Suppose I have a lookup table mylookup(lkid, lkname, ...)
Then I have other 2 tables:
mytab(id, parentid, name, lookname, ...)
yourtab(id, parentid, ...) --id, parentid coming from mytab
Then I have a update try to get lkid and it's parent lkid, the sql like:
update yourtab set columnx =
case
when (select lkid from mylookup a join mytab b on a.lkname = b.lkname where b.id = c.parentid ) > 3
and (select lkid from mylookup a join mytab b on a.lkname = b.lkname where b.id = c.parentid ) >
(select lkid from mylookup where lkname = c.lkname )
then 1
else 0
end
from yourtab c
this sql performance is not good. (select lkid from mylookup a join mytab b on a.lkname = b.lkname where b.id = c.parentid ) is called 2 times for each row in yourtb.
how to rewrite sql to improve performance for this case?
Maybe something like this..
It's untested would need a SQLFiddle with some sample data to really understand inner workings.
Update yourtab set columnx =
case when not null (
SELECT lkid
FROM mylookup a
INNER join mytab b
on a.lkname = b.lkname
WHERE b.id = c.parentid
and a.lkid > 3
and a.lkname > c.lkname) then 1
else 0
end
FROM yourtab c
update yourtab
set columnx = 1
FROM yourtab c
join mylookup a
on a.lkname > c.lkname
and a.lkid > 3
join mytab b
on a.lkname = b.lkname
and b.id = c.parentid
where columnx <> 1
update yourtab
set columnx = 0
FROM yourtab c
left join mylookup a
on a.lkname > c.lkname
and a.lkid > 3
left join mytab b
on a.lkname = b.lkname
and b.id = c.parentid
where columnx <> 0
and ( a.lkname is null or b.id is null )
Two statements but is saves taking a lock if the value is already correct
Related
I am trying to pull data out and chuck it into a Stimulsoft report. The problem I am having is that I need it to output to two columns. I also need every "manager" record to show even if the count assigned to said record is NULL.
This is what i have at the moment:
DECLARE #ManagerCount INT = (( SELECT Count(*) FROM AM WHERE dbo.AM.AMCurrent = 1 AND dbo.AM.OmitInReport = 0 ) + 1) / 2
DECLARE #tmp_AM1 TABLE (AMID INT, AMName NVARCHAR(100), ID INT)
INSERT INTO #tmp_AM1 SELECT AMID, AMName, row_number() over (order by AMID ) FROM AM
WHERE dbo.AM.AMCurrent = 1 AND dbo.AM.OmitInReport = 0
SELECT * FROM (
SELECT ta.id AS id1, ta.AMName AS ManagerName1, COUNT(*) AS ManagerCount1 FROM #tmp_AM1 tA INNER JOIN Job J ON tA.AMID = j.AMID
WHERE ta.ID BETWEEN 1 AND #ManagerCount AND j.jobStatusID != 5
GROUP BY ta.ID, ta.AMName
) a
LEFT JOIN
(
SELECT ta.id AS id2,ta.AMName AS ManagerName2, COUNT(*) AS ManagerCount2 FROM #tmp_AM1 tA INNER JOIN Job J ON tA.AMID = j.AMID
WHERE ta.ID > #ManagerCount AND j.jobStatusID != 5
GROUP BY ta.AMName, ta.ID
) b ON a.id1 + #ManagerCount = b.id2
Which ends up returning something like:
There are 18 managers so 9 per column, but this code doesn't show them all since anything that doesn't have a count in the first left join, won't show, and therefore the same row in column 2 doesn't show.
