASE statement to replace similar-to wctr_id's with a single new wctr_id, so we can treat them as the same
               WHEN     s.wctr_id     IN ('H211','H212')
               THEN     'Z211'
               ELSE     s.wctr_id
          END
,          CASE
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-12)
               THEN     '12 mos. back'
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-24)
               THEN     '24 mos. back'
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-36)
               THEN     '36 mos. back'
               ELSE '37 mos. back +'
          END
ORDER BY     o.part_nbr
,          CASE
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-12)
               THEN     '12 mos. back'
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-24)
               THEN     '24 mos. back'
               WHEN a.ord_min_start     >     ADD_MONTHS(SYSDATE,-36)
               THEN     '36 mos. back'
               ELSE '37 mos. back +'
          END
,          s.major_seq_nbr
;that produces this output:
`                       MIN_
                       PLAN_
PART_NBR   OP_N WCTR   SETUP TIME_PERIOD
ABC-123    0081 X516      .0 12 mos. back
ABC-123    0130 Z211      .5 12 mos. back
ABC-123    0170 X928      .0 12 mos. back
ABC-123    0324 Y000      .1 12 mos. back
ABC-123    0120 H214      .3 24 mos. back
ABC-123    0122 T203     2.5 24 mos. back
ABC-123    0130 Z211      .5 24 mos. back
ABC-123    0170 X928      .0 24 mos. back
ABC-123    0324 Y000      .1 24 mos. back
ABC-123    0324 Y000      .1 37 mos. back +The output is correct; the problem is, can we get the same results without 2 copies of the wctr CASE expression, and 3 copies of the time_period CASE expression?
Is that a fair description of the problem?
For starters, you can use column aliases in the ORDER BY clause. Unfortunately, that's the only place where you can use column aliases in the same query where they are defined . You can compute them in a sib-query, and then use them in super-queries.
In other words, wouldn't it be nice if we had a table just like op_detail, but with an additional column called wctr? Make a view: either a regular view (especially if wctr, computed the same way, will be used in several different queries) or an in-line view, as shown below.
Wouldn't it be great if there was a table that was just like a, except that it had time_period alreay computed? Make a view: either a regular view or an in-line view, as shown below.
SELECT     o.part_nbr
,     s.major_seq_nbr          AS op_nbr
,     s.wctr
,     MIN(s.plan_su_rsrc)          AS min_plan_setup
     -- lots of other agtgregates go here in real query
,     a2.time_period
FROM     ord_master      o
,     (          -- Begin in-line view s, with data from op_detail
             SELECT  op_detail.*      -- or list all columns needed in main query
          ,     CASE          -- replace similar-to wctr_id's with a single new wctr_id, so we can treat them as the same
                   WHEN     wctr_id        IN ('H211','H212')
                   THEN     'Z211'
                   ELSE     wctr_id
               END                    AS wctr
          FROM     op_detail
     ) s          -- End in-line view s, with data from op_detail
,     (           -- Begin in-line view a2, to compute time_period
          SELECT  a.*
          ,     CASE
                   WHEN ord_min_start     >     ADD_MONTHS(SYSDATE,-12)
                       THEN     '12 mos. back'
                       WHEN ord_min_start     >     ADD_MONTHS(SYSDATE,-24)
                       THEN     '24 mos. back'
                       WHEN ord_min_start     >     ADD_MONTHS(SYSDATE,-36)
                       THEN     '36 mos. back'
                       ELSE '37 mos. back +'
               END                         AS time_period
          FROM     a          -- in real query, a is an in-line view
     ) a2          -- End in-line view a2, to compute time_period
WHERE     o.ord_nbr     = s.ord_nbr
AND     o.sub_ord_nbr     = s.sub_ord_nbr
AND     o.ord_nbr     = a2.ord_nbr
AND     o.sub_ord_nbr     = a2.sub_ord_nbr
AND     o.ord_type     = 'AA'
AND     o.ord_stat     IN ('CL','OP')
AND     s.oper_stat     IN ('CL','CP')
GROUP BY     o.part_nbr
,          s.major_seq_nbr
,          s.wctr
,          a2.time_period
ORDER BY     o.part_nbr
,          a2.time_period
,          s.major_seq_nbr
;In your example, wctr is the same as wctr_id except for two values. If it was more complicated (e.g., if there were not just 2 values that need to be mapped to 'Z211', but also 3 values to were equivalent to 'Z987', and 1 that was to be treated like 'A012', and 5 that should appear as 'A666', and ...), then you could code all that in a humongeous CASE expression (a nested CASE expression, if it involves more that 128 branches), but it really belongs in a table, where you have the raw wctr_id (like 'H211') in one column, and its corresponding equivalent (such as 'Z211') in another. I know you said you can't create new tables; I'm just pointing this out so you can tell the people who won't let you create new tables that it's a bad idea, likely to cause errors and certain to waste time in the future.