Include date boundaries to segment select results - sql

Question

I want to return in the same row the results of a particular measure on different time range.
For instance "Last 30 days", "Last 7 days", "Last 3 days".
On doing that I've initially used the UNION function, and created multiple sub-queries for every time range. The downside of doing that is I'm collecting the same numbers three times, with a consequent increase in running time.
A colleague suggested me to use the CASE function to segment the time range. I've initially thought to implement it as follows:
select tp.Name,
case when pub.Date between DATEADD(day, -31, getdate()) and DATEADD(day, -1, getdate())
then SUM(Impressions) else 0 end 'Last 30 days Impressions',
case when pub.Date between DATEADD(day, -31, getdate()) and DATEADD(day, -1, getdate())
then SUM(Revenue * rler.Rate) else 0 end 'Last 30 days Revenues',
case when pub.Date between DATEADD(day, -8, getdate()) and DATEADD(day, -1, getdate())
then SUM(Impressions) else 0 end 'Last 7 days Impressions',
case when pub.Date between DATEADD(day, -8, getdate()) and DATEADD(day, -1, getdate())
then SUM(Revenue * rler.Rate) else 0 end 'Last 7 days Revenues',
case when pub.Date between DATEADD(day, -4, getdate()) and DATEADD(day, -1, getdate())
then SUM(Impressions) else 0 end 'Last 3 days Impressions',
case when pub.Date between DATEADD(day, -4, getdate()) and DATEADD(day, -1, getdate())
then SUM(Revenue * rler.Rate) else 0 end 'Last 3 days Revenues'
from ...
where ...
group by tp.Name, tp.Kind, pub.Date
order by 'Last 30 days Impressions'
Unfortunately this will return a row for each Name, Kind and Date which is not what I want. The issue I think relies on the pub.Date in the GROUP BY call. What should I do to overcome the issue?

Answer 1

You can't quite implement your colleague's suggestion because the times overlap. You can do non-overlapping ranges, something like this:
select tp.Name,
(case when pub.Date between DATEADD(day, -3, getdate()) and DATEADD(day, -1, getdate())
then 'Last 3 days Impressions'
when pub.Date between DATEADD(day, -7, getdate()) and DATEADD(day, -1, getdate())
then '4-7 days Impressions'
when pub.Date between DATEADD(day, -31, getdate()) and DATEADD(day, -1, getdate())
then '8-31 days Impressions'
else 'Older'
end) as TimeRange,
SUM(Impressions) as NumImpressions,
. . .
from . . .
where . . .
group by tp.Name,
(case when pub.Date between DATEADD(day, -3, getdate()) and DATEADD(day, -1, getdate())
then 'Last 3 days Impressions'
when pub.Date between DATEADD(day, -7, getdate()) and DATEADD(day, -1, getdate())
then '4-7 days Impressions'
when pub.Date between DATEADD(day, -31, getdate()) and DATEADD(day, -1, getdate())
then '8-31 days Impressions'
else 'Older'
end)

Answer 2

I'm answering my own question. The result can be achieve using the following code:
select tp.Name,
sum(case when pub.Date between DATEADD(day, -31, getdate()) and DATEADD(day, -1, getdate())
then Impressions else 0 end) 'Last 30 days Impressions',
sum(case when pub.Date between DATEADD(day, -31, getdate()) and DATEADD(day, -1, getdate())
then (Revenue * rler.Rate) else 0 end) 'Last 30 days Revenues',
sum(case when pub.Date between DATEADD(day, -8, getdate()) and DATEADD(day, -1, getdate())
then Impressions else 0 end) 'Last 7 days Impressions',
sum(case when pub.Date between DATEADD(day, -8, getdate()) and DATEADD(day, -1, getdate())
then (Revenue * rler.Rate) else 0 end) 'Last 7 days Revenues',
sum(case when pub.Date between DATEADD(day, -4, getdate()) and DATEADD(day, -1, getdate())
then Impressions else 0 end) 'Last 3 days Impressions',
sum(case when pub.Date between DATEADD(day, -4, getdate()) and DATEADD(day, -1, getdate())
then (Revenue * rler.Rate) else 0 end) 'Last 3 days Revenues'
from ...
where ...
group by tp.Name, tp.Kind
order by 'Last 30 days Impressions' desc