SQL:seleziona record distinti in un campo con i record più alti da un altro campo

SQL Server 2005+, utilizzando CTE:

WITH rows AS (
  SELECT t.id,
         t.staff_id,
         t.skill_id,
         t.mainskill,
         ROW_NUMBER() OVER (PARTITION BY t.staff_id ORDER BY t.mainskill DESC) AS rank
    FROM TABLE t)
  SELECT r.id,
         r.staff_id,
         r.skill_id,
         r.mainskill
    FROM rows r
   WHERE r.rank = 1
ORDER BY r.staff_id

SQL Server 2005+, equivalente non CTE:

  SELECT r.id,
         r.staff_id,
         r.skill_id,
         r.mainskill
    FROM (SELECT t.id,
                 t.staff_id,
                 t.skill_id,
                 t.mainskill,
                 ROW_NUMBER() OVER (PARTITION BY t.staff_id ORDER BY t.mainskill DESC) AS rank
            FROM TABLE t) r
   WHERE r.rank = 1
ORDER BY r.staff_id

Entrambi utilizzano ROW_NUMBER , disponibile solo da SQL Server 2005.