Im trying to make a query to get the difference of avg(score1+score2/2) from 'current' and the most recent 'archived' . to do a chart on Oracle Apex
Table name: myTable
id | score1 | score2 | status | date
------------------------------------------
1 | 10 | 20 | current| 07/09/19
2 | 20 | 30 |archived| 04/09/19
3 | 15 | 35 |archived| 02/09/19
wanted the result: (avg(score1 + score2/2) where status = 'current') - (avg(score1 + score2/2) where status = 'archived' only the most recent)
Im tried
Hmmm . . . one method is conditional aggregation:
select max(case when status = 'current' then score_avg end), as current_score,
max(case when status = 'archive' then score_avg end), as last_archive_score,
(max(case when status = 'current' then score_avg end) -
max(case when status = 'archive' then score_avg end)
) as diff
from (select t.*,
row_number() over (partition by status order by date desc) as seqnum,
(score1 + score2) / 2 as score_avg
from t
) t
where seqnum = 1;
I am guessing that you really want (score1 + score2) / 2). However, if you want score1 + score2 / 2, then use that expression instead.
u want this?
select status , avg(score1 + score2/2) from you_table
group by status
or
select (select avg(score1 + score2/2) from you_table
where status='current')-(select avg(score1 + score2/2) from you_table
where status='archived') diff from dual
One option would be using
min/max(score1) keep (dense_rank first order by "date" desc) over (partition by status)
to compute the archived case, and an ordinary arithmetic average computation for current case (depending on the sample data, there exists only one row for current case )
with myTable( id, score1, score2, status, "date" )as
(
select 1, 10, 20, 'current' , date'2019-09-07' from dual union all
select 2, 20, 30, 'archived', date'2019-09-04' from dual union all
select 3, 15, 35, 'archived', date'2019-09-02' from dual
), t as
(
select
case when status = 'current' then ( score1 + score2 ) / 2 end as curr,
case when status = 'archived' then
(
(
min(score1) keep (dense_rank first order by "date" desc) over (partition by status)+
min(score2) keep (dense_rank first order by "date" desc) over (partition by status)
)/2
)
end as arch
from myTable
)
select max(curr)-max(arch) as "Avg.Result"
from t;
Demo
Related
Let's say I have a table like this:
user_id
order
action
1
1
start
1
2
other
1
3
other
1
4
end
1
5
other
2
1
start
2
2
other
2
3
end
2
4
other
2
5
start
2
6
other
2
7
end
And I want to create a new column that flags the rows that appear between "start" and "end" events for each user (ordering by "order"):
user_id
order
action
is_between_start_and_end
1
1
start
NULL
1
2
other
1
1
3
other
1
1
4
end
NULL
1
5
other
NULL
2
1
start
NULL
2
2
other
1
2
3
end
NULL
2
4
other
NULL
2
5
start
NULL
2
6
other
1
2
7
end
NULL
How can I achieve this?
Consider below approach
select * except(grp),
if(
countif(action = 'end') over (partition by user_id, grp order by `order`) = 0
and action != 'start', 1, null
) as is_between_start_and_end
from (
select *,
countif(action = 'start') over (partition by user_id order by `order`) as grp
from your_table
)
if applied to sample data in y our question - output is
This can be solved with windows functions.
