Question

I have a table of data for a particular stock that contains every single transaction going back a few years. I want to group this data on a per-day basis, so that I can see things like the daily volume, etc.

The best way I have come up with so far is:

select 
datepart(year, date),
datepart(month, date),
datepart(day, date), 
sum(volume) from hi 
group by 
datepart(year, date),
datepart(month, date),
datepart(day, date)
order by
datepart(year, date),
datepart(month, date),
datepart(day, date)

My SQL is somewhat limited, so I’m wondering if there’s a more efficient way to do this?

Answer 1

SELECT DateAdd(dd, sub.DayCount, 0) as TheDate, SUM(sub.Volume)
FROM
(SELECT DateDiff(dd, 0, date) as DayCount, Volume FROM hi)
as Sub
GROUP BY Sub.DayCount
ORDER BY Sub.DayCount

As far as efficiency goes, both queries must read the whole table once. They incur the same IO. You should messure IO with SET STATISTICS IO ON to see if that’s a problem.

Since it seems that IO is the real issue, one option is to create an index which includes both Volume, DayCount (and no other columns).