Related
There was a question asked several years ago with a similar title (Using WITH clause with INSERT statement in POSTGRESQL), but I need to figure out a more complicated situation.
I've used the example from POSTGRESQL for using a "With" statement (https://www.postgresql.org/docs/current/queries-with.html). Let's say I made a pre-computed table and wanted to insert the output from the query into it (top_region, total_sales, product_units, and product_sales), how would I do that?
Precomputed table
INSERT INTO top (top_region, total_sales, product_units, product_sales)
select top_region, total_sales, product_units, product_sales
from #not sure here
WITH regional_sales AS (
SELECT region, SUM(amount) AS total_sales
FROM orders
GROUP BY region
), top_regions AS (
SELECT region
FROM regional_sales
WHERE total_sales > (SELECT SUM(total_sales)/10 FROM regional_sales)
)
SELECT region,
product,
SUM(quantity) AS product_units,
SUM(amount) AS product_sales
FROM orders
WHERE region IN (SELECT region FROM top_regions)
GROUP BY region, product;
My query
WITH sales_data AS (
SELECT s.customer_id, s.product_id, s.quantity, pr.list_price, pr.category_id, c.state_id
FROM order_products s
INNER JOIN Products pr ON (s.product_id = pr.product_id)
INNER JOIN Customer c ON (s.customer_id = c.customer_id)
),
t_20cat AS (
SELECT category_id, ROW_NUMBER() OVER (ORDER BY SUM(list_price*quantity) DESC) AS "rank", SUM(list_price*quantity) AS total_sales FROM sales_data
GROUP BY category_id ORDER BY total_sales DESC LIMIT 20
),
t_20cust AS (
SELECT customer_id, ROW_NUMBER() OVER (ORDER BY SUM(list_price*quantity) DESC) AS "rank", SUM(list_price*quantity) AS total_sales FROM sales_data
GROUP BY customer_id ORDER BY total_sales DESC LIMIT 20
),
t_20cc AS (
SELECT customer_id, t_20cust.rank as customer_rank, category_id, t_20cat.rank as category_rank FROM t_20cat, t_20cust
)
SELECT t_20cc.*, COALESCE(SUM(sales_data.quantity), 0) AS quantity_sold, COALESCE(SUM(sales_data.list_price*sales_data.quantity), 0) AS dollar_value
FROM t_20cc
LEFT JOIN sales_data ON (
t_20cc.customer_id = sales_data.customer_id AND t_20cc.category_id = sales_data.category_id
)
GROUP BY t_20cc.customer_id, t_20cc.customer_rank, t_20cc.category_id, t_20cc.category_rank
INSERT INTO top_20(customer_id, customer_rank, category_id, category_rank
select category_id, category_rank, category_id, category_rank
from t_20cc
)
Move the other SELECT query into its own CTE (region_summary), then reference that in the INSERT statement's SELECT ... FROM clause:
WITH regional_sales AS (
SELECT
region,
SUM(amount) AS total_sales
FROM
orders
GROUP BY
region
),
top_regions AS (
SELECT
region
FROM
regional_sales
WHERE
total_sales > ( SELECT SUM( total_sales ) / 10 FROM regional_sales )
),
region_summary AS (
SELECT
region,
product,
SUM(quantity) AS product_units,
SUM(amount) AS product_sales
FROM
orders
WHERE
region IN ( SELECT region FROM top_regions )
GROUP BY
region,
product
)
INSERT INTO top ( top_region, total_sales, product_units, product_sales )
SELECT
region AS top_region,
product AS total_sales,
product_units,
product_sales
FROM
region_summary;
I'm trying to combine range.
