2.10 Transaction Processing

There are two ways to control transactions through Microsoft SQL Server:

setting isolation levels
using transaction control methods

Each of these methods is described below.

2.10.1 Setting Isolation Levels

You can use the isolation() method of RWDBConnection to set the isolation level of the SQL server. Table 5 shows the mapping between the IsolationType argument you pass and the isolation level set by the SQL server. Take special care when setting isolation levels as this effects all users of the SQL server.

Table 5 -- Setting the isolation level

RWDBConnection::IsolationType	Microsoft SQL Server Isolation Level
Unknown	Level 1 - `SQL_TXN_READ_UNCOMMITTED`
ANSILevel1	Level 1 - `SQL_TXN_READ_UNCOMMITTED`
ANSILevel2	Level 1 - `SQL_TXN_READ_COMMITTED`
ANSILevel3	Level 3 - `SQL_TXN_REPEATABLE_READ` or: `SQL_TXN_SERIALIZABLE`

To determine the isolation level, call RWDBConnection::isolation() without an argument.

2.10.2 Using Transaction Control Methods

You can explicitly control transactions through the following methods:

RWDBConnection::beginTransaction()

RWDBConnection::rollbackTransaction()

RWDBConnection::rollbackTransaction(RWCString savepoint)

RWDBConnection::commitTransaction()

RWDBConnection::setSavepoint(RWCString savepoint)

These methods are implemented using the following Transact-SQL transaction statements:

"begin transaction"
"rollback transaction"
"rollback transaction savepoint"
"commit transaction"
"save transaction savepoint"

An application can add the DBTools.h++ transaction methods to its code to take explicit control of its transaction blocks. Transactions may be nested and the savepoint feature is supported. The savepoint feature allows a current transaction to be partially rolled back to a marked point.

NOTE: At the time of this writing, conflicting information was found in the MS SQL Server ODBC driver documentation regarding the use of Transact-SQL transaction statements. Please see the Microsoft SQL Server readme for more information regarding transaction savepoints.

The example below demonstrates the use of the savepoint feature and the other transaction processing methods of DBTools.h++.

// Assume we have a table myTable(c int) with no rows in it.
RWDBInserter ins= myTable.inserter(); 

cn2.beginTransaction("outmost");   // Begin transaction. 
                                   // Transaction name is allowed.
                                   // Note: unlike savepoint name,
                                   // transaction name is not used
                                   // by server.

(ins << 1 ).execute(cn2);          // First insertion
cn2.beginTransaction("inner");     // Nested transaction begins
...
cn2.setSavepoint("svp1");          // Save first insertion

(ins << 2 ).execute(cn2);          // Second insertion
(ins << 3 ).execute(cn2);          // Third insertion

cn2.rollbackTransaction("svp1");   // Roll back second and 
                                   // third insertions.
cn2.commitTransaction("inner");    // Pairing inner transaction
cn2.commitTransaction("outmost");  // Commit transaction on
                                   // part that is not rolled back.

// The above program results in myTable holding one row of data.
// Its value is 1.