with tbl as (
Select 1 as user_id, 1 as order_it,"start" as action
Union all select 1 , 2 ,"other"
Union all select 1 , 3 ,"other"
Union all select 1 , 4 ,"end"
Union all select 1 , 5 ,"other"
Union all select 2 , 1 ,"start"
Union all select 2 , 2 ,"other"
Union all select 2 , 3 ,"end"
Union all select 2 , 4 ,"other"
Union all select 2 , 5 ,"start"
Union all select 2 , 6 ,"other"
Union all select 2 , 7 ,"end"
),
helper as (
Select *,
countif(action="end") over win_before as ends,
countif(action="start") over win_before as starts,
first_value(if(action="end" or action="start",action,null) ignore nulls) over (partition by user_id order by order_it rows between current row and unbounded following) as end_to_come
from tbl
window win_before as (partition by user_id order by order_it rows between unbounded preceding and current row)
order by user_id,order_it
)
select *,
if(end_to_come="end" and starts-ends=1,1,null) as is_between_start_and_end
from helper
order by user_id,order_it
This should work but could surely be more optimized
with input as (
select 1 user_id, 1 as order_, 'start' action union all
select 1, 2, 'other' union all
select 1, 3, 'other' union all
select 1 , 4 , 'end' union all
select 1 , 5 , 'other' union all
select 2 , 1 , 'start' union all
select 2 , 2 , 'other' union all
select 2 , 3 , 'end' union all
select 2 , 4 , 'other' union all
select 2 , 5 , 'start' union all
select 2 , 6 , 'other' union all
select 2 , 7 , 'end'
)
select
*,
if (
order_ > max(if(action = 'start', order_, null))
over(partition by user_id order by order_ range between unbounded preceding and current row) and
order_ < min(if(action = 'end', order_, null))
over(partition by user_id order by order_ range between current row and unbounded following) and
coalesce(order_ not between
max(if(action = 'end', order_, null))
over(partition by user_id order by order_ range between unbounded preceding and 1 preceding)
and min(if(action = 'start', order_, null))
over(partition by user_id order by order_ range between 1 following and unbounded following), true)
, 1, null) as flag
from input
order by 1,2
Edit: It should also take into account weird cases where for instance a 3rd user has end > other > start > other > end > other in that order. The flag should only apply to the 4th item. If you have start > other > start > other > end however, it's unclear if items 2,3,4 or 4 or 2,4 should be flagged. I think it would only flag 4 here
Edit2: This version should flag 2,3,4
if (
order_ > max(if(action = 'start', order_, null))
over(partition by user_id order by order_ range between unbounded preceding and 1 preceding) and
order_ < min(if(action = 'end', order_, null))
over(partition by user_id order by order_ range between current row and unbounded following) and
coalesce(max(if(action = 'start', order_, null))
over(partition by user_id order by order_ range between unbounded preceding and 1 preceding) >
max(if(action = 'end', order_, null))
over(partition by user_id order by order_ range between unbounded preceding and current row),true)
, 1, null) as flag
I have a table like this:
id
status
grade
123
Overall
A
123
Current
B
234
Overall
B
234
Current
D
345
Overall
C
345
Current
A
May I know how can I display how many ids is fitting with the condition:
The grade is sorted like this A > B > C > D > F,
and the Overall grade must be greater than or equal to the Current grade
Is it need to use CASE() to switch the grade to a number first?
e.g. A = 4, B = 3, C = 2, D = 1, F = 0
In the table, there should be 345 is not match the condition. How can I display the tables below:
qty_pass_the_condition
qty_fail_the_condition
total_ids
2
1
3
and\
fail_id
345
Thanks.
As grade is sequential you can do order by desc to make the number. for the first result you can do something like below
select
sum(case when GradeRankO >= GradeRankC then 1 else 0 end) AS
qty_pass_the_condition,
sum(case when GradeRankO < GradeRankC then 1 else 0 end) AS
qty_fail_the_condition,
count(*) AS total_ids
from
(
select * from (
select Id,Status,
Rank() over (partition by Id order by grade desc) GradeRankO
from YourTbale
) as a where Status='Overall'
) as b
inner join
(
select * from (
select Id,Status,
Rank() over (partition by Id order by grade desc) GradeRankC
from YourTbale
) as a where Status='Current'
) as c on b.Id=c.Id
For second one you can do below
select
b.Id fail_id
from
(
select * from (
select Id,Status,
Rank() over (partition by Id order by grade desc) GradeRankO
from Grade
) as a where Status='Overall'
) as b
inner join
(
select * from (
select Id,Status,
Rank() over (partition by Id order by grade desc) GradeRankC
from Grade
) as a where Status='Current'
) as c on b.Id=c.Id
where GradeRankO < GradeRankC
You can use pretty simple conditional aggregation for this, there is no need for window functions.
A Pass is when the row of Overall has grade which is less than or equal to Current, with "less than" being in A-Z order.
Then aggregate again over the whole table, and qty_pass_the_condition is simply the number of non-nulls in Pass. And qty_fail_the_condition is the inverse of that.
SELECT
qty_pass_the_condition = COUNT(t.Pass),
qty_fail_the_condition = COUNT(*) - COUNT(t.Pass),
total_ids = COUNT(*)
FROM (
SELECT
t.id,
Pass = CASE WHEN MIN(CASE WHEN t.status = 'Overall' THEN t.grade END) <=
MIN(CASE WHEN t.status = 'Current' THEN t.grade END)
THEN 1 END
FROM YourTable t
GROUP BY
t.id
) t;
To query the actual failed IDs, simply use a HAVING clause:
SELECT
t.id
FROM YourTable t
GROUP BY
t.id
HAVING MIN(CASE WHEN t.status = 'Overall' THEN t.grade END) >
MIN(CASE WHEN t.status = 'Current' THEN t.grade END);
db<>fiddle
UniqueId
ITEM
DATE
1
A
2022-01-01
2
A
2022-01-02
3
B
2022-01-03
4
B
2022-01-04
5
A
2022-01-05
6
A
2022-01-06
7
B
2022-01-07
8
B
2022-01-08
9
A
2022-01-09
10
A
2022-01-10
11
A
2022-01-11
I have above table where the item is changing from A to B and then B to A (etc).
