Question

This is simple in theory, but difficult for me to figure out.

I have two SQL Server tables:

1. List of purchases with purchase date and total

   select total, date from purchases
   

2. list of miles traveled for a specific job and date of travel

   select travelDate, miles from trips
   

EDIT: To keep the answer/discussion on my question, I am rephrasing the requirement.

I need to figure out the total miles driven between each purchase.

I want more accuracy than an overall average.

I can manually get this by summing all miles from trips in between each purchases date. Now, I just want to automate the process.

The grouping should be such that all trips dates greater than purchases date A and less than purchases date B are part of the purchases date A group.

Answer 1

Curing my denseness, I see that your request is quite reasonable by treating the problem as “replacement fuel cost”—thus using the fuel cost of the next fueling rather than the previous cost to buy the fuel actually used (which gets really complicated, really fast). Volume then doesn’t matter. Try this on for size.

SELECT
   T.*,
   P.*, -- from previous purchase
   N.*, -- from next purchase (NULL if none yet)
   TripCost = N.Total * T.Miles / M.MilesThisFill
FROM
   dbo.Trips T
   CROSS APPLY (
      SELECT TOP 1 *
      FROM dbo.Purchases P
      WHERE P.[Date] < T.travelDate
      ORDER BY P.[Date] DESC
   ) P
   CROSS APPLY (
      SELECT TOP 1 *
      FROM dbo.Purchases P
      WHERE P.[Date] > T.travelDate
      ORDER BY P.[Date]
   ) N
   CROSS APPLY (
      SELECT Sum(miles) MilesThisFill
      FROM dbo.Trips T2
      WHERE
         T2.[Date] > P.[Date]
         AND T2.[Date] < N.[Date]
   ) M;

Or here’s a version that thinks very differently about the problem but should give the same result. Let me know which one performs better, would ya? (SET STATISTICS IO ON; SET STATISTICS TIME ON;)

WITH PSeq AS (
   SELECT
      Seq = Row_Number() OVER (ORDER BY [Date]),
      *
   FROM dbo.Purchases
), Slices AS (
   SELECT
      FromDate = P.[Date],
      ToDate = N.[Date],
      N.Total
   FROM
      PSeq P
      INNER JOIN PSeq N
         ON P.Seq + 1 = N.Seq
), TotalMiles AS (
    SELECT
       S.FromDate,
       Sum(T.Miles) MilesThisFill
    FROM
      Slices S
      INNER JOIN dbo.Trips T
         ON T.travelDate BETWEEN S.FromDate AND S.ToDate
    GROUP BY
       S.FromDate 
)
SELECT
   T.travelDate,
   S.FromDate,
   S.ToDate,
   TripCost = S.Total * T.Miles / M.MilesThisFill
FROM
   Slices S
   INNER JOIN dbo.Trips T
      ON T.travelDate BETWEEN S.FromDate AND S.ToDate 
   INNER JOIN dbo.TotalMiles M
      ON S.FromDate = L.FromDate;

I apologize in advance for any typos or errors… I haven’t tested the code.

And just for laughs, here’s the first query transmogrified into a version that would work even on SQL Server 2000!

SELECT
   T.travelDate,
   T.Miles,
   T.ToDate,
   TripCost = P.Total * T.Miles / M.MilesThisFill
FROM
   (
      SELECT
         T.travelDate,
         T.Miles,
         ToDate = (
            SELECT TOP 1 P.Date
            FROM dbo.Purchases P
            WHERE P.[Date] > T.travelDate
            ORDER BY P.[Date]
         )
      FROM
         dbo.Trips T
   ) T
   INNER JOIN (
      SELECT
         ToDate = (
            SELECT TOP 1 P.Date
            FROM dbo.Purchases P
            WHERE P.[Date] > T2.travelDate
            ORDER BY P.[Date]
         ),
         MilesThisFill = Sum(T2.Miles)
      FROM dbo.Trips T2
      GROUP BY
         (
            SELECT TOP 1 P.Date
            FROM dbo.Purchases P
            WHERE P.[Date] > T2.travelDate
            ORDER BY P.[Date]
         )
   ) M ON T.ToDate = M.ToDate
   INNER JOIN dbo.Purchases P
      ON T.ToDate = P.[Date];

This actually exposes that I might not need to look up the previous purchase date in my first query, if I do it right… so here’s a final version:

WITH TripData AS (
   SELECT
      T.Miles,
      T.travelDate,
      ToDate = (
         SELECT TOP 1 P.[Date]
         FROM dbo.Purchases P
         WHERE P.[Date] > T.travelDate
         ORDER BY P.[Date]
      )
   FROM
      dbo.Trips T
)
SELECT
   T.*,
   P.*,
   TripCost = P.Total * T.Miles / M.MilesThisFill
FROM
   TripData T
   INNER JOIN dbo.Purchases P
      ON T.ToDate = P.[Date]
   INNER JOIN (
      SELECT
         T2.ToDate,
         Sum(T2.Miles) MilesThisFill
      FROM TripData T2
      GROUP BY
         T2.ToDate
   ) M ON T.ToDate = M.ToDate;

Note: the order of the TripCost expression is important, because Miles and TotalMiles are integers. If you put P.Total last, you will get wrong answers because Miles / TotalMiles will be converted to an integer.