Results of SELECT * FROM #tmp_AM1:
DECLARE #tmp_AM1 TABLE (AMID INT, AMName NVARCHAR(100), ID INT)
INSERT INTO #tmp_AM1 SELECT AMID, AMName, row_number() over (order by AMID ) FROM AM
WHERE dbo.AM.AMCurrent = 1 AND dbo.AM.OmitInReport = 0
SELECT * FROM (
SELECT ta.id AS id1, ta.AMName AS ManagerName1, COUNT(*) AS ManagerCount1 FROM #tmp_AM1 tA INNER JOIN Job J ON tA.AMID = j.AMID
WHERE ta.ID BETWEEN 1 AND #ManagerCount AND j.jobStatusID != 5
GROUP BY ta.ID, ta.AMName
) a
LEFT OUTER JOIN
(
SELECT ta.id AS id2,ta.AMName AS ManagerName2, COUNT(*) AS ManagerCount2 FROM #tmp_AM1 tA INNER JOIN Job J ON tA.AMID = j.AMID
WHERE ta.ID > #ManagerCount AND j.jobStatusID != 5
GROUP BY ta.AMName, ta.ID
) b ON a.id1 + #ManagerCount = b.id2 where ManagerName2 IS Null and ManagerCount2 IS NULL
just you want to use LEFT OUTER JOIN for select row even there is have any null values.,
Since the two subqueries are pretty much identical, except the where-statement, I would consiter rewriting it into one single query. I'm not sure why you need the same columns outputed into different columns in the result, but something like this might work:
WITH cte AS (
SELECT
ta.id AS id
,ta.AMName AS ManagerName
,COUNT(*) AS ManagerCount
,CASE WHEN ta.ID BETWEEN 1 AND #ManagerCount THEN 0 ELSE 1 END AS something
FROM
#tmp_AM1 tA
INNER JOIN Job J ON tA.AMID = j.AMID
WHERE
j.jobStatusID != 5
GROUP BY
ta.ID
,ta.AMName
,CASE WHEN ta.ID BETWEEN 1 AND #ManagerCount THEN 0 ELSE 1 END
)
SELECT
CASE WHEN something = 0 THEN cte.id ELSE null END AS id1
,CASE WHEN something = 0 THEN cte.ManagerName ELSE null END AS ManagerName1
,CASE WHEN something = 0 THEN cte.ManagerCount ELSE null END AS ManagerCount1
,CASE WHEN something = 1 THEN cte.id ELSE null END AS id2
,CASE WHEN something = 1 THEN cte.ManagerName ELSE null END AS ManagerName2
,CASE WHEN something = 1 THEN cte.ManagerCount ELSE null END AS ManagerCount2
FROM
cte
Probably not the best approach but i got the correct output using:
DECLARE #ManagerCount INT = (( SELECT Count(*) FROM AM WHERE dbo.AM.AMCurrent = 1 AND dbo.AM.OmitInReport = 0 ) + 1) / 2
DECLARE #tmp_AM1 TABLE (AMID INT, AMName NVARCHAR(100), ID INT)
INSERT INTO #tmp_AM1 SELECT AMID, AMName, row_number() over (order by AMID ) FROM AM
WHERE dbo.AM.AMCurrent = 1 AND dbo.AM.OmitInReport = 0
ORDER By AMName
SELECT ManagerName1, ManagerName2, ManagerCount1, ManagerCount2 FROM (
SELECT AMID, ta.id AS id1, ta.AMName AS ManagerName1 FROM #tmp_AM1 tA
WHERE (ta.ID BETWEEN 1 AND #ManagerCount)
) a
LEFT JOIN
(
SELECT AMID, ISNULL(COUNT(*), 0) AS ManagerCount1 FROM Job j
INNER JOIN tblJobOutcome jO ON j.JobOutcomeID = jo.JobOutcomeID AND jO.JobOutcomeID != 5
GROUP BY AMID
) a1 ON a.AMID = a1.AMID
LEFT JOIN
(
SELECT AMID, ta.id AS id2, ta.AMName AS ManagerName2 FROM #tmp_AM1 tA
WHERE (ta.ID > #ManagerCount)
) b ON a.id1 + #ManagerCount = b.id2
LEFT JOIN
(
SELECT AMID, ISNULL(COUNT(*), 0) AS ManagerCount2 FROM Job j
INNER JOIN tblJobOutcome jO ON j.JobOutcomeID = jo.JobOutcomeID AND jO.JobOutcomeID != 5
GROUP BY AMID
) b1 ON b.AMID = b1.AMID
Gives me the correct output in two columns.
gives me this:
I'm quite stuck with this problem for sometime now..
How do I sort column A depending on the contents of Column B?
I have this sample:
ID count columnA ColumnB
-----------------------------------
12 1 A B
13 2 C D
14 3 B C
I want to sort it like this:
ID count ColumnA ColumnB
-----------------------------------
12 1 A B
14 3 B C
13 2 C D
so I need to sort the rows if the previous row of ColumnB = the next row of ColumnA
I'm thinking a loop? but can't quite imagine how it will work...
I was thinking it will go like this (maybe)
SELECT
a.ID, a.ColumnA, a.ColumnB
FROM
TableA WITH a (NOLOCK)
LEFT JOIN
TableA b WITH (NOLOCK) ON a.ID = b.ID AND a.counts = b.counts
WHERE
a.columnB = b.ColumnA
the above code isn't working though and I was thinking more on the lines of...
DECLARE #counts int = 1
DECLARE #done int = 0
--WHILE #done = 0
BEGIN
SELECT
a.ID, a.ColumnA, a.ColumnB
FROM
TableA WITH a (NOLOCK)
LEFT JOIN
TableA b WITH (NOLOCK) ON a.ID = b.ID AND a.counts = #counts
WHERE
a.columnB = b.ColumnA
set #count = #count +1
END
If this was a C code, would be easier for me but T-SQL's syntax is making it a bit harder for a noobie like me.