WITH a AS (
select '2017-09-16 07:12:57' as begat,'2017-09-16 11:30:22' as endat
union
select '2017-09-18 17:05:21' ,'2017-09-19 13:18:01'
union
select '2017-09-19 15:34:40' ,'2017-09-22 13:29:37'
union
select '2017-09-22 12:24:16' ,'2017-09-22 13:18:29'
union
select '2017-09-28 09:48:54' ,'2017-09-28 13:39:13'
union
select '2017-09-20 13:52:43' ,'2017-09-20 14:14:43'
), b AS (
SELECT *, lag(endat) OVER (ORDER BY begat) < begat OR NULL AS step
FROM a
)
, c AS (
SELECT *, count(step) OVER (ORDER BY begat) AS grp
FROM b
)
SELECT min(begat), coalesce( max(endat), 'infinity' ) AS range
FROM c
GROUP BY grp
ORDER BY 1
Result
1 "2017-09-16 07:12:57";"2017-09-16 11:30:22"
2 "2017-09-18 17:05:21";"2017-09-19 13:18:01"
3 "2017-09-19 15:34:40";"2017-09-22 13:29:37"
4 "2017-09-22 12:24:16";"2017-09-22 13:18:29"
5 "2017-09-28 09:48:54";"2017-09-28 13:39:13"
positions 3,4 intersect (endata> next begat)
How do I make the union of all the intersections into one large interval
I need result
1 "2017-09-16 07:12:57";"2017-09-16 11:30:22"
2 "2017-09-18 17:05:21";"2017-09-19 13:18:01"
3 "2017-09-19 15:34:40";"2017-09-22 13:29:37"
4 "2017-09-28 09:48:54";"2017-09-28 13:39:13"
Hey I would suggest using the following process :
1- Identify when a row is new, so you give a value of 1 to values that do not overlap (CTE b)
2- Sequence together the rows that have overlaps with others. This way you can see have a common identifier that will allow you to MAX and MIN begat and endat (CTE c)
3- For each sequence, give the MIN of begat and the MAX of endat so you will have your final values
WITH a AS (
select '2017-09-16 07:12:57' as begat,'2017-09-16 11:30:22' as endat
union
select '2017-09-18 17:05:21' ,'2017-09-19 13:18:01'
union
select '2017-09-19 15:34:40' ,'2017-09-22 13:29:37'
union
select '2017-09-22 12:24:16' ,'2017-09-22 13:18:29'
union
select '2017-09-28 09:48:54' ,'2017-09-28 13:39:13'
union
select '2017-09-20 13:52:43' ,'2017-09-20 14:14:43'
)
, b AS (
SELECT
begat
, endat
, (begat > MAX(endat) OVER w IS TRUE)::INT is_new
FROM a
WINDOW w AS (ORDER BY begat ROWS BETWEEN UNBOUNDED PRECEDING AND 1 PRECEDING)
)
, c AS (
SELECT
begat
, endat
, SUM((is_new)) OVER (ORDER BY begat) seq
FROM b
)
SELECT
MIN(begat) beg_at
, MAX(endat) end_at
FROM c
GROUP BY seq
EDITED
If you need speed you can use a psql function:
create or replace function append_ranges_in_a() returns setof a
language plpgsql
as
$BODY$
declare
v_current a%rowtype;
v_new a%rowtype;
v_first boolean:=true;
begin
for v_current in select begat, endat from a order by begat, endat
loop
if v_first then
v_first := false;
v_new.begat := v_current.begat;
elsif v_new.endat < v_current.begat then
return next v_new;
v_new.begat := v_current.begat;
end if;
v_new.endat := greatest(v_current.endat,v_new.endat);
end loop;
return next v_new;
return;
end;
$BODY$;
select * from append_ranges_in_a()
I test it with ~ 400000 rows:
delete from a;
insert into a (begat, endat)
select time::text, (time+'1 day'::interval)::text
from (select t+(round(random()*23.0)||' hours')::interval as time
from generate_series('1401-01-01'::timestamp,'2018-08-21'::timestamp,'1 day'::interval) t
) t;
select count(*) from a;
select * from append_ranges_in_a() offset 100000 limit 10
and it is twice fast as O(n^2) pure SQL version.
OLD slow solution:
You can use a recursive WITH query https://www.postgresql.org/docs/current/static/queries-with.html to construct the result row by row.
