Can't use case & aggregation correctly - sql

I have the following table
Cash_table
ID Cash Rates Amount
1 50 3 16
2 100 4 25
3 130 10 7
3 130 10 6
4 13 7 1.8
5 30 8 2.5
5 30 10 1
6 10 5 2
What I want as a result is to cumulate all the entries that have a Count(id)>1 like this:
ID New_Cash New_Rates New_Amount
1 50 3 16
2 100 4 25
3 130 10+10 130/(10+10)
4 13 7 1.8
5 30 8+10 30/(8+10)
6 10 5 2
So I only want to change the rows where Count(id)>1 and leave the rest like it was.
For the rows with count(id)>1 I want to sum up the rates and take the cash and divide it by the sum of the rates. The Rates alone aren't a problem since I can sum them up and group by id and get the desired result.
The problem is with the New_Amount column:
I am trying to do it with a case statement but it isn't working:
select id,
cash as new_cash,
sum(rates) as new_rates,
(case count(id)
when 1 then amount
else cash/sum(nvl(rates,null))
end) as new_amount
from Cash_table
group by id

As the cash value is always the same for an ID, you can group by that as well:
select id,
cash as new_cash,
sum(rates) as new_rates,
case count(id)
when 1 then max(amount)
else cash/sum(rates)
end as new_amount
from cash_table
group by id, cash
order by id
ID NEW_CASH NEW_RATES NEW_AMOUNT
---------- ---------- ---------- ----------
1 50 3 16
2 100 4 25
3 130 20 6.5
4 13 7 1.8
5 30 18 1.66666667
6 10 5 2
The first branch of the case expression needs an aggregate because you aren't grouping by amount; and the sum(nvl(rates,null)) can just be sum(rates). If you're expecting any null rates then you need to decide how you want the amount to be handled, but nvl(rates,null) isn't doing anything.
You can do the same thing without a case expression if you prefer, manipulating all the values - which might be more expensive:
select id,
cash as new_cash,
sum(rates) as new_rates,
sum(amount * rates)/sum(rates) as new_amount
from cash_table
group by id, cash
order by id

Related

Use each row's computed result to next row and further

I have a table listed below and want to calculate new stock. initially new stock is 0
id qty price operation newstock
1 20 23 buy
2 10 12 sale
3 15 14 buy
4 6 13 buy
and so on
want output like- in case of buy- newstock+qty
and in case of sale- newstock-qty
id qty price operation newstock
1 20 23 buy 23 (eg- 0+23= 23)
2 10 12 sale 13 (previous row 23-10)
3 15 14 buy 28 (previous row 13+15)
4 6 13 buy 32 ( previous row 28+6)
I used cross apply also but not getting result.
I think you want a cumulative sum:
select t.*,
sum(case when operation = 'buy' then qty else -qty end) over (order by id) as newstock
from t

Decreasing of the sum by percentage (subtraction of percentage) SQL

Substraction of percentage
I have got a table where is one record. Table looks like:
cust_code, SUM
1 25
I need to calculate a subtraction like this.
cust_code, ID, SUM
1 1 25
1 2 23
1 3 21.16
1 4 19.47
1 5 17.91
1 6 16.48
1 7 15.16
1 8 13.95
. . .
. . .
. . .
1 15 7.78
where value of sum in record 2 is subtracted by 8% of record 1,
value of sum in record 3 is subtracted by 8% of record 2,
value of sum in record 4 is subtracted by 8% of record 3, etc.
Max ID will be 15.
It should be single query, I can use any additional external table (containing for example simple counter from 0 to 15).
Best regards,
Volcano
Easy to do with a recursive CTE:
WITH cte AS
(SELECT cust_code, 1 AS id, sum FROM cust
UNION ALL
SELECT cust_code, id + 1, sum * 0.92 FROM cte WHERE id < 15)
SELECT * FROM cte ORDER BY id;
cust_code id sum
---------- ---------- ----------
1 1 25.0
1 2 23.0
1 3 21.16
1 4 19.4672
1 5 17.909824
1 6 16.4770380
1 7 15.1588750
1 8 13.9461650
1 9 12.8304718
1 10 11.8040340
1 11 10.8597113
1 12 9.99093444
1 13 9.19165969
1 14 8.45632691
1 15 7.77982076

