How to build date change query using presto sql - sql

Question

I need to make a tracker containing item name and price date change.
The date shows when the price is changed. I used lead, but it shows each day not when the day has changed.
for instance,
item name = A
price date = 2022-11-21, price = $4
price date = 2022-11-25, price = $3
price date = 2022-11-30, price = $4
The expectation for the result is:
start date, next date, price
2022-11-21 2022-11-24 $4
2022-11-25 2022-11-29 $3
2022-11-30 2023-02-14 (current date) $4
Any help would be appreciated.
#Updated:
The dataset is containing daily price
for instance :
item name = A
price date = 2022-11-21, price = $4
price date = 2022-11-22, price = $4
price date = 2022-11-23, price = $4
price date = 2022-11-24, price = $4
price date = 2022-11-25, price = $3
price date = 2022-11-26, price = $3
price date = 2022-11-27, price = $3
price date = 2022-11-29, price = $3
price date = 2022-12-01, price = $4
Query :
select
item_name,
supplier_name,
price_date,
price,
lead(price_date) over (partition by item_name order by price) as next_price from
price
Result:
price = $4, 2022-11-21, 2022-11-22
price = $4, 2022-11-22, 2022-11-23
price = $4, 2022-11-23, 2022-11-24
price = $3, 2022-11-25, 2022-11-26
price = $3, 2022-11-26, 2022-11-27
price = $3, 2022-11-27, 2022-11-29
price = $3, 2022-11-29, 2022-11-30
price = $4, 2022-12-01, 2023-02-14
While my expectation is :
price = $4, 2022-11-21, 2022-11-25
price = $3, 2022-11-26, 2022-11-30
price = $4, 2023-12-01, 2023-02-14

Answer 1

You can try using gaps-and-island approach - i.e. introduce a column representing the change in price, use cumulative sum to calculate groups and use group by to calculate the results:
-- sample data
with dataset(date, price) as (
values (date '2022-11-21', 4),
(date '2022-11-22', 4),
(date '2022-11-23', 4),
(date '2022-11-24', 4),
(date '2022-11-25', 3),
(date '2022-11-26', 3),
(date '2022-11-27', 3),
(date '2022-11-29', 3),
(date '2022-12-01', 4)
)
-- query
select arbitrary(price) price,
min(date) as start_dt,
max(date) as end_dt
from (
select date, price, sum (change) over (order by date) as grp
from (
select *, if(lag(price) over (order by date) != price, 1, 0) as change
from dataset)
)
group by grp;
Output:
price
start_dt
end_dt
4
2022-11-21
2022-11-24
3
2022-11-25
2022-11-29
4
2022-12-01
2022-12-01
Few notes:
since your actual data is partitioned - do not forget to add item_name for window partitioning and final group by
if you really need current date as the end_dt for final row in partition there are several ways to achieve that - not very fun with inserting missing dates (see this for inspiration) or you can just roll up another subquery which will check if lead(end_dt) over (partition ... order by end_dt) is null and use current date for end_dt