|
I have a need to execute a query that pulls all data from one table regardless of whether matching data (matched by a foreign key) is present in a second table. This would be a classic LEFT JOIN query except that if a matching record is found in the RIGHT table (table2), then I need to pull only the latest such record from that table.
In particular, I need to pull one field, called VE (maybe more in the future) from the RIGHT table, but only from the latest record in the RIGHT table that matches the data in the LEFT table. The basic scenario I have is an object described in the LEFT table and the results of a process performed on that object in the RIGHT table. There may be more than one process result for that object in the right table, but only the latest one matters to me for the purposes of this query.
To solve this problem I used the following query based on the advice of someone from Code Project, but it didn't quite work out as planned. The ID in this query is an autonumber field that I'm using to retrieve the latest record.
SELECT Details.*, Results.VE
FROM Details LEFT JOIN [SELECT Results.VE, Results.SerialNumber, Max(Results.ID) From Results Group By Results.VE, Results.SerialNumber]. AS Results ON Details.SN = Results.SerialNumber);
It worked for the most part, but there were 7 more records in this result table from this query than are present in the 'LEFT' table. It appears that the duplications were occuring in the subquery, such that if there was an object that had two (or more) process results with different VE values, then the duplication occurred. For other duplications, this query correctly retrieved only the last record.
I appreciate any help.
Thanks.
|
|
|
|
|
Most likely the reason is in your group by section. You group on two columns so you'll get as many rows as there are valid combinations present. But then on your join you use only one column (SerialNumber).
If you want the latest record based on maximum ID and the maximum of id should be fetched based on serialnumber, it may be simpler to use correlated scalar at select portion (there may be typos):
SELECT Details.*,
(SELECT Results.VE
From Results r1
WHERE Details.SN = r1.SerialNumber
AND ID = (SELECT MAX(ID)
FROM Results r2
WHERE Details.SN = r1.SerialNumber)
FROM Details;Hope this helps,
Mika
The need to optimize rises from a bad design.
My articles[ ^]
|
|
|
|
|
Thanks. Your explanation about the GROUP BY helped me to understand why I was getting some duplicates, but the solution does not work; the result set was so huge that I had to cancel the attempt to execute the query.
|
|
|
|
|
Few questions:
- do you want to get as many rows in the output as there are rows in Details-table (no where conditions)?
- how many rows there are in details table?
- how many rows there are in Results table?
- is SerialNumber indexed in Results-table?
- is ID indexed in results table?
The need to optimize rises from a bad design.
My articles[ ^]
|
|
|
|
|
Mika,
1. Ultimately, no, I want to retrieve rows from the Details table based on certain conditions, say date range or whether one of the fields in the table is equal to (OR LIKE) a certain parameter. But, I thought that retrieving all rows from the Details table (and no more than the total number of rows) was the simplest condition.
2. The sample database I have-and perhaps a small database at that-has 10080 rows in the Details table.
3. 10500 rows. Nearly all objects in the Details table have undergone the process whose results are recorded in the results table (sometimes more than once).
4. No
5. Yes, ID is indexed in the results table.
Thanks.
|
|
|
|
|
Okay,
So the performance is affected because you fetch so many records (not a real world situation, but although a good test case).
Anyway, since the scalar is correlated, it's executed as many times as there are rows in the result (in this case 10080 times). If the result is limited, it will greately affect performance.
Another thing. Try adding a new index with for column SerialNumber (or SerialNumber, ID). Leave the existing ID index as it is. I would guess that you'll see quite different performance.
Also you could have an index on Details.SN if not already indexed.
Mika
The need to optimize rises from a bad design.
My articles[ ^]
|
|
|
|
|
Thanks again, Mika, for your help.
For your information, this is the solution-suggested by someone at another site - that worked:
SELECT Details.*, r1.VE
FROM Details LEFT JOIN [SELECT r1.* FROM Results r1 INNER JOIN (SELECT SerialNumber, Max(ID) As maxID FROM Results GROUP BY SerialNumber) AS r2
ON r1.SerialNumber = r2.SerialNumber AND r1.ID = r2.maxID]. AS r1 ON (Details.SN = r1.SerialNumber);
The actual query I'm using is a bit more complex than this as it involves several more conditions in the LEFT JOIN statement and some more or less simple WHERE queries to limit the information returned from the Details table, but the basic idea of this query does solve the problem I had described.