The the most recent item in the table based on the date is A (the last row).
I need to find the date on which this last item (A) was started to be in effect.
So in this case the item A was in effect from 2022-01-09 onwards (UniqueId 9).
How can I find the UniqueId or the date of item A, where it got changed to be in effect (Row 9)?
Thank you.
with data as (
select *,
last_value(item) over (order by "date") as last_item,
lag(item) over (order by "date") as prev_item
from T
)
select
max(case when item = last_item and item <> prev_item then "date" end) as max_date
from data;
or
with data as (
select *,
case when item <> lag(item) over (order by "date")
and item = last_value(item) over (order by "date")
then 1 end as flag
from T
)
select max("date") as last_transition_date
from data
where flag = 1;
https://dbfiddle.uk/?rdbms=sqlserver_2019&fiddle=bd5f6398c0167d74c26a67fafac5225e
Supposing you need all the data:
with data as (
select *,
case when item <> lag(item) over (order by "date")
and item = last_value(item) over (order by "date")
then 1 end as flag
from T
)
select *,
max(case when flag = 1 then "date" end) over () as last_transition_date
from data;
Getting a flag using a comparison of current item with previous item in time, using LAG() is indeed the way.
But it's absolutely sufficient to get the highest date and highest unique (as both are sorted ascending together) where the obtained flag is 1:
WITH
-- your input
indata(UniqueId,ITEM,DATE) AS (
SELECT 1,'A',DATE '2022-01-01'
UNION ALL SELECT 2,'A',DATE '2022-01-02'
UNION ALL SELECT 3,'B',DATE '2022-01-03'
UNION ALL SELECT 4,'B',DATE '2022-01-04'
UNION ALL SELECT 5,'A',DATE '2022-01-05'
UNION ALL SELECT 6,'A',DATE '2022-01-06'
UNION ALL SELECT 7,'B',DATE '2022-01-07'
UNION ALL SELECT 8,'B',DATE '2022-01-08'
UNION ALL SELECT 9,'A',DATE '2022-01-09'
UNION ALL SELECT 10,'A',DATE '2022-01-10'
UNION ALL SELECT 11,'A',DATE '2022-01-11'
)
-- real query starts here; replace following comma with "WITH"
,
w_change_ind AS (
SELECT
*
, CASE WHEN LAG(item) OVER(ORDER BY date) <> item
THEN 1
ELSE 0
END AS chg_ind
FROM indata
)
SELECT
MAX(uniqueid) AS uqid
, MAX(date) AS dt
FROM w_change_ind
WHERE chg_ind=1
;
-- out uqid | dt
-- out ------+------------
-- out 9 | 2022-01-09
Based on your description, this is one way to do what you want.
select top 1 * from table1
where item ='A'
order by uniqueid desc
If this is not what you want, then you will have to provide additional information.
How can I get this data set in Image 1 to look like the data in Image 2. Basically rather than having each purchase on its own line I want to group by Name and have all that persons purchases on one line. They can buy a max of 5 items and my database is about 30 million lines worth of purchases.
P.S The date order is not important
You can use row_number() and conditional aggregation:
select name,
max(case when seqnum = 1 then item end) as item_1,
max(case when seqnum = 1 then date end) as date_1,
max(case when seqnum = 2 then item end) as item_2,
max(case when seqnum = 2 then date end) as date_2,
max(case when seqnum = 3 then item end) as item_3,
max(case when seqnum = 3 then date end) as date_3
from (select t.*,
row_number() over (partition by name order by date asc) as seqnum
from t
) t
group by name;
You can use PIVOT with row_number as follows:
Select * from
(select t.*,
row_number() over (partition by name order by date_purchased) rn
from your_table t
) PIVOT
(Max(item_purchased), max(date_purchased) For rn in (1,2,3));
I have a set that looks something like this
ID date_IN date_out
1 1/1/18 1/2/18
1 1/3/18 1/4/18
1 1/5/18 1/8/18
2 1/1/18 1/5/18
2 1/7/18 1/9/18
I began by
SELECT ID, date_IN, Date_out, lead(date_out) over ( partition by (ID)
order by ID) as next_out
From table
And get something like this...
