I am trying to use a correlated subquery, but I am trying to limit it to the "best" record. When I use SQL very similiar to what follows, I get two rows per BigTable.identifier, and I wish to have only one. In the 'UNION' statement, the second half is more desirable than the first half. However, sometimes the first half will be needed. Any ideas? Here's the code:
select
BigTable.identifier,
Correlated.ID,
Correlated.Effective_Date,
Correlated.Period_Number
from
BigTable
inner join
(
select
TOP 2147483647
Table3.identifier,
Table4.Effective_Date,
Table4.Period_Number
from
Table3
inner join Table4 on Table3.matching_key = Table4.matching_key
where
Table4.Period_Number = 0
order by Table4.Effective_Date desc
UNION
select
TOP 2147483647
Table3.Identifer,
Table4.Effective_Date,
Table4.Period_Number
from
Table3
inner join Table5 on Table3.matching-key = Table5.matching-key
inner join Table4 on Table5.key1 = Table4.key1 and
Table5.key2 = Table4.key2
where
Table4.period_number = 1
order by Table4.Effective_Date desc
) as Correlated
on BigTable.identifier = Correlated.identifier
If each sub-query in that UNION had some condition which EXCLUDED the row if it was less-preferred, you would never see the less-preferred rows in the UNION.
So, if each were to have a NOT EXISTS (.... a better row in the other side of the union ....), you would eliminate less-preferred rows at the root.
I'm not clear on how you want to use effective date. Assuming you mean that you prefer Period=1 but if the Effective date is less you prefer Period=0, then something like this might work.
select
BigTable.identifier,
Correlated.ID,
Correlated.Effective_Date,
Correlated.Period_Number
from
BigTable
inner join
(
select
TOP 2147483647
Table3.identifier,
Table4.Effective_Date,
Table4.Period_Number
from
Table3
inner join Table4 on Table3.matching_key = Table4.matching_key
where
Table4.Period_Number = 0
AND NOT EXISTS
(select 1
from Table5 T5 inner join Table4 T4
on T5.key1 = T4.key1 and T5.key2 = T4.key2
where Table3.matching-key = T5.matching-key
and (T4.Effective_Date >= Table4.Effective_Date and T4.Period_Number = 1)
)
order by Table4.Effective_Date desc
UNION
select
TOP 2147483647
Table3.Identifer,
Table4.Effective_Date,
Table4.Period_Number
from
Table3
inner join Table5 on Table3.matching-key = Table5.matching-key
inner join Table4 on Table5.key1 = Table4.key1 and
Table5.key2 = Table4.key2
where
Table4.period_number = 1
AND NOT EXISTS
(select 1
from Table4 T4
where Table3.matching-key = T4.matching-key
and (T4.Period_Number > 0)
and (T4.Effective_Date > Table4.Effective_Date and T4.Period_Number = 0)
)
order by Table4.Effective_Date desc
) as Correlated
on BigTable.identifier = Correlated.identifier
Related
I have a table table1. with columns sn, rt and type
I want to get rows with different rt (i.e rt = 1,2,3) column conditions
(SELECT *
FROM table1
WHERE sn = 'testing' AND rt = 1 AND type = 'pump'
ORDER BY id DESC
LIMIT 1)
UNION
(SELECT *
FROM table1
WHERE sn = 'testing' AND rt = 2 AND type = 'pump'
ORDER BY id DESC
LIMIT 1)
UNION
(SELECT *
FROM table1
WHERE sn = 'testing' AND rt = 3 AND type = 'pump'
ORDER BY id DESC
LIMIT 1)
Currently i am trying the above.
Which is the effective way to get the rows
Use ROW_NUMBER() window function:
SELECT t.*
FROM (
SELECT *, ROW_NUMBER() OVER (PARTITION BY rt ORDER BY id DESC) rn
FROM table1
WHERE sn = 'testing' AND type = 'pump' AND rt IN (1, 2, 3)
) t
WHERE t.rn = 1
You can omit AND rt IN (1, 2, 3) if 1, 2 and 3 are the only possible values for rt.
