[SOLVED] Why connection in Python's DB-API does not have "begin" operation?

Why connection in Python's DB-API does not have "begin" operation?

Working with cursors in mysql-python I used to call "BEGIN;", "COMMIT;", and "ROLLBACK;" explicitly as follows:

try:
    cursor.execute("BEGIN;")
    # some statements
    cursor.execute("COMMIT;")
except:
    cursor.execute("ROLLBACK;")

then, I found out that the underlying connection object has the corresponding methods:

try:
    cursor.connection.begin()
    # some statements
    cursor.connection.commit()
except:
    cursor.connection.rollback()

Inspecting the DB-API PEP I found out that it does not mention the begin() method for the connection object, even for the extensions.

Mysql-python, by the way, throws the DeprecationWarning, when you use the method. sqlite3.connection, for example, does not have the method at all.

And the question is why there is no such method in the PEP? Is the statement somehow optional, is it enough to invoke commit() instead?

Solution

Decided to answer myself:

A thread about DB API 2.0 transactions in python-list and the following excerpt from the noticeable book SQL The Complete Reference make me think that DB API implements SQL1 standard behaviour:

The first version of the SQL standard (SQL1) defined an implicit transaction mode, based on the transaction support in the early releases of DB2. In implicit mode, only the COMMIT and ROLLBACK statements are supported. A SQL transaction automatically begins with the first SQL statement executed by a user or a program and ends when a COMMIT or ROLLBACK is executed. The end of one transaction implicitly starts a new one.

Explicit transaction mode (the SQL2 and SQL:1999) seems to be handy when the RDBSM supports autocommit mode and the current connection is in that mode, but DB API just does not reflect it.