Question

I have a table containing some names and their associated ID, along with a snapshot:

snapshot, systemid, name[, some, other, columns]

I need to identify all the unique names that a systemid has had across all snapshots, but only where there has been at least once change.

For example, with the data:

'DR1', 0, 'MOUSE_SPEED'
'DR1', 1, 'MOUSE_POS'
'DV8', 0, 'MOUSE_BUTTONS'
'DV8', 1, 'MOUSE_POS'
'DR6', 0, 'MOUSE_BUTTONS'
'DR6', 1, 'MOUSE_POS'
'PP2', 0, 'MOUSE_SPEED'
'PP2', 1, 'MOUSE_POS'

…I’d like a query that will return (in any order):

0, 'MOUSE_SPEED'
0, 'MOUSE_BUTTONS'

Additionally, it would be useful to have the inverse – a list of systemids that have remained stable across all snapshots (that is, where the name has never changed).

I am using PostgreSQL v8.4.2.

EDIT: Updated to reflect comments (sorry for the original less-than-perfect post, I am new here!).

Answer 1

Following is with SQL Server but it does not use any SQL Server specific constructs. It should be portable to postgresql.

SQL Statement

SELECT  DISTINCT t1.id, t1.name
FROM    @Table t1
        INNER JOIN (
          SELECT  t.id 
          FROM    (
                    SELECT  DISTINCT id, name
                    FROM    @Table
                  ) t
          GROUP BY t.id 
          HAVING COUNT(*) > 1
        ) t2 ON t2.id = t1.id

Test data

DECLARE @Table TABLE (snapshot INTEGER, id INTEGER, name VARCHAR(32))

INSERT INTO @TABLE
SELECT 1, 0, 'MOUSE_SPEED'
UNION ALL SELECT 1, 1, 'MOUSE_POS'
UNION ALL SELECT 1, 2, 'KEYBOARD_STATE'
UNION ALL SELECT 2, 0, 'MOUSE_BUTTONS'
UNION ALL SELECT 2, 1, 'MOUSE_POS'
UNION ALL SELECT 2, 2, 'KEYBOARD_STATE'
UNION ALL SELECT 3, 0, 'MOUSE_SPEED'
UNION ALL SELECT 3, 1, 'MOUSE_POS'
UNION ALL SELECT 3, 2, 'KEYBOARD_STATE'