Use aggregation only on rows where count(ID) is greater than one

Hi I have the following table
Cash_table
ID Cash Rates
1 50 3
2 100 4
3 70 10
3 60 10
4 13 7
5 20 8
5 10 10
6 10 5
What I want as a result is to cumulate all the entries that have a Count(id)>1 like this:
ID New_Cash New_Rates
1 50 3
2 100 4
3 (70+60)/(10+10) 10+10
4 13 7
5 (20+10)/(8+10) 8+10
6 10 5
So I only want to change the rows where Count(id)>1 and leave the rest like it was.
For the rows with count(id)>1 I want to sum up the rates and take the sum of the cash and divide it by the sum of the rates. The Rates alone aren't a problem since I can sum them up and group by id and get the desired result.
The problem is with the cash column:
I am trying to do it with a case statement but it isn't working:
select id, sum(rates) as new_rates, case
when count(id)>1 then sum(cash)/nullif(sum(rates),0))
else cash
end as new_cash
from Cash_table
group by id
You only need group by id and aggregate:
select
id,
sum(cash) / (case count(*) when 1 then 1 else sum(rates) end) as new_cash,
sum(rates) as new_rates
from Cash_table
group by id
order by id
See the demo.
You can aggregate rate and cash columns by sum() function with grouping by id
select
id,
sum(cash)/decode( sum( nvl(rates,0) ), 0 ,1, sum( nvl(rates,0) )) as new_cash,
sum(rates) as new_rates
from cash_table
group by id
there's no nullif() function in Oracle, use nvl() instead
switch case part ( where decode() function is used ) against the
possibility of division by zero

sql best strategy to partition same values based on temporal sequence

I have data that looks like this, where there are multiple values for each ID that correspond to an ascending date variable:
ID LEVEL DATE
1 10 10/1/2000
1 10 11/20/2001
1 10 12/01/2001
1 30 02/15/2002
1 30 02/15/2002
1 20 05/17/2002
1 20 01/04/2003
1 30 07/20/2003
1 30 03/16/2004
1 30 04/15/2004
I want to acquire a count per each ID/LEVEL/DATE block that looks like this:
ID LEVEL COUNT
1 10 3
1 30 2
1 20 2
1 30 3
The problem is that if I use the count windows function and partition by level, it groups 30 together regardless of the temporal sequence. I want the count for level 30 both before and after 20 to be distinct. Does anyone know how to do that?
A standard gaps and islands solution using ROW_NUMBER(), if it's available on your particular DBMS...
WITH
ordered AS
(
SELECT
*,
ROW_NUMBER() OVER (PARTITION BY id ORDER BY date) AS set_ordinal,
ROW_NUMBER() OVER (PARTITION BY id, level ORDER BY date) AS grp_ordinal
FROM
yourData
)
SELECT
id,
level,
set_ordinal - grp_ordinal,
MIN(date),
COUNT(*)
FROM
ordered
GROUP BY
id,
level,
set_ordinal - grp_ordinal
ORDER BY
id,
MIN(date)
Visualising the effect of the two row numbers...
ID LEVEL DATE set_ordinal grp_ordinal set-grp GROUP
-- ----- ---------- ----------- ----------- ------- --------
1 10 10/01/2000 1 1 0 1,10,0
1 10 11/20/2001 2 2 0 1,10,0
1 10 12/01/2001 3 3 0 1,10,0
1 30 02/15/2002 4 1 3 1,30,3
1 30 02/15/2002 5 2 3 1,30,3
1 20 05/17/2002 6 1 5 1,20,5
1 20 01/04/2003 7 2 5 1,20,5
1 30 07/20/2003 8 3 5 1,30,5
1 30 03/16/2004 9 4 5 1,30,5
1 30 04/15/2004 10 5 5 1,30,5

Oracle SQL find row crossing limit

I have a table which has four columns as below
ID.
SUB_ID. one ID will have multiple SUB_IDs
Revenue
PAY where values of Pay is always less than or equal to Revenue
select * from Table A order by ID , SUB_ID will have data as below
ID SUB_ID REVENUE PAY
100 1 10 8
100 2 12 9
100 3 9 7
100 4 11 11
101 1 6 5
101 2 4 4
101 3 3 2
101 4 8 7
101 5 4 3
101 6 3 3
I have constant LIMIT value 20 . Now I need to find the SUB_ID which Revenue crosses the LIMIT when doing consecutive SUM using SUB_ID(increasing order) for each ID and then find total Pay ##. In this example
for ID 100 Limit is crossed by SUB ID 2 (10+12) . So total Pay
is 17 (8+9)
for ID 101 Limit is crossed by SUB ID 4
(6+4+3+8) . So total Pay is 18 (5+4+2+7)
Basically I need to find the row which crosses the Limit.
Fiddle: http://sqlfiddle.com/#!4/4f12a/4/0
with sub as
(select x.*,
sum(revenue) over(partition by id order by sub_id) as run_rev,
sum(pay) over(partition by id order by sub_id) as run_pay
from tbl x)
select *
from sub s
where s.run_rev = (select min(x.run_rev)
from sub x
where x.id = s.id
and x.run_rev > 20);