I create the table
The first row is the candidate first row (ending where ending), but the row is not "ready"
Then I look at the next row (step) and if it is not intersecting I add a ready row,
Also I add a not ready row with the current (last) observed range
When I do not have more rows I calculate the last row
I retain ready rows and the last row
Here is the code
CREATE TABLE a as
select '2017-09-16 07:12:57' as begat,'2017-09-16 11:30:22' as endat
union
select '2017-09-18 17:05:21' ,'2017-09-19 13:18:01'
union
select '2017-09-19 15:34:40' ,'2017-09-22 13:29:37'
union
select '2017-09-22 12:24:16' ,'2017-09-22 13:18:29'
union
select '2017-09-28 09:48:54' ,'2017-09-28 13:39:13'
union
select '2017-09-20 13:52:43' ,'2017-09-20 14:14:43';
WITH RECURSIVE t(begat, endat, ready, step) AS (
select * from (
select *,false,1 from a order by begat, endat limit 1) a
UNION ALL
SELECT new_rows.*
FROM (SELECT * FROM t ORDER BY begat DESC limit 1) t,
lateral (SELECT * FROM a ORDER BY begat, endat OFFSET step LIMIT 1) a,
lateral (
SELECT t.begat, t.endat, true as ready, step WHERE t.endat < a.begat
UNION SELECT CASE WHEN t.endat < a.begat THEN a.begat ELSE t.begat END, greatest(a.endat, t.endat), false, step+1
) new_rows
)
select begat, endat
from (
select begat, endat, ready, row_number() over (order by begat desc, endat desc)=1 is_last
from t
order by begat, endat) t
where ready or is_last;
i using range type
https://www.postgresql.org/docs/9.3/static/rangetypes.html
WITH tmp AS (
-- preparation range type
select begat, coalesce( endat, 'infinity' ) as endAt, tsrange( begat, coalesce( endat, 'infinity' ) ) as rg
from (
select '2017-09-11 17:13:03'::timestamp as begat ,'2017-09-12 12:24:09'::timestamp as endat union
select '2017-09-19 15:34:40','2017-09-20 11:04:45' union
select '2017-09-20 08:32:00','2017-09-22 13:28:37' union
select '2017-09-20 13:52:43','2017-09-20 14:14:43' union
select '2017-09-21 12:24:16','2017-09-21 13:28:29' union
select '2017-09-22 12:24:16','2017-09-22 13:28:29' union
select '2017-09-22 12:34:16','2017-09-23 13:28:29' union
select '2017-09-22 12:25:16','2017-09-24 13:28:29' union
select '2017-09-28 09:48:54','2017-09-28 13:39:13' union
select '2017-09-28 14:22:16','2017-09-28 15:52:15' union
select '2017-10-05 12:17:45','2017-10-06 12:35:38' union
select '2017-10-06 16:20:44','2017-10-07 10:11:09' union
select '2017-10-07 20:38:32','2017-10-09 14:42:29' union
select '2017-10-12 18:22:14','2017-10-12 20:52:45'
) a
),a as (
-- group intersecting range
select l.*
from tmp l left join tmp r on l.begAt > r.begAt and r.rg #> l.rg
where r.begAt is null
),
b AS (
SELECT *, lag(endat) OVER (ORDER BY begat) < begat OR NULL AS step
FROM a
)
, c AS (
SELECT *, count(step) OVER (ORDER BY begat) AS grp
FROM b
)
SELECT min(begat), coalesce( max(endat), 'infinity' ) AS range
FROM c
GROUP BY grp
ORDER BY 1
I need to get a tree of related nodes given a certain node, but not necessary top node. I've got a solution using two CTEs, since I am struggling to squeeze it all into one CTE :). Might somebody have a sleek solution to avoid using two CTEs? Here is some code that I was playing with:
DECLARE #temp AS TABLE (ID INT, ParentID INT)
INSERT INTO #temp
SELECT 1 ID, NULL AS ParentID
UNION ALL
SELECT 2, 1
UNION ALL
SELECT 3, 2
UNION ALL
SELECT 4, 3
UNION ALL
SELECT 5, 4
UNION ALL
SELECT 6, NULL
UNION ALL
SELECT 7, 6
UNION ALL
SELECT 8, 7
DECLARE #startNode INT = 4
;WITH TheTree (ID,ParentID)
AS (
SELECT ID, ParentID
FROM #temp
WHERE ID = #startNode
UNION ALL
SELECT t.id, t.ParentID
FROM #temp t
JOIN TheTree tr ON t.ParentID = tr.ID
)
SELECT * FROM TheTree
;WITH Up(ID,ParentID)
AS (
SELECT t.id, t.ParentID
FROM #temp t
WHERE t.ID = #startNode
UNION ALL
SELECT t.id, t.ParentID
FROM #temp t
JOIN Up c ON t.id = c.ParentID
)
--SELECT * FROM Up
,TheTree (ID,ParentID)
AS (
SELECT ID, ParentID
FROM Up
WHERE ParentID is null
UNION ALL
SELECT t.id, t.ParentID
FROM #temp t
JOIN TheTree tr ON t.ParentID = tr.ID
)
SELECT * FROM TheTree
thanks
Meh. This avoids using two CTEs, but the result is a brute force kludge that hardly qualifies as "sleek" as it won’t be efficient if your table is at all sizeable. It will:
Recursively build all possible hierarchies
As you build them, flag the target NodeId as you find it
Return only the targeted tree
I threw in column “TreeNumber” on the off-chance the TargetId appears in multiple hierarchies, or if you’d ever have multiple values to check in one pass. “Depth” was added to make the output a bit more legible.
