My table looks like this:
metro_region, value, date
with multiple values (one for each date) across the month of November. There are about 100 metro regions.
I want my report to have the following data:
Metro_region Today Yesterday 2daysAgo 3dayAgo
MetroRegionA 40.1 54.3 64.8 48.1
MetroRegionB 31.1 53.1 97.8 43.2
What I tried:
select
metro_region,
date,
LAG(value,3) over (Partition by metro order by metro) as "3daysAgo",
LAG(value,2) over (Partition by metro order by metro) as "2daysAgo",
LAG(value,1) over (Partition by metro order by metro) as "Yesterday",
value as Today
from mytable
where date = curdate();
I suspect I'm not partitioning properly...or merely grossly missing how to use LAG
...any insights are appreciated!
You need results of today and the last 3 days, so you must not filter only for today.
First get the results for 4 days and then filter for today.
Also, in every partition you must sort by date
and not metro
:
select metro_region, date, `3daysAgo`, `2daysAgo`, `Yesterday`, Today
from (
select metro_region,
date,
LAG(value,3) over (partition by metro_region order by date) as `3daysAgo`,
LAG(value,2) over (partition by metro_region order by date) as `2daysAgo`,
LAG(value,1) over (partition by metro_region order by date) as `Yesterday`,
value as Today
from mytable
where date between curdate() - interval 3 day and curdate()
) t
where date = curdate()
See a simplified demo.