Question

I am trying to get the results of the query below:

select distinct
       REQ_ID
     , ID
     , MAX(STEP) as step
     , SUBSTRING(p1.NAME,9, len(p1.NAME)) as _index_
     , p2.value as location

from HISTORY h

LEFT JOIN Parameter p1 
on     p1.WP_ID=h.ID
   AND (   (    p1.NAME like 'name_' 
            AND p1.VALUE like h.ID COLLATE DATABASE_DEFAULT ) 
        OR (    p1.NAME like 'name__' 
            AND p1.VALUE like h.ID COLLATE DATABASE_DEFAULT) )

LEFT JOIN Parameter p2 
on     p2.WP_ID=h.ID 
   AND p2.PA_NAME = 'Location' + (SUBSTRING(p1.NAME,9, len(p1.NAME)) ) 

WHERE h.ROLE = 'rock'

GROUP BY REQ_ID, ID, step, p1.name, p2.value

The problem is that the query is returning more than 1 results (in my case there are 6) and the MAX(STEP) value is not returning the max, I can see values like 0,1,3,0,1,2
Is there a way to get only the result with the max(step)? the step field is varchar(1)

Answer 1

When dealing with first record in a group, the often simplest approach is to use ROW_NUMBER(). Without knowing more about your schema and data, I can say that the following query will work, but it may be possible to simplifiy it further…

WITH
  sequenced_data AS
(
  SELECT
    REQ_ID,
    ID,
    ROW_NUMBER() OVER (PARTITION BY req_id, id ORDER BY step DESC) AS sequence_id,
    STEP,
    SUBSTRING(p1.NAME,9, len(p1.NAME)) as _index_,
    p2.value as location
  FROM
    HISTORY    h
  LEFT JOIN
    Parameter  p1
      ON  p1.WP_ID = h.ID
      AND (   (p1.NAME like 'name_' AND p1.VALUE  like h.ID COLLATE DATABASE_DEFAULT ) 
           OR (p1.NAME like 'name__' AND p1.VALUE  like h.ID COLLATE DATABASE_DEFAULT)
          )
  LEFT JOIN
    Parameter  p2
      ON  p2.WP_ID = h.ID
      AND p2.PA_NAME = 'Location' + (SUBSTRING(p1.NAME,9, len(p1.NAME)) ) 
  WHERE
    h.ROLE = 'rock'
  GROUP BY
    REQ_ID,
    ID,
    STEP,
    SUBSTRING(p1.NAME,9, len(p1.NAME)),
    p2.value
)
SELECT
  *
FROM
  sequenced_data
WHERE
  sequence_id = 1

This will give only one row per req_id, id combination, and will always pick the row with the highest step value.

NOTE: I don’t know if the GROUP BY is still needed. I left it in simply because you had it.