Database

22-Oct-01 6:48

I have a couple of questions about using BeginTrans() in a try catch block. Below is an example of how I use ADO to modify a record in my Application. This application was originally made to use JET and I am tyring to migrate to MSDE. When I was using JET to maintain consistancy between the records of one connection and the records of another connection in the same app using a different thread I had to flush the cache in Jet ( using a call to pJetEngine->RefreshCache( m_pConnection)) and then call BeginTrans, update the data and call CommitTrans. In JET each ConnectionPtr has a cache on the database that is independent of other open connections. Using this procedure the thread will see and modify the current data in the db and not what was cached for this connection. Do we have to do someting like this with MSDE? Is there a way to flush the cache? Also is it possible that BeginTrans / CommitTrans will throw an exception? If so how do I handle BeginTrans inside a try / catch block? Do I call RollbackTrans in each of my catch routines? Could RollbackTrans() throw an exception?

_RecordsetPtr pRst = NULL;
try
{
TESTHR(pRst.CreateInstance(__uuidof(Recordset)));

pRst->Open(_bstr_t(strQuery), m_pConnection, adOpenStatic ,
adLockOptimistic, adCmdText);

m_pConnection->BeginTrans();

int count = pRst->GetRecordCount();
if ( count == 1 ) {
// Modify record using PutCollect
}

pRst->Update();
pRst->Close();
pRst=NULL;
m_pConnection->CommitTrans();
}
catch (_com_error &e)
{
GenerateError(e.Error(), e.ErrorMessage(), e.Description(),__FILE__,__LINE__);
}
catch( CException* e)
{
GenerateError(e,__FILE__,__LINE__);
}
catch (...)
{
GenerateError(__FILE__,__LINE__);
}

Wanderley M22-Oct-01 23:11

22-Oct-01 23:11

Hi
Do I call RollbackTrans in each of my catch routines?

Looking at your sample, the answer is yes. But I have one question: why your try/catch block and begintrans/committrans enclose all your recordset operation?

Regards,

Wanderley

John M. Drescher23-Oct-01 7:14

23-Oct-01 7:14

There are several cases in my database where I need to guarantee that the data read from one ConnectionPtr is the latest data written to the database. This generally happens when I need to generate an ID for something and don't use an AutoNumber. One example of this is when I need to generate a Patient ID when the user has not entered a valid one and the case has been forcibly closed because of inactivity.

As I said in my original description, because of caching that occurs in the connection at least when using JET to gaurantee you have the latest data from one ConnectionPtr to another you have to flush then cache (on reads) and do begintrans/committrans on all writes that need to be the latest. There is a MS KB article Q200300 that explains the problem as it refers to JET, it is a couple of years old and I have no idea if it applies to MSDE also.

Below is taken from the KB article:
In certain situations you may need to use two separate ADO connections when writing and reading data from an Access database using the Jet OLEDB Provider. For example, if you have two separate processes that are writing and reading to the same Access database, there is no way to share a single connection. In this situation, you can synchronize writes and reads with separate ADO connections if you follow these guidelines:

The writer must start a transaction, using ADO's Connection.BeginTrans, prior to writing the data.

The writer must make the database updates and then commit the transaction (using ADO's Connection.CommitTrans).

The reader must call JRO.JetEngine.RefreshCache passing in it's connection prior to attempting to read the data.

Wanderley M23-Oct-01 11:50

23-Oct-01 11:50

Hi
I asked that because I use a different approach for try/catch and begin/rollback/committrans and even for custom generated PKIDs (I create it at stored procedure level).

Sorry, I couldn't help this time. Smile | :)

Regards,

Wanderley

John M. Drescher23-Oct-01 12:34

23-Oct-01 12:34

Thanks. Actually you have been very helpful. I have not used any stored procedures (except the system ones) yet.. I've only started to migrate from JET to MSDE last friday and T-SQL is new to me.

Wanderley M23-Oct-01 22:45

SQL Server 7.0 SP3 and money datatype

23-Oct-01 22:45

Hi
One more thing: do you need to be sure that you have the latest data to create unique IDs? Do you keep a table of IDs and increment it whenever you need?

I said I have a different approach because I just use try/catch blocks where it's more likely (is that right? english is my second language Smile | :)

) to have trouble. In your sample, I would use it only when opening the recordset. And I use begin/commit/rollback when I'm actually writing data (before addnew and after update, for example).

Regards,

Wanderley

Michael Martin21-Oct-01 20:46

Michael Martin

21-Oct-01 20:46

I have built a database that has some tables that use the money datatype. The pain in the arse in when I retrieve the data using SQL it always returns 4 decimal places.

