4

I have a database containing time stamped data values, with multiple samples per day, i.e.

05/01/11 01:00:00 - 1.23  
05/01/11 01:12:34 - 0.99  
....  
05/01/11 23:59:59 - 2.34  
05/02/11 00:11:22 - 4.56

etc

I am trying to get a list of the first sample of each month. I have managed to get a list of every sample on the 1st of each month:

SELECT
    "RecordTime", "FormattedValue"
FROM
    CDBHistoric
WHERE
    "Id" = 12345 AND EXTRACT (DAY FROM "RecordTime") = 1

This gives data like

03/01/11 00:00:01 - 1.23  
03/01/11 00:12:34 - 0.99  
....  
04/01/11 00:00:34 - 2.34  
04/01/11 00:11:22 - 4.56

but I have been unable to get the first value of each group.

Saurabh Gokhale
  • 53,625
  • 36
  • 139
  • 164
Tim Connell
  • 41
  • 2
  • 1
    Are you using Oracle? Are you simply trying to get the first record for each day? – p.campbell May 06 '11 at 02:10
  • 1
    What DBMS? The answer in SQL Server will be much different than the one for Oracle, which will be much different from the one for MySQL, etc... `SQL` isn't enough information to provide an answer. – Ken White May 06 '11 at 02:13
  • @Ken White: the code is Standard SQL-92 so I think the `sql` tag is enough (i.e. why not post an answer using Standard SQL-92?) – onedaywhen May 06 '11 at 07:09

1 Answers1

2

Here's a solution for SQL Server. The idea is to use a correlated subquery to identify the minimum RecordTime for each year and month. It should be easy to translate the idea to your flavour of DBMS.

WITH Data AS (
    SELECT CONVERT(DATETIME, '1-May-2011 01:00:00') AS RecordTime, 1.23 AS FormattedValue
    UNION ALL SELECT CONVERT(DATETIME, '1-May-2011 01:12:34'), 0.99
    UNION ALL SELECT CONVERT(DATETIME, '1-May-2011 23:59:59'), 2.34
    UNION ALL SELECT CONVERT(DATETIME, '2-May-2011 00:11:22'), 4.56
)
SELECT D1.*
FROM Data D1
WHERE D1.RecordTime = (SELECT MIN(RecordTime)
                       FROM Data D2
                       WHERE YEAR(D2.RecordTime) = YEAR(D1.RecordTime)
                       AND MONTH(D2.RecordTime) = MONTH(D1.RecordTime))

Edit: If your dataset is large then joining on the subquery should be substantially faster, but might be harder to translate into your flavour of DBMS:

SELECT D1.*
FROM Data D1
INNER JOIN (
    SELECT MIN(D2.RecordTime) AS FirstMonthlyRecordTime
    FROM Data D2
    GROUP BY YEAR(D2.RecordTime), MONTH(D2.RecordTime)
) AS SubQuery
    ON D1.RecordTime = SubQuery.FirstMonthlyRecordTime

Edit 2: The following code is the Standard SQL-92 version of my first query above (tested on mySQL):

SELECT T1.RecordTime, T1.FormattedValue
FROM CDBHistoric T1
WHERE T1.RecordTime = (SELECT MIN(T2.RecordTime)
                       FROM CDBHistoric T2
                       WHERE EXTRACT(YEAR FROM T1.RecordTime) = EXTRACT(YEAR FROM T2.RecordTime)
                       AND EXTRACT(MONTH FROM T1.RecordTime) = EXTRACT(MONTH FROM T2.RecordTime))
srgerg
  • 18,719
  • 4
  • 57
  • 39