Concatenate Multiple Row Values into 1 Row, with SQL for iSeries

Question

First, I need to thank Kent Milligan and his article at http://www.mcpressonline.com/sql/techtip-combining-multiple-row-values-into-a-single-row-with-sql-in-db2-for-i.html for getting me as far in this problem as I have. But now I need to expand on what he has done here.

To avoid you having to go to his article, the problem he addressed was concatenating string data from multiple rows into a single row in the resulting table. For example:

Table Cars:

• Make Model
• Ford Fusion
• Chevy Tahoe
• Honda Odyssey
• Ford Taurus
• Ford Focus
• Chevy Malibu

Results:

• Make Model
• Chevy Malibu, Tahoe
• Ford Focus, Taurus, Fusion
• Honda Odyssey

This was done with the SQL statement:

WITH numbered_sets(make, model, curr, prev) AS (
   SELECT make, model,
       ROW_NUMBER() OVER (PARTITION BY make ORDER BY model) AS curr,
       ROW_NUMBER() OVER (PARTITION BY make ORDER BY model) -1 AS prev
   FROM inventory)
SELECT make,
       MAX (TRIM(L ',' FROM
             CAST(SYS_CONNECT_BY_PATH(model, ',') AS VARCHAR(256)) ))
FROM numbered_sets
START WITH curr = 1
CONNECT BY make = PRIOR make AND prev = PRIOR curr
GROUP BY make

I was able to adapt that to my own table, and get most of the way where I wanted to get. But for my purposes, I have an additional column I need to include for the grouping. For example:

Table Cars:

• Make Type Model
• Ford Sedan Fusion
• Chevy SUV Tahoe
• Honda Minivan Odyssey
• Ford Sedan Taurus
• Ford Sedan Focus
• Chevy Sedan Malibu
• Ford SUV Escape
• Ford SUV Explorer
• Chevy Sedan Impala

For the Results, I’d be looking for:

• Make Type Model
• Chevy Sedan Malibu, Impala
• Chevy SUV Tahoe
• Ford Sedan Fusion, Taurus, Focus
• Ford SUV Escape, Explorer
• Honda Minivan Odyssey

Does anyone have any thoughts on what all I need to add to the original statement to be able to add the TYPE column, and GROUP on that accordingly? I’ve tried a handful of things, but I suspect I need to do something with the CONNECT_BY_PATH statement, I’m just not sure what.

Thank you

Answer 1

I think you just need to integrate type at the correct points thoughtout the query.

without being able to test, I think this would be close; but I may have missed something...

WITH numbered_sets(make, type, model, curr, prev) AS (
   SELECT make, type, model,
       ROW_NUMBER() OVER (PARTITION BY make, Type ORDER BY Make, Type, model) AS curr,
       ROW_NUMBER() OVER (PARTITION BY make, type ORDER BY Make, type, model) -1 AS prev
   FROM inventory)
SELECT make, Type
       MAX (TRIM(L ',' FROM
             CAST(SYS_CONNECT_BY_PATH(model, ',') AS VARCHAR(256)) ))
FROM numbered_sets
START WITH curr = 1
CONNECT BY make = PRIOR make AND prev = PRIOR curr and type = prior type
GROUP BY make, type

Perhaps we need do change the connect by to do a concat before connect by... though I can't see why this would help yet...

CONNECT BY concat(make,type) = PRIOR concat(make,type) AND prev = PRIOR curr