* _mysql.c: raise ProgrammingError for certain methods if they
are used on Result object created with conn.use_result(),
since they don't make sense.
* setup.py: Reimplement quote stripping around directory names.
This seems to be a bug in older versions of mysql_config
* cursors.py: If a unicode query string is passed, encode it in
the connection's character set.
and it seems to cause problems on some platforms or versions of MySQL
(most likely 3.23's mysql_config)
* removed use of weak references; I don't think they were needed any more,
and they caused problems in some of the threaded tests.
* applied patch #1188359: If converters are passed in, use a copy().
* Version 1.2.1c2.
* Reworked Unicode support; please test
* Make Binary() create an array('c')
* Clean up Cursors a bit.
* Add a TimeDelta factory function, even though not in the API spec.
win32 now requires win32pipe to build. Build has only been
tested on Linux, but AFAIK, os.popen() works properly on
all other platforms, and certainly must work on anything
remotely POSIX/UNIX-like. I am assuming that it will work
on Mac OS X as well.
* setup.py: changed include and library path order
* MySQLdb/connections.py: fix super class reference in autocommit()
* Tested against MySQL-4.1.4a with InnoDB tables
* _mysql.c: restored missing mysql_errno() call (bug #1023466)
* _mysql.c: Only try to use SSL support if HAVE_OPENSSL is defined.
Raise NotSupportedError if ssl is requested but not available.
* Turn autocommit off initially
* Add support for mysql_autocommit, _commit, and _rollback API functions
(new in MySQL-4.1)
* Remove Connection.begin(); use SQL BEGIN or START TRANSACTION instead
* pytimes (standard datetime module) is now the default implementation
* Detect and handle MySQL-4.1 and newer TIMESTAMP (looks like DATETIME)
* UnicodeType and ObjectType now always handled (required features)
* Ditch support for L at the end of long integer
* Remove z and crypt libs if building for Windows
* Version 1.1.2
* The MySQL FIELD_TYPE converter can now be a sequence of 2-tuples.
Item 0 is a bit mask (using FLAG.*) which must be matched. This should
be an integer. Item 1 is the conversion function. If item 0 is not an
integer, then this function is considered the default converter for
this FIELD_TYPE. Note that these tuples are considered when the query
has been executed and the result is available, so it doesn't add a
per-row overhead.
* As a result of the above, BINARY BLOB fields are now returned as
character arrays using the array.array class. Non-BINARY BLOB fields
(i.e. TEXT) are returned as strings. If unicode is enabled, they are
returned as unicode strings.
* Bump version to 0.9.3b3.