A more complex solution like #John’s might do, and more and subtler tricks could be done with more detailed table sturctures.
DECLARE #startNode INT = 4
;WITH cteAllTrees (TreeNumber, Depth, ID, ParentID, ContainsTarget)
AS (
SELECT
row_number() over (order by ID) TreeNumber
,1
,ID
,ParentID
,case
when ID = #startNode then 1
else 0
end ContainsTarget
FROM #temp
WHERE ParentId is null
UNION ALL
SELECT
tr.TreeNumber
,tr.Depth + 1
,t.id
,t.ParentID
,case
when tr.ContainsTarget = 1 then 1
when t.ID = #startNode then 1
else 0
end ContainsTarget
FROM #temp t
INNER JOIN cteAllTrees tr
ON t.ParentID = tr.ID
)
SELECT
TreeNumber
,Depth
,ID
,ParentId
from cteAllTrees
where TreeNumber in (select TreeNumber from cteAllTrees where ContainsTarget = 1)
order by
TreeNumber
,Depth
,ID
Here is a technique where you can select the entire hierarchy, a specific node with all its children, and even a filtered list and how they roll.
Note: See the comments next to the DECLAREs
Declare #YourTable table (id int,pt int,name varchar(50))
Insert into #YourTable values
(1,null,'1'),(2,1,'2'),(3,1,'3'),(4,2,'4'),(5,2,'5'),(6,3,'6'),(7,null,'7'),(8,7,'8')
Declare #Top int = null --<< Sets top of Hier Try 2
Declare #Nest varchar(25) = '|-----' --<< Optional: Added for readability
Declare #Filter varchar(25) = '' --<< Empty for All or try 4,6
;with cteP as (
Select Seq = cast(1000+Row_Number() over (Order by name) as varchar(500))
,ID
,pt
,Lvl=1
,name
From #YourTable
Where IsNull(#Top,-1) = case when #Top is null then isnull(pt,-1) else ID end
Union All
Select Seq = cast(concat(p.Seq,'.',1000+Row_Number() over (Order by r.name)) as varchar(500))
,r.ID
,r.pt
,p.Lvl+1
,r.name
From #YourTable r
Join cteP p on r.pt = p.ID)
,cteR1 as (Select *,R1=Row_Number() over (Order By Seq) From cteP)
,cteR2 as (Select A.Seq,A.ID,R2=Max(B.R1) From cteR1 A Join cteR1 B on (B.Seq like A.Seq+'%') Group By A.Seq,A.ID )
Select Distinct
A.R1
,B.R2
,A.ID
,A.pt
,A.Lvl
,name = Replicate(#Nest,A.Lvl-1) + A.name
From cteR1 A
Join cteR2 B on A.ID=B.ID
Join (Select R1 From cteR1 where IIF(#Filter='',1,0)+CharIndex(concat(',',ID,','),concat(',',#Filter+','))>0) F on F.R1 between A.R1 and B.R2
Order By A.R1
I need to write a T-SQL group by query for a table with multiple dates and seq columns:
DROP TABLE #temp
CREATE TABLE #temp(
id char(1),
dt DateTime,
seq int)
Insert into #temp values('A','2015-03-31 10:00:00',1)
Insert into #temp values('A','2015-08-31 10:00:00',2)
Insert into #temp values('A','2015-03-31 10:00:00',5)
Insert into #temp values('B','2015-09-01 10:00:00',1)
Insert into #temp values('B','2015-09-01 10:00:00',2)
I want the results to contains only the items A,B with their latest date and the corresponding seq number, like:
id MaxDate CorrespondentSeq
A 2015-08-31 10:00:00.000 2
B 2015-09-01 10:00:00.000 2
I am trying with (the obviously wrong!):
select id, max(dt) as MaxDate, max(seq) as CorrespondentSeq
from #temp
group by id
which returns:
id MaxDate CorrespondentSeq
A 2015-08-31 10:00:00.000 5 <-- 5 is wrong
B 2015-09-01 10:00:00.000 2
How can I achieve that?