Any help is greatly appreciated!
Edit: sample code
drop table tablea
create table TableA(
id int,
colA varchar(10),
colb varchar(10),
counts int
)
insert INTO TableA
(id, cola, colb, counts)
select 12, 'Bad', 'Cat', 3
insert INTO TableA
(id, cola, colb, counts)
select 13, 'Apple', 'Bad', 1
insert INTO TableA
(id, cola, colb, counts)
select 14, 'Cat', 'Dog', 2
select * FROM TableA
SELECT a.ID, a.ColA, a.ColB
FROM TableA a WITH (NOLOCK)
LEFT JOIN TableA b WITH (NOLOCK)
ON a.ID = b.ID
Where a.colB = b.ColA
ORDER BY a.ColA ASC
you just need to add ORDER BY clause
-- SELECT a.ID, a.ColumnA, a.ColumnB
-- FROM TableA WITH a (NOLOCK)
-- LEFT JOIN TableA b WITH (NOLOCK)
-- ON a.ID = b.ID
-- and a.counts = b.counts
-- Where a.columnB = b.ColumnA
ORDER BY a.ColumnA ASC
This is all you need. Sometimes you have to think simple
select * from table A
order by columnA asc
Is there such thing like conditional join:
SELECT *
FROM TABLE1 A
IF (a=='TABLE2') THEN INNER JOIN TABLE2 B ON A.item_id=B.id
ELSE IF (a=='TABLE3') THEN INNER JOIN TABLE3 C ON A.item_id=C.id
While a is a field in TABLE1.
I like to use this in stored procedures without using dynamic sql (without writing query as string and EXEC(#query)).
EDIT: I can't write:
IF (a=='TABLE2) THEN queryA
ELSE IF (a=='TABLE3') THEN queryB
Because a is a field of TABLE1.
EDIT: Modified answer based on comment below:
You could try to get clever with some left joins. This will return more columns, so you'd probably want to be more discriminating than just SELECT *.
SELECT *
FROM TABLE1 A
LEFT JOIN TABLE2 B
ON A.item_id = B.id
AND A.a = 'TABLE2'
LEFT JOIN TABLE3 C
ON A.item_id = C.id
AND A.a = 'TABLE3'
WHERE (B.id IS NOT NULL AND A.a = 'TABLE2')
OR (C.id IS NOT NULL AND A.a = 'TABLE3')
Updated the query as requried:
SELECT * FROM
(
SELECT *
FROM TABLE1 A INNER JOIN TABLE2 B
ON A.a='TABLE2' --This will eleminate the table rows if the value of A.a is not 'TABLE2'
AND A.item_id=B.id) A,
(SELECT * FROM
INNER JOIN TABLE3 C
ON A.a='TABLE3' --This will eleminate the table rows if the value of A.a is not 'TABLE3'
AND A.item_id=C.id
) B
) a
I have a problem with recursive CTE query
Let's say that I have that category tree (Category table)
In my CTE query, I search for all children of the 1 category:
(that query works fine)
with mq as
(
select c.Id as parent, c.Id as child
from dbo.Category c
where c.Id = 1
union all
select q.child, c.Id
from mq q
inner join dbo.Category c on q.child = c.IdParentCategory
)
The output
Then, I want to get that Category ID, wchih doesn't have a child: categories 9,10,12,14,15
with mq as
(
select c.Id as parent, c.Id as child
from dbo.Category c
where c.Id = 1
union all
select q.child, c.Id
from mq q
inner join dbo.Category c on q.child = c.IdParentCategory
where child in
(
select c1.Id
from dbo.Category c1
where not exists(select c2.Id
from dbo.Category c2
where c2.Id = c1.IdParentCategory)
)
)
but the output is wrong:
why ? Any ideas will be helpful !
if I separate the query from CTE, everything is OK
declare #tab table
(parent int, child int);
insert into #tab
select * from mq
delete from #tab
where child in (
select c1.parent
from #tab c1
where not exists(select c2.parent from #tab c2 where c2.parent = c1.child)
)
with mq as
(
select c.Id as parent, c.Id as child
from dbo.Category c
where c.Id = 1
union all
select q.child, c.Id
from mq q
inner join dbo.Category c on q.child = c.IdParentCategory
)
select child from mq where child not in (select parent from mq)
Would seem to give the output you want - in fact your description of the problem almost took this form.
Pseudo-code as follows:
update TABLEA a, TABLEB b
set a.addr = 'aaa',
b.name = 'bbb'
from TABLEA a, TABLEB b
where a.id = b.id and a.id = 1
You can only UPDATE one table. So, you can change your SQL to the following:
UPDATE tableA a
SET a.addr = 'aaa'
WHERE exists
(SELECT b.id
FROM tableB b
WHERE b.id = a.id)