Home | Back
/*
  Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.

  The MySQL Connector/C++ is licensed under the terms of the GPLv2
  <http://www.gnu.org/licenses/old-licenses/gpl-2.0.html>, like most
  MySQL Connectors. There are special exceptions to the terms and
  conditions of the GPLv2 as it is applied to this software, see the
  FLOSS License Exception
  <http://www.mysql.com/about/legal/licensing/foss-exception.html>.

  This program is free software; you can redistribute it and/or modify
  it under the terms of the GNU General Public License as published
  by the Free Software Foundation; version 2 of the License.

  This program is distributed in the hope that it will be useful, but
  WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
  or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
  for more details.

  You should have received a copy of the GNU General Public License along
  with this program; if not, write to the Free Software Foundation, Inc.,
  51 Franklin St, Fifth Floor, Boston, MA 02110-1301  USA
*/

GA  1.1.0 - 2010-09-13

- Added Driver::threadInit() and Driver::threadEnd() methods. Every thread
  of a threaded client must call Driver::threadInit() at the very start
  of the thread before it does anything else with Connector/C++ and every
  thread must call Driver::threadEnd() when it finishes. You can find an
  example demonstrating the use in examples/pthreads.cpp. It is strongly
  discouraged to share connections between threads. It is theoretically
  possible, if you set certain undocumented mutexes, but it is not
  supported at all. Use one connection per thread. Do not have two
  threads using the same connection at the same time. Please
  check the C API notes on threading on the MySQL manual. Connector/C++
  wraps the C API. (Lawrin, Andrey, Ulf)

- Fixed ResultSetMetaData::getColumnType() to return sql::DataType::VARCHAR
  even for multi-type character sets. (Fix from Andrey)

- Fixed ResultSetMetaData::getColumnDisplaySize() to return number
  of characters instead of number of bytes. However, there are cases
  when the MySQL server returns wrong or insufficient meta data, see also
  http://bugs.mysql.com/bug.php?id=6399 . Application developers can rely
  on getColumnDisplaySize() returning at least the number of required
  characters. Application developers shall not expect the exact number
  for a given result set. The driver cannot get any closer but
  returning the maximum number for a given result set (Fix from Andrey)

- Changed FindMySQL.cm so it looks for mysqlclient in the ENV{MySQL_DIR}\lib,
  as that is where it's located in (some) Connector/C distros. (Lawrin)

- Fixed bug in CmakeLists.txt - it didn't really use ENV{BOOST_ROOT}, boost wasn't
  searched in the directory set in that environment variable. (Lawrin)

- Fixed http://bugs.mysql.com/bug.php?id=51562 : error messages did not get
  reported properly if error occured after query execution and before fetching
  data within the underlying C API call mysql_use|store_result. (Andrey)

- BIT fields are now correctly decoded (Workbench bug #36239). (Andrey)

- Connection::getClientOption(const sql::SQLString & optionName, void * optionValue)
  now accepts the optionName values "metadataUseInfoSchema",
  "defaultStatementResultType", "defaultPreparedStatementResultType",
  "characterSetResults". In the previous version only "metadataUseInfoSchema"
  was allowed. The same is true for Connection::setClientOption().

- Fixed Bug #45048 "prepared statements corrupt the heap". ResultBind is back in
  the prepared statement, but shared with the result set. (Andrey)

- get_driver_instance() is only available in dynamic library builds, static builds
  won't have this symbol. This is done, because someone might decided to load the
  dll with LoadLibrary/dlopen and needs an entry point to start using the library
  (some kind of a Driver Manager). For those, who don't use cmake for building
  you need to define mysqlcppconn_EXPORTS if you are loading dynamically and want
  to use that entry point. (Andrey)

- ABI change - added support for optional run-time dynamic loading of the MySQL Client Library.
  This is useful if you want to re-distribute a certain client library with
  the connector and make sure that it gets used. It also allows you to
  use a different client library for every connection. As it is an advanced
  feature it is not enabled by default. By default the MySQL Client Library
  will be linked at compile time just as in every version before.

- Fixed bug in ResultSetMetaData for normal statements and prepared ones, getScale
  and getPrecision did return wrong results. (Andrey)

- Fixed http://bugs.mysql.com/bug.php?id=45846 . Excluding dynamically generated
  and platform specific header files source packages generated using cpack (Ulf).

- Connection map property OPT_RECONNECT is now of type boolean (was long long).

- We now check LDFLAGS, CXXFLAGS and CPPFLAGS from the environment for every binary
  we generate. (Ulf, Kent)

- Fixed http://bugs.mysql.com/bug.php?id=45843: cmake error if configuring an out
  of source build [when not calling cmake in the root directory]. (Ulf)

- Fixed http://bugs.mysql.com/bug.php?id=44931: missing includes when using GCC 4.4.
  Note that GCC 4.4 is not yet in use for any official MySQL builds. (Ulf, Lawrin)

- Fixed a performance issue of Prepared Statements. Reading large result sets has
  been slow - think O(n^2). Patch by Frank Schoenheit from the OpenOffice.org
  Base team. (Ulf)

- Exchanged most use cases of std::auto_ptr with boost::scoped_ptr. Migrated from
  own solution to boost::scoped_array. In addition, migrated from another own
  solution to boost::shared_ptr/weak_ptr for guarding access around result sets.
  (Andrey)

- API incompatible change: ConnectPropertyVal is no more a struct by a typedef
  that uses boost::variant. This will break your build but is pretty easy to fix.
  If you had code like
  ---
  {
    sql::ConnectPropertyVal tmp;
    tmp.str.val=passwd.c_str();
    tmp.str.len=passwd.length();
    connection_properties["password"] = tmp;
  }
  ---
  Your new code you will like this:
  ---
  connection_properties["password"] = sql::ConnectPropertyVal(passwd);
  ---
  Which is simpler. (Andrey)

- Fixed a bug in PS, which in combination of preparing a stored procedures without
  any parameters was leading to exception, which was a problem, and was leading
  to another problem which manifested itself with double free. All similar code
  snippets in the Connector were fixed. (Andrey)

- The driver makes use of some Boost components (http://www.boost.org).
  You need to have Boost 1.34.0 or newer installed on your system in
  order to compile the driver from source. Binary distributions have no
