Counting consecutive numbers into one row

Question

Hello everyone,
I have recently discovered that we can use Max ( Decode ()) function of Oracle to pivot the results of a table. I have executed this just fine. However, pivoting a table is just one part of the solution that I need. The pivoting function results to something like this:
01,02,03,05,06,07,08,09,10,11,12,13,14,16,17,20,21,23,25What I actually need is something like this:
1-17, 20-21,23,25I really don't know how to start solving this but so far I have the below query:
SELECT DISTINCT
     MAX(DECODE(wldw.wafernumber,'01', '01'))
      || MAX(DECODE(wldw.wafernumber,'02', ',02'))
      || MAX(DECODE(wldw.wafernumber,'03', ',03'))
      || MAX(DECODE(wldw.wafernumber,'04', ',04'))
      || MAX(DECODE(wldw.wafernumber,'05', ',05'))
      || MAX(DECODE(wldw.wafernumber,'06', ',06'))
      || MAX(DECODE(wldw.wafernumber,'07', ',07'))
      || MAX(DECODE(wldw.wafernumber,'08', ',08'))
      || MAX(DECODE(wldw.wafernumber,'09', ',09'))
      || MAX(DECODE(wldw.wafernumber,'10', ',10'))
      || MAX(DECODE(wldw.wafernumber,'11', ',11'))
      || MAX(DECODE(wldw.wafernumber,'12', ',12'))
      || MAX(DECODE(wldw.wafernumber,'13', ',13'))
      || MAX(DECODE(wldw.wafernumber,'14', ',14'))
      || MAX(DECODE(wldw.wafernumber,'15', ',15'))
      || MAX(DECODE(wldw.wafernumber,'16', ',16'))
      || MAX(DECODE(wldw.wafernumber,'17', ',17'))
      || MAX(DECODE(wldw.wafernumber,'18', ',18'))
      || MAX(DECODE(wldw.wafernumber,'19', ',19'))
      || MAX(DECODE(wldw.wafernumber,'20', ',20'))
      || MAX(DECODE(wldw.wafernumber,'21', ',21'))
      || MAX(DECODE(wldw.wafernumber,'22', ',22'))
      || MAX(DECODE(wldw.wafernumber,'23', ',23'))
      || MAX(DECODE(wldw.wafernumber,'24', ',24'))
      || MAX(DECODE(wldw.wafernumber,'25', ',25'))  AS WAFERS     
FROM a_wiplothistory wl
JOIN Container C ON (wl.containerid = c.containerid OR wl.containerid= c.splitfromid )
JOIN a_wiplotdetailshistory wld ON wl.wiplothistoryid = wld.wiplothistoryid
JOIN a_wiplotdetailswafershistory wldw ON wld.wiplotdetailshistoryid = wldw.wiplotdetailshistoryid
WHERE c.containername = :lotThanks for helping guys.
Edited by: 1001275 on May 15, 2013 6:28 PM

Answer 1

Hi,
1001275 wrote:
Hello everyone,
I have recently discovered that we can use Max ( Decode ()) function of Oracle to pivot the results of a table. I have executed this just fine. However, pivoting a table is just one part of the solution that I need...You said it!
First, you need some way of grouping consecutive rows together (1-17 in one group, 20-21 in anoter, 23 as a group all by itself, and so on).
Then you need GROUP BY to get infmation about each goup, such as the smallest and largest number in the group.
Finally, you need to combine all that information into one big string. This is actually an example of String Aggregation , rather than pivoting. The two are closely related. Pivot means you're taking 1 column on multiple rows, and putting them into multiple columns on one row. String Aggregation is taking 1 column on multple row, and concatenating all their contents into one big string column.
Here's one way to do it:
WITH     got_group_id     AS
     SELECT     wafernumber
     ,     ROW_NUMBER () OVER (ORDER BY wafernumber)
                  - wafernumber          AS group_id
     FROM     wldw
,     got_group_info     AS
     SELECT       TO_CHAR (MIN (wafernumber))
            || CASE
                     WHEN  COUNT (*) > 1
                THEN  '-' || TO_CHAR (MAX (wafernumber))
                 END             AS group_label
     ,       ROW_NUMBER () OVER (ORDER BY  MIN (wafernumber))
                         AS group_num
     FROM      got_group_id
     GROUP BY  group_id
SELECT  SUBSTR ( SYS_CONNECT_BY_PATH (group_label, ',')
            , 2
            )     AS txt
FROM    got_group_info
WHERE     CONNECT_BY_ISLEAF     = 1
START WITH     group_num      = 1
CONNECT BY     group_num     = PRIOR group_num + 1
     AND     prior sys_guid () is not null
;I find the first part of this query to be the trickiest. I used the Fixd Difference technique to assign a common group_id to consecutive rows. See {message:id=9953384} and/or {message:id=9957164} foran explantaion of the Fixed Difference technique.
Next, in sub-query got_group_info, I used aggregate functions to produce a group_label, such as '1-17', and to assign consecutive numbers to each group. This is also a little tricy, because it involves nesting an aggregate function (MIN in this case) inside an analytc function (ROW_NUMBER).
Finally, I used SYS_CONNECT_BY_PATH to do the string aggregation.
Output:
TXT
1-17,20-21,23,25Whenever you have a question, please post CREATE TABLE and INSERT statements for some sample data. For example:
CREATE TABLE     wldw
(       wafernumber     NUMBER (3)     PRIMARY KEY
INSERT INTO wldw (wafernumber) VALUES ( 1);
INSERT INTO wldw (wafernumber) VALUES ( 2);
INSERT INTO wldw (wafernumber) VALUES ( 3);
INSERT INTO wldw (wafernumber) VALUES ( 4);
INSERT INTO wldw (wafernumber) VALUES ( 5);
INSERT INTO wldw (wafernumber) VALUES ( 6);
INSERT INTO wldw (wafernumber) VALUES ( 7);
INSERT INTO wldw (wafernumber) VALUES ( 8);
INSERT INTO wldw (wafernumber) VALUES ( 9);
INSERT INTO wldw (wafernumber) VALUES (10);
INSERT INTO wldw (wafernumber) VALUES (11);
INSERT INTO wldw (wafernumber) VALUES (12);
INSERT INTO wldw (wafernumber) VALUES (13);
INSERT INTO wldw (wafernumber) VALUES (14);
INSERT INTO wldw (wafernumber) VALUES (15);
INSERT INTO wldw (wafernumber) VALUES (16);
INSERT INTO wldw (wafernumber) VALUES (17);
INSERT INTO wldw (wafernumber) VALUES (20);
INSERT INTO wldw (wafernumber) VALUES (21);
INSERT INTO wldw (wafernumber) VALUES (23);
INSERT INTO wldw (wafernumber) VALUES (25);I realize that your table (and your query) are a lot more complicated, but it looks like you can show the part you don't already understand using just this one table with this one column.
Also, whenever you have a question, say which version oif Oracle you'e using (e.g., 11.2.0.2.0).
The query above should work in Oracle 10.1 and up. I got the wong results in the main query in Oracle 10.2, however. (Oracle 10.2 has a lot of bugs related to CONNECT BY.) It worked fine in version 11.1.
If you're using Oracle 11.2, you'll want to use LISTAGG, not SYS_CONNECT_BY_PATH, to do the string aggregation.
For more about string aggregation in various versions of Oracle, see this Oracle Base page.