Thanks again.
|
|
|
|
|
I've got a stored procedure running across 2 servers which either inserts or updates multiple rows of data into a series of tables from a single source table. I've run into a problem where, if the source table doesn't contain a record for a given ID, MS DTC throws an exception and bombs out of the transaction. e.g.
<br />
UPDATE MyTable<br />
SET p.Column1 = pm.Column1<br />
FROM<br />
Property p<br />
inner join Address a<br />
on p.PropertyRef = a.PropertyRef<br />
inner join remotedb.PropertyMaster pm<br />
on pm.PropertyRef = p.PropertyRef<br />
errors with 'The Microsoft Distributed Transaction Coordinator (MS DTC) has cancelled the distributed transaction.'. Has anyone encountered this - and ideally know of a solution / workaround?
TIA
Dave
It definitely isn't definatley
modified on Monday, October 13, 2008 12:02 PM
|
|
|
|
|
Few issues:
- in the statement, you have a reference to
src, but it's not defined. Typo? - I know that the source cannot have more than one row if update is used like this (or at least the results may be incorrect), but could it also be that the source row must have exactly one matching row. But then again, just guessing
The need to optimize rises from a bad design.
My articles[ ^]
|
|
|
|
|
Yeah the src part was a typo (now updated in the original question).
I think you're probably right in your second point; it appears that DTC is expecting exactly one record set but since it's getting none it just decides to not work. Frustrating!
It definitely isn't definatley
modified on Monday, October 13, 2008 1:23 PM
|
|
|
|
|
I made a simple test case and ran it in both SQL Server 2008 and 2005 without DTC and had no problems. The test was the following:
CREATE TABLE Test1 (
Column1 varchar(50)
);
INSERT INTO Test1 (Column1)
VALUES ('ABC');
CREATE TABLE Test2 (
Column2 varchar(50)
);
UPDATE Test1
SET Test1.Column1 = Test2.Column2
FROM Test2
UPDATE Test1
SET Test1.Column1 = a.Column2
FROM Test2 a
INNER JOIN Test2 b
ON a.Column2 = b.Column2
In all cases I received 0 rows updated as expected. I was curious since typically if DTC makes a stop it's because an error has occured in some node and it simply rolls back all pending transactions. So I was thinking that there may be another error behind the one you received, but I had no confirmation for that.
One thing you could test is to modify your statement like this:
UPDATE MyTable
SET p.Column1 = (SELECT pm.Column1
FROM Property p
inner join Address a
on p.PropertyRef = a.PropertyRef
inner join remotedb.PropertyMaster pm
on pm.PropertyRef = p.PropertyRef)
Perhaps there's a difference in the behaviour or at least you get some other info.
The need to optimize rises from a bad design.
My articles[ ^]
|
|
|
|
|
Cheers Mika,
I actually tried something similar on a couple of different machines and only found the problem on the original server so it's almost certainly down to a configuration issue (even though my client's DBA team reckon it's not). I've sent them a copy of DTCPing to do some diagnostics so hopefully that'll find the issue. Failing all else I'll get them to reinstall MSDTC and hopefully that'll mend everything.
Thanks for you help
Dave
It definitely isn't definatley
|
|
|
|
|
You're welcome.
If you find out it's a configuration issue and not a broken install, I'd be really glad if you could share the info.
I'm having a hard time to understand what configuration option would modify the behaviour so dramatically since DTC is just a controller and it doesn't have to understand what actually happens under the hood. Just needs to know if everything went as expected or not.
Hopefully you find the solution one way or another without too much pain.
Mika
The need to optimize rises from a bad design.
My articles[ ^]
|
|
|
|
|
Despite protests from my client's DBA team, the server was upgraded to SP2 (previously SP1) and a hotfix was applied http://support.microsoft.com/kb/937517/en-us[^] - apparently there's a known problem which boils down to an incorrect attention signal being passed to the linked server which forces the linked server to rollback the transaction.