Is there some way that I force SQL Server to only store 2 decimal places? Or is the something I can do in the SQL statement to truncate/round it to 2 decimal places? Either will do as I only store 2 decimal places worth of value in there anyway.

Michael Martin
Pegasystems Pty Ltd
Australia
martm@pegasystems.com
+61 413-004-018

"Don't belong. Never join. Think for yourself. Peace"
- Victor Stone

Re: SQL Server 7.0 SP3 and money datatype

Wanderley M22-Oct-01 3:32

Re: SQL Server 7.0 SP3 and money datatype

22-Oct-01 3:32

Hi
Is there a specific reason to use money? You could use numeric instead of money - this way you can define how many decimal places you need.

Do you need to make calculations after you retrieve the data? If not, you could use SELECT STR(<your field="">, 8, 2) FROM [...]

HTH

Regards,

Wanderley

Michael Martin22-Oct-01 15:32

Michael Martin

22-Oct-01 15:32

Wanderley

thanks for the information looks like what I need. There isn't any reason to use money other than I was lazy, saw it and used it. I will try it as numeric, otherwise I will use the SELECT trick. Big Grin | :-D

Michael Martin
Pegasystems Pty Ltd
Australia
martm@pegasystems.com
+61 413-004-018

"Don't belong. Never join. Think for yourself. Peace"
- Victor Stone

MSDE 2000 Question

John M. Drescher21-Oct-01 8:57

21-Oct-01 8:57

I'm having a problem migrating from JET to MSDN 2000 in my application. I keep getting timeout errors and the following error message "cannot start more transactions on this session." Is this the result of the 5 connection limit on MSDE? I am only opening and using the program in one application that has multiple threads and opens the database with a connection for each thread that needs db access.

Re: MSDE 2000 Question

John M. Drescher22-Oct-01 5:48

ADO DataGrid, MS Access & Server-side cursors

22-Oct-01 5:48

I think I found the problem. It was with my transactions. I called _pConnection->BeginTrans() inside a try / catch block. When an exception was raised because of a duplicate index, the m_pConnection->CommitTrans() was not being called and the transaction was still open. The next call (and all subsequent) call to BeginTrans would fail.

Jim Stim18-Oct-01 22:53

Jim Stim

18-Oct-01 22:53

Can anyone tell me why recordset data from an Access table can't be displayed on a DataGrid control if the recordset cursor location is set to the server?

Re: ADO DataGrid, MS Access & Server-side cursors

Rashid Thadha21-Oct-01 22:40

Rashid Thadha

21-Oct-01 22:40

For a recordset to be bound to a grid, it must support bookmarks. A server side dynamic cursor on SQL Server does not support bookmarks. Try using a keyset cursor.

also see msdn article Q224192

http://support.microsoft.com/support/kb/articles/Q224/1/92.ASP

How to age out subscribers accounts?

Jason Jystad18-Oct-01 9:37

Jason Jystad

18-Oct-01 9:37

I have a web classifieds system that I wrote for a client and I am having some trouble with the aging out of accounts. THis system is built with ASP and interfaces with MSSQL Server 7.0.

What happens is this, a person subscribes for a period of 90 days. They pay their fee and get the run of the system for 90 days. When the 90 days are up they are no longer allowed access. They can still log in, but they are only given access to the pay area so they can re-up their subscription. If they do not re-up then after a while their account goes away entirely.

So I have three states of user:
Non-Subscriber
Subscriber
Expired Subscriber

Non-subscribers become subscribers by entering their user info and paying, that is simple enough and it works fine.

Subscribers last for 90 days and then they become Expired Subscribers. (They get flagged in the database as expired)

Expired Subscribers last for another 30 days and then they go away. (They are deleted from the database)

So my question:
What is the most efficient way to test for the age of the accounts? (I am comfortable with databases, but I do not claim to be an expert so I am curious about what you guys think SQL wise)

Also, how would you guys get the system to run the check routine regularly? Say once a day. This is an ASP based system and it is hosted on a virtual hosting provider so I cannot just set up a scheduled job to run on the server like I normally would. I am going to have to find a tricky way to do it.

TIA!

Jason Jystad
Cito Technologies
www.citotech.net

>------------------------------------------------<
"Luckily," he went on, "you have come to exactly the right place with your interesting problem, for there is no such word as 'impossible' in my dictionary. In fact," he added, brandishing the abused book, "everything between 'herring' and 'marmalade' seems to be missing."
-- Dirk Gently (Douglas Adams)
>------------------------------------------------<

Re: How to age out subscribers accounts?

Wanderley M19-Oct-01 8:42