TOP 1 or LIMIT 1 not working in a Correlated Subquery - SPS 11 - sql

Question

I am trying to work on a query to fill the Price for a product where it is zero with non-zero value fro previous record. I tried to write a simple correlated subquery but its not working.
var_1 = select * from "XYZ"."PRD_TEST" where price <> 0 order by period desc;
var_out = select a.product,a.period, ( select price from :var_1 b where a.product = b.product and a.period > b.period and b.period <> 0 limit 1 ) as price from "XYZ"."PRD_TEST" a;
PRODUCT PERIOD PRICE
A 1 100
A 2 0 - to be filled with 100
A 3 0 - to be filled with 100
A 4 5
A 5 0 - to be filled with 5
I tried to replace the sub-query with scalar function but it does not take table as a Parameter.
I tried to achieve the output using Left outer join and Row_number but it's too expensive and runs for a long time.
I am looking for a best option to fetch only 1 record in the subquery just like TOP 1.

Answer 1

you could use a scalar subquery like this ( select ... limit 1 is not considered scalar in HANA, unfortunately):
Do begin
var_1 = select * from (( Select 'A' product, '1' period, '100' price from sys.dummy )
Union ( Select 'A' product, '2' period, '0' price from sys.dummy )
Union ( Select 'A' product, '3' period, '0' price from sys.dummy )
Union ( Select 'A' product, '4' period, '5' price from sys.dummy )
Union ( Select 'A' product, '5' period, '0' price from sys.dummy )) order by period desc;
var_out = ( select a.product,
a.period,
( select max(price)
from :var_1 b
where a.product = b.product
and a.period > b.period
and b.period <> 0
and b.period = ( select max(period) from :var_1 c
where a.product = c.product
AND a.period > c.period
and c.period <> 0
and c.price <> 0
)) as price
from :var_1 a where price = '0' )
union (select product, period, price from :var_1 where price <> '0' );
select * from :var_out order by product, period;
end
Tested on sps12
Added after comment.
Why don't you just try? It is very simple.
Because I was curious I tried it on my HCP trial instance, takes about 1 sec for on milion rows. I included a ifnull to avoid rows with null-values where there is no price in earlier periods.
Here is the coding:
drop table var_1;
create column table var_1 as
(
select
cast ( 'Prod' || prd.GENERATED_PERIOD_START as nvarchar(20) ) product,
cast ( per.GENERATED_PERIOD_START as decimal(2)) period,
cast ( case when rand() < '0.5' then rand() * '100' else '0' end
as decimal(5,2)) as price -- ~50% of price is 0
from series_generate_integer(1,0,1000000/13) as prd, --~1Mio records
series_generate_integer(1,0,13) as per --12 periods + period 0
);
merge delta of var_1;
select * from var_1
order by product, period
limit 100;
do begin sequential execution -- don't let parallel execution influence the runtime-measurement
declare start_timestamp timestamp;
start_timestamp = current_timestamp;
var_out = ( select a.product,
a.period,
ifnull ((select max(price)
from var_1 b
where a.product = b.product
and a.period > b.period
and b.period <> 0
and b.period = ( select max(period) from var_1 c
where a.product = c.product
AND a.period > c.period
and c.period <> 0
and c.price <> 0
)),'0.0') as price
from var_1 a where price = '0' )
union (select product, period, price from var_1 where price <> '0' );
select nano100_between(:start_timestamp, (select current_timestamp from dummy) )/10000 as runtime_millisec from dummy;
select * from :var_out
order by product, period
limit 100;
end