SQL：通过分区聚合

Question

Query: Considering only Italian routes, for each category of goods and for each year

• select the average daily income for each month

• the total monthly income since the beginning of the year

SQL:

SELECT
  gc.GoodCategory,
  tm.Month,
  tm.Year,
  SUM(ro.Income) / COUNT(DISTINCT tm.Date),
  SUM(ro.Income) OVER (PARTITION BY gc.GoodCategory, tm.Year 
    ORDER BY tm.Month ROWS UNBOUNDED PRECEDING)

FROM FactRoutes ro,
     DimLocation dp,
     DimLocation ds,
     DimGoodCategory gc,
     DimTime tm

WHERE ro.DepartureID = dp.LocationID
    AND ro.DestinationID = ds.LocationID
    AND ro.GoodCategoryID = gc.GoodCategoryID
    AND ro.GoodTimeID = tm.GoodTimeID
    AND dp.State = 'Italy'
    AND ds.State = 'Italy'

GROUP BY gc.GoodCategory,
     tm.Month,
     tm.Year;

But facing the below error

Column 'FactRoutes.Income' is invalid in the select list
because it is not contained in either an aggregate function
or the GROUP BY clause.

whats the better way to handle it?

Answer 1

I think that you want:

SELECT
  gc.GoodCategory,
  tm.Month,
  tm.Year,
  SUM(ro.Income) / COUNT(DISTINCT tm.Date),
  SUM(SUM(ro.Income)) OVER (PARTITION BY gc.GoodCategory, tm.Year ORDER BY tm.Month)
FROM FactRoutes ro
INNER JOIN DimLocation dp ON ro.DepartureID = dp.LocationID
INNER JOIN DimLocation ds ON ro.DestinationID = ds.LocationID
INNER JOIN DimGoodCategory gc ON ro.GoodCategoryID = gc.GoodCategoryID
INNER JOIN DimTime tm ON ro.GoodTimeID = tm.GoodTimeID
WHERE dp.State = 'Italy' AND ds.State = 'Italy'
GROUP BY gc.GoodCategory, tm.Month, tm.Year;

The main point is that, in order to make your query is not a valid aggregate query, you need to use an aggregate function within the window function, like SUM(SUM(ro.Income)) OVER (...) instead of just SUM(ro.Income) OVER(...) , so you get a window sum over the previous groups of records.

Other notable points:

• always use explicit joins (with the ON keyword) rather than old-school, implicit joins (with commas in the FROM clause), whose syntax has fallen out of favor for decades

• ROWS UNBOUNDED PRECEDING is not needed; your window function has an ORDER BY clause so that's what it does anyway