*feels vindicated 
It definitely isn't definatley
|
|
|
|
|
Great  !
Hopefully that solves the problem. Based on the hotfix info it could be exactly what you were experiencing.
Also confirmed that it wasn't intended behaviour as we expected 
The need to optimize rises from a bad design.
My articles[ ^]
|
|
|
|
|
hy,
how can i create a transaction so that when ever i update a row of a table no other user can change it or access it..
its like transaction on a account , only one user can operate on an account at a time.
i tried the following sql statement LOCK [ TABLE ] name IN lock_mode
but it gives an error "incorrect syntax near table"
should i create a stored procedure and then use SET TRANSACTION ISOLATION LEVEL SERIALATION after creating a transaction.
plz help
|
|
|
|
|
|
Ashfield wrote: This is the 3rd time you have asked the question, the answer is NOT going to change!!!!!
"The clue train passed his station without stopping." - John Simmons / outlaw programmer
"Real programmers just throw a bunch of 1s and 0s at the computer to see what sticks" - Pete O'Hanlon
"Not only do you continue to babble nonsense, you can't even correctly remember the nonsense you babbled just minutes ago." - Rob Graham
|
|
|
|
|
Hi..
I m using mysql database.
In my table adapter, I have 2 queries.
Query(1)
INSERT INTO comment
(commentID, reviewID, commentStatus, submitterID, commentDate, lastUpadatedDate)
VALUES (?commentID, ?reviewID, ?commentStatus, ?submitterID, ?commentDate, ?lastUpadatedDate)
Query(2)
INSERT INTO commentdetail
(commentDetailID, commentID, commentType, locator1, locator2, comment, remark)
VALUES (?commentDetailID, ?commentID, ?commentType, ?locator1, ?locator2, ?comment, ?remark)
i can add the Query(1).
but everytime i tried to add Query(2) displays the following error.
The query builder failed.
Attempted to read or write protected memory. This is often an indication that other memory is corrupt.
i need urgently..
Could anyone know how to solve this error,Please?
thanks in advance.
|
|
|
|
|
I use .NET 2.0 and VS 2005
|
|
|
|
|
Haven't used MySql really so just guessing here.
In the second statement, you have a potentially reserved word comment and also remark. Try removing them from the statement:
INSERT INTO commentdetail
(commentDetailID, commentID, commentType, locator1, locator2)
VALUES (?commentDetailID, ?commentID, ?commentType, ?locator1, ?locator2)
The need to optimize rises from a bad design.
My articles[ ^]
|
|
|
|
|
Hi,
Can someone please help me with an issue I am having?
I have created a Table named "Countries" The table has a CountryID and a Country. What needs to happen is that the user must be able to add his/her country in the database. The problem I am having is the following:
The CountryID is NOT auto numbered. I need to write a trigger that executes BEFORE the insert procedure. It basically needs to check what the latest record ID is, if there are none than the CountryID should be 1 if there are records the trigger must take the last record and add 1.
I have never worked with triggers so I am in a bit of a pickle. Any help will be appreciated!!
Illegal Operation
|
|
|
|
|
Why can't you change the countryid field into an identity (autonumber) field.
Triggers work on the inserted record and therefore fire after insert. Identity field is the designed methedology, you are trying to make a trigger do the job of another function.
Alternatively you can create an insert proc that gets the number before the insert begins. This is subject to corruption in high volume tables (2 attempts at the same time - very unlikely but must be considered).
Never underestimate the power of human stupidity
RAH
|
|
|
|
|
Ok, point taken. What happens when you use auto numbering and you have 10 entries in your table, now you need to delete entry number 7. What are you going to do to fix the table??
Illegal Operation
|
|
|
|
|
Define "fix" the table. Guessing you are stressing about the missing sequential number 7 from the identity field.
This is Database 101. ID fields must NEVER have intelligent data, they are used to identify the record and link it to other tables. The sooner you get this idea the happier you will be. If you are using the ID field other than to locate the record (eg sorting) then your design is tragically flawed.
I recommend you get a book on database design and data structuring (mine are so old they are no longer published).
Never underestimate the power of human stupidity
RAH
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