EDIT
The dt datetime column has duplicated values (exactly same date!)
I am using SQL Server 2005
You can use a ranking subselect to get only the highest ranked entries for an id:
select id, dt, seq
from (
select id, dt, seq, rank() over (partition by id order by dt desc, seq desc) as r
from #temp
) ranked
where r=1;
SELECT ID, DT, SEQ
FROM (
SELECT ID, DT, SEQ, Row_Number()
OVER (PARTITION BY id ORDER BY dt DESC, seq DESC) AS row_number
FROM temp
) cte
WHERE row_number = 1;
Demo : http://www.sqlfiddle.com/#!3/3e3d5/5
With trial and errors maybe I have found a solution, but I'm not completely sure this is correct:
select A.id, B.dt, max(B.seq)
from (select id, max(dt) as maxDt
from #temp
group by id) as A
inner join #temp as B on A.id = B.id AND A.maxDt = B.dt
group by A.id, B.dt
Select id, dt, seq
From #temp t
where dt = (Select Max(dt) from #temp
Where id = t.Id)
If there are duplicate rows, then you also need to specify what the query processor should use to determine which of the duplicates to return. Say you want the lowest value of seq,
Then you could write:
Select id, dt, seq
From #temp t
where dt = (Select Max(dt) from #temp
Where id = t.Id)
and seq = (Select Min(Seq) from #temp
where id = t.Id
and dt = t.dt)
I want these two SQL statements in one SQL command. Help :)
Statement #1:
SELECT SUM(nrofitems) as totItems
FROM tblSets
WHERE moduleexamID = 20
Statement #2:
SELECT TOP (cast(totItems as int)) questions
FROM tblQuestions
WHERE moduleexamID = 20
ORDER BY NEWID()
create table #tblSets (
moduleexamID int,
moduleId int,
nrofitems int
)
go
create table #tblQuestions (
moduleexamID int,
body varchar(1024)
)
go
insert into #tblQuestions values
(20,'aaaaaa'),
(20,'bbbbbb'),
(20,'cccccc'),
(20,'dddddd'),
(21,'eeeeee'),
(21,'ffffff'),
(20,'gggggg')
go
insert into #tblSets values
(20,1,1),
(20,2,2),
(21,1,1),
(22,1,3)
go
select top (
select sum(s.nrofitems)
from #tblSets s
where s.moduleexamID=20
) *, newid() as id
from #tblQuestions q
where q.moduleexamID=20
order by id
go
Just use ROW_NUMBER().
Something like this:
SELECT * FROM
(
SELECT tblQuestions.*,
ROW_NUMBER() OVER (ORDER BY NEWID()) as RN
FROM tblQuestions
WHERE moduleexamID = 20
) as T1
WHERE RN<=
ISNULL(
(SELECT SUM(nrofitems) as totItems
FROM tblSets
WHERE moduleexamID = 20
),0);
You may try following also:
;WITH a AS (
SELECT moduleexamID, SUM(nrofitems) as totItems
FROM tblSets
GROUP BY moduleexamID
)
SELECT b.questions
FROM a
CROSS APPLY (
SELECT TOP (cast(a.totItems as int)) questions
FROM tblQuestions
WHERE moduleexamID = a.moduleexamID
ORDER BY CHECKSUM(NEWID())
) b
WHERE a.moduleexamID = 20