You'll want to do JOINs between the table and itself, then specify the conditions in the singular WHERE clause. I've done a rough sketch of it below:
SELECT t1.*, t2.*, t3.*
FROM table1 AS t1
JOIN table1 AS t2
ON t1.sn = t2.sn AND t1.type = t2.type -- If you've got more reasonable connections between your datapoints, use them here
JOIN table1 AS t3
ON t.sn = t3.sn AND t1.type = t3.type
WHERE t1.rt = 1 AND t2.rt = 2 AND t3.rt = 3
ORDER BY t1.id DESC
LIMIT 1
Depending on what other requirements you have, you may have to tweak some parts of that. If you want results when t1 has a value but t2 or t3 doesn't you can use a LEFT JOIN instead.
Below sample query is a part of my main query. I found SORT operator in below query is consuming 30% of the cost.
To avoid SORT, there is need of creation of Indexes. Is there any other way to optimize this code.
SELECT TOP 1 CONVERT( DATE, T_Date) AS T_Date
FROM TableA
WHERE ID = r.ID
AND Status = 3
AND TableA_ID >ISNULL((
SELECT TOP 1 TableA_ID
FROM TableA
WHERE ID = r.ID
AND Status <> 3
ORDER BY T_Date DESC
), 0)
ORDER BY T_Date ASC
Looks like you can use not exists rather than the sorts. I think you'll probably get a better performance boost by use a CTE or derived table instead of the a scalar subquery.
select *
from r ... left outer join
(
select ID, min(t_date) as min_date from TableA t1
where status = 3 and not exists (
select 1 from TableA t2
where t2.ID = t1.ID
and t2.status <> 3 and t2.t_date > t1.t_date
)
group by ID
) as md on md.ID = r.ID ...
or
select *
from r ... left outer join
(
select t1.ID, min(t1.t_date) as min_date
from TableA t1 left outer join TableA t2
on t2.ID = t1.ID and t2.status <> 3
where t1.status = 3 and t1.t_date < t2.t_date
group by t1.ID
having count(t2.ID) = 0
) as md on md.ID = r.ID ...
It also appears that you're relying on an identity column but it's not clear what those values mean. I'm basically ignoring it and using the date column instead.
Try this:
SELECT TOP 1 CONVERT( DATE, T_Date) AS T_Date
FROM TableA a1
LEFT JOIN (
SELECT ID, MAX(TableA_ID) AS MaxAID
FROM TableA
WHERE Status <> 3
GROUP BY ID
) a2 ON a2.ID = a1.ID AND a1.TableA_ID > coalesce(a2.MAXAID,0)
WHERE a1.ID = r.ID AND a1.Status = 3
ORDER BY T_Date ASC
The use of TOP 1 in combination with the unexplained r alias concern me. There's almost certainly a MUCH better way to get this data into your results that doesn't involve doing this in a sub query (unless this is for an APPLY operation).
Can someone explain this to me?
I have two huge tables on which I wish to LEFT JOIN, but I think it would be more efficient to filter at the 'on' rather than the 'where'.
Select * from Table1 t1 left join Table2 t2 on t1.Id = t2.Id and t1.Enabled = 1
Rather than
Select * from Table1 t1 left join Table2 t2 on t1.Id = t2.Id where t1.Enabled = 1
The above results are not the same. No matter what I put in as filtering, it stays unaffected:
Select * from Table1 t1 left join Table2 t2 on t1.Id = t2.Id and t1.Enabled = 1
yields exactly the same results as
Select * from Table1 t1 left join Table2 t2 on t1.Id = t2.Id and t1.Enabled = 0
Also, Table1 might only have a few records where Enabled = 1, so it would be inefficient to join millions of records first and then filter to find only 10.
I can do a sub select first, but I somehow I feel it is not the right way:
Select * from (Select * from Table1 where Enabled = 1) a left join Table2 t2 on a.Id = t2.Id
Types of Joins
Left outer join All rows from the first-named table (the "left"
table, which appears leftmost in the JOIN clause) are included.
Unmatched rows in the right table do not appear.
When you move the condition to the where then t1.Enabled = 1 is applied
Do you have actual performance problems with the first query?
Table1:
Shop
Manager
Date
Table2:
Shop
Date
Sales
I need to get Table2 with Manager field from Table1. I did the following trick:
select
t1.[Shop]
,t1.[Date]
,t1.[Sum]
,t2.[Manager]
from t1
left join t2
on t1.[Shop] = t2.[Shop]
and t2.[Date] = (select max(t2.[Date]) from t2
where t2.[Shop] = t1.[Shop]
and t2.[Date] < t1.[Date])
It works, but subquerying is very slow, so I wonder if there is more elegant and fast way to do so?