ID date_IN date_out next_out
1 1/1/18 1/2/18 1/4/18
1 1/3/18 1/4/18 1/8/18
1 1/5/18 1/8/18 Null
2 1/1/18 1/5/18 1/9/18
2 1/7/18 1/9/18 Null
The problem I’m going to to have is that in my actual data many of the ID’s have A LOT of entries. The goal is to have all of the date_out’s appear on one row per ID....
ID date_IN date_out next_out next_out1 etc. etc.
1 1/1/18 1/2/18 1/4/18 1/8/18 X X
2 1/1/18 1/5/18 1/7/18 X Null Null
Is there a way to loop the lead() through the entire partition, order by ID drop everything but the first row then move on to the next ID?
Here is one approach, which assumes that you only expect to have a maximum of three date pairs per ID. You may assign a row number and then aggregate by ID:
WITH cte AS (
SELECT ID, date_IN, date_out,
ROW_NUMBER() OVER (PARTITION BY ID ORDER BY date_IN) rn
FROM yourTable
)
SELECT
ID,
MAX(CASE WHEN rn = 1 THEN date_IN END) AS date_IN,
MAX(CASE WHEN rn = 1 THEN date_out END) AS date_out,
MAX(CASE WHEN rn = 2 THEN next_IN END) AS next_in_1,
MAX(CASE WHEN rn = 2 THEN date_out END) AS next_out_2,
MAX(CASE WHEN rn = 3 THEN date_IN END) AS next_in_2,
MAX(CASE WHEN rn = 3 THEN date_out END) AS next_out_2
FROM cte
GROUP BY ID
No need to do a loop but use the offset option. Below is lifted from the documentation.
offset
Optional. It is the physical offset from the current row in the table.
If this parameter is omitted, the default is 1.
example; lead(date_out) means next value
lead(date_out, 2) means 2nd row after current row
lead(date_out, 3) 3rd row after current row and so on.
in your code; use below snippet;
lead(date_out) over ( partition by (ID) order by ID) as next_out,
lead(date_out, 2) over ( partition by (ID) order by ID) as next_out2,
lead(date_out, 3) over ( partition by (ID) order by ID) as next_out3
WITH TAB AS(
SELECT 1 ID, CAST('2018/01/01' AS DATE) DATE_IN, CAST('2018/01/02' AS DATE) DATE_OUT FROM DUAL
UNION
SELECT 1, CAST('2018/01/03' AS DATE) , CAST('2018/01/04' AS DATE) FROM DUAL
UNION
SELECT 1, CAST('2018/01/05' AS DATE) , CAST('2018/01/08' AS DATE) FROM DUAL
UNION
SELECT 1, CAST('2018/01/09' AS DATE) , CAST('2018/01/10' AS DATE) FROM DUAL
UNION
SELECT 1, CAST('2018/01/11' AS DATE) , CAST('2018/01/12' AS DATE) FROM DUAL
UNION
SELECT 2, CAST('2018/01/01' AS DATE) , CAST('2018/01/05' AS DATE) FROM DUAL
UNION
SELECT 2, CAST('2018/01/07' AS DATE) , CAST('2018/01/09' AS DATE) FROM DUAL
) --select * from tab;
, LEAF_CALC AS( --CONNECTING THE DATE_OUTS
SELECT
ID
,SYS_CONNECT_BY_PATH(DATE_OUT, '$') HRCHY
, LEVEL LVL
, CONNECT_BY_ISLEAF ISLEAF
FROM TAB
CONNECT BY PRIOR DATE_OUT < DATE_IN
START WITH ID = 1
) --SELECT * FROM LEAF_CALC;
, DATA_SORT AS( --ADDING ALL DATE_OUTS IN 1 ROW
SELECT
P.ID, P.HRCHY
FROM LEAF_CALC P,
(SELECT ID, MAX(LVL) MAXLVL FROM
LEAF_CALC
GROUP BY ID) C
WHERE P.ID = C.ID
AND P.LVL = C.MAXLVL
)--SELECT * FROM DATA_SORT
--SEGREGATING ALL DATES USING REGEXP_SUBSTR
SELECT
ID
, REGEXP_SUBSTR(HRCHY, '[^$]+', 1, 1) DATE_IN
, REGEXP_SUBSTR(HRCHY, '[^$]+', 1, 2) NEXT_OUT
, REGEXP_SUBSTR(HRCHY, '[^$]+', 1, 3) NEXT_OUT2
, COALESCE(REGEXP_SUBSTR(HRCHY, '[^$]+', 1, 4), 'NA') NEXT_OUT3
, COALESCE(REGEXP_SUBSTR(HRCHY, '[^$]+', 1, 5), 'NA') NEXT_OUT4
FROM DATA_SORT;