Using LAG() function to find past value - sql

Question

I've used the Lag Function to find the previous value. However I have run into an issue that requires a more complex query.
Here is my scenario.
Our table currently keeps month end data for each record. With the exception of the last 95 days. we like to keep daily records fro the last 95 days. This is what I mean by month end and daily records
ID Date Amount
123 10/31/2019 52
123 11/31/2019 56
123 12/31/2019 59
123 01/25/2020 32
123 01/26/2020 28
123 ... ..
123 03/12/2020 103
Imagine that the ... represent a daily record for id: 123 up until yesterday.
My task worked perfectly for our month end historical data, but i ran into an issue with our daily historical data
what I want is to be able to get the last value from the last day of the previous month for all months.
this is what I currently have for my query
Select ID, Date, Amount,LAG(Amount, 1, 0) OVER(PARTITION BY ID
ORDER BY id,
Date)
AS SharePreviousBalance from dbo.shares
where date >= 20191031
This is the output I would like to have, but my current query does not work the way i want it to work:
ID Date Amount SharePreviousBalance
123 10/31/2019 52 0
123 11/31/2019 56 52
123 12/31/2019 59 56
123 ... .. ..
123 01/25/2020 32 0
123 01/26/2020 28 0
123 01/27/2020 28 0
123 ... .. ..
123 01/31/2020 28 59
123 ... .. ..
123 02/15/2020 28 0
123 ... .. ..
123 02/29/2020 25 28
123 ... .. ..
123 03/05/2020 29 0
123 ... .. ..
123 03/10/2020 30 0
123 ... .. ..
123 03/12/2020 103 25
Any Ideas?
Thank you

Answer 1

With a little conditional logic, you can still do this with lag():
select
t.*,
case when date = eomonth(date) then
coalesce(
lag(amount) over(
partition by id, case when date = eomonth(date) then 1 else 0 end
order by date
),
0
)
end SharePreviousBalance
from mytable t
The idea is to build a partition for "end-of-month" rows (ie rows whose date is the last day of a month). Within that partition, an end-of-month row can access the previous end of month with lag().
Demo on DB Fiddle - I added a few rows to your sample data:
ID | Date | Amount | SharePreviousBalance
--: | :--------- | -----: | -------------------:
123 | 2019-10-31 | 52 | 0
123 | 2019-11-30 | 56 | 52
123 | 2019-12-31 | 59 | 56
123 | 2020-01-20 | 28 | null
123 | 2020-01-25 | 32 | null
123 | 2020-01-26 | 28 | null
123 | 2020-01-31 | 28 | 59
123 | 2020-02-12 | 103 | null
123 | 2020-02-28 | 103 | null
123 | 2020-02-29 | 103 | 28
If you also want to show the value of the previous end of month for the current date, then add that row to the "end-of-month" partition:
select
t.*,
case when date in (eomonth(date), cast(getdate() as date)) then
coalesce(
lag(amount) over(
partition by
id,
case when date in (eomonth(date), cast(getdate() as date)) then 1 else 0 end
order by date
),
0
)
end SharePreviousBalance
from mytable t
order by id, date