Some sample data to play around: http://pastebin.com/uLN6x5JE
may seem like a round about way but join on a single condition is typically faster
select t12.[Shop], t12.[Date], t12.[Sum]
, t12.[Manager]
from
( select t1.[Shop], t1.[Date], t1.[Sum]
, t2.[Manager]
, row_number() over (partition by t2.[Shop] order by t2.[Date] desc) as rn
from t1
join t2
on t2.[Shop] = t1.[Shop]
and t1.[Date] < t1.[Date]
) as t12
where t12.rn = 1
union
select t1.[Shop], t1.[Date], t1.[Sum]
, null as [Manager]
from t1
left join t2
on t2.[Shop] = t1.[Shop]
and t1.[Date] < t1.[Date]
group by t1.[Shop], t1.[Date], t1.[Sum]
having count(*) = 1
You may get much better performance by adding a covering index on t2 if you don't already have one:
create index T2ShopDate on t2 ([Shop], [Date]) include ([Manager])
Here is a version that uses a CTE to find all maximum manager dates first and then join back to t2 to get the manager:
;with MaxDates ([Shop], [Date], [Sum], [MaxMgrDate]) as
(
select
t1.[Shop]
,t1.[Date]
,t1.[Sum]
,max(t2.[Date])
from t1
left join t2
on t2.[Shop] = t1.[Shop]
and t2.[Date] < t1.[Date]
group by
t1.[Shop]
,t1.[Date]
,t1.[Sum]
)
select
MaxDates.[Shop]
,MaxDates.[Date]
,MaxDates.[Sum]
,t2.[Manager]
from MaxDates
inner join t2
on t2.[Date] = MaxDates.[MaxMgrDate]
You might be able to remove the second join back to t2 by using row_number():
;with MaxDates ([Shop], [Date], [Sum], [Manager], [RowNum]) as
(
select
t1.[Shop]
,t1.[Date]
,t1.[Sum]
,t2.[Manager]
,row_number() over (partition by (t1.[Shop]) order by t2.[Date] desc)
from t1
left join t2
on t2.[Shop] = t1.[Shop]
and t2.[Date] < t1.[Date]
)
select *
from MaxDates
where RowNum = 1
Totally confused and I have been working at this for 2 hours
I thought restriction on the left side of the join are honored
On this query I am getting [docSVsys].[visibility] 1 and <> 1
I thought this would restrict [docSVsys].[visibility] to 1
select top 1000
[docSVsys].[sID], [docSVsys].[visibility]
,[Table].[sID],[Table].[enumID],[Table].[valueID]
from [docSVsys] with (nolock)
left Join [DocMVenum1] as [Table] with (nolock)
on [docSVsys].[visibility] in (1)
and [Table].[sID] = [docSVsys].[sID]
and [Table].[enumID] = '140'
and [Table].[valueID] in (1,7)
This works
select top 1000
[docSVsys].[sID], [docSVsys].[visibility]
,[Table].[sID],[Table].[enumID],[Table].[valueID]
from [docSVsys] with (nolock)
left Join [DocMVenum1] as [Table] with (nolock)
on [Table].[sID] = [docSVsys].[sID]
and [Table].[enumID] = '140'
and [Table].[valueID] in (1,7)
where [docSVsys].[visibility] in (1)
I am just having a really off day as I had it in my mind the left side honored the join
SELECT *
FROM A
LEFT JOIN B ON Condition
is equivalent to
SELECT *
FROM A
CROSS JOIN B
WHERE Condition
UNION ALL
SELECT A.*, NULL AS B
FROM A
WHERE NOT EXISTS (SELECT * FROM B WHERE Condition)
Some rough pseudo-code...
Note, that all rows from A get through. It's just that the columns from B can be NULL if the join fails for some particular row of A.
Put the filter on docSVsys into the WHERE clause.
LEFT JOINs preserve all rows from the left (first) table, no matter what. The condition in the ON clause is only for matching which rows from the right/second table should be paired with rows from the left/first table.
If you want to exclude some rows from the firs table, use the WHERE clause:
select top 1000
[docSVsys].[sID], [docSVsys].[visibility]
,[Table].[sID],[Table].[enumID],[Table].[valueID]
from [docSVsys] with (nolock)
left Join [DocMVenum1] as [Table] with (nolock)
on [Table].[sID] = [docSVsys].[sID]
and [Table].[enumID] = '140'
and [Table].[valueID] in (1,7)
where [docSVsys].[visibility] in (1)