Question

I have a table that i wanted to create some charts.
Basically i need to know how many registers i have in interval of 10 minutes and how many of them are already pos-processed. With this 2 queries i have the info i need, but i need they 2 in just 1 result table.

First query

SELECT  
    (
        CAST(DATEPART(HOUR, m.Ocr_DataHora) AS NVARCHAR) +
        ':' +
        CAST((DATEPART(MINUTE, m.Ocr_DataHora) % 6) AS NVARCHAR) +
        '0'         
    )                                       AS Hora,
    COUNT(*)                                AS Movimentações
FROM
    Integracao m
WHERE
        m.Ocr_DataHora      >=  '2012-09-17 00:00:00.000'
    AND m.Ocr_DataHora      <   '2012-09-18 00:00:00.000'
GROUP BY        
    DATEPART(HOUR, m.Ocr_DataHora),
    (DATEPART(MINUTE, m.Ocr_DataHora) % 6)
ORDER BY    
    DATEPART(HOUR, m.Ocr_DataHora),
    (DATEPART(MINUTE, m.Ocr_DataHora) % 6)

Second query

SELECT  
    (
        CAST(DATEPART(HOUR, s.Ocr_DataHora) AS NVARCHAR) +
        ':' +
        CAST((DATEPART(MINUTE, s.Ocr_DataHora) % 6) AS NVARCHAR) +
        '0'         
    )                                       AS Hora,
    COUNT(*)                                AS Sucesso
FROM
    Integracao s
WHERE
        s.Veiculo_Modelo    <>  ''
    AND s.Ocr_DataHora      >=  '2012-09-17 00:00:00.000'
    AND s.Ocr_DataHora      <   '2012-09-18 00:00:00.000'
GROUP BY        
    DATEPART(HOUR, s.Ocr_DataHora),
    (DATEPART(MINUTE, s.Ocr_DataHora) % 6)

How can i join them if my common field is generated dynamically?

Answer 1

As others have already told you, you could just use the two queries as derived tables (subselects) and join them using each one’s Hora column. It doesn’t matter if the column is computed in the query, it’s perfectly legal to use it in a join condition.

However, I think you can get the same results more efficiently, by calculating both counts in a single query (i.e. without subqueries). Here’s how you can do that:

SELECT  
    (
        CAST(DATEPART(HOUR, m.Ocr_DataHora) AS NVARCHAR) +
        ':' +
        CAST((DATEPART(MINUTE, m.Ocr_DataHora) % 6) AS NVARCHAR) +
        '0'         
    )                                                  AS Hora,
    COUNT(*)                                           AS Movimentações,
    COUNT(CASE WHEN s.Veiculo_Modelo <> '' THEN 1 END) AS Sucesso
FROM
    Integracao m
WHERE
        m.Ocr_DataHora      >=  '2012-09-17 00:00:00.000'
    AND m.Ocr_DataHora      <   '2012-09-18 00:00:00.000'
GROUP BY        
    DATEPART(HOUR, m.Ocr_DataHora),
    (DATEPART(MINUTE, m.Ocr_DataHora) % 6)
ORDER BY    
    DATEPART(HOUR, m.Ocr_DataHora),
    (DATEPART(MINUTE, m.Ocr_DataHora) % 6)

As you can see, the query doesn’t filter rows on s.Veiculo_Modelo <> ''. However, that condition is used when counting rows for the Sucesso column. The argument of the second COUNT is a CASE expression. It returns a value (arbitrarily chosen to be 1, but could be anything really) if the specified condition is met and NULL otherwise. Since COUNT omits nulls, the result will be the same as with your dedicated query calculating and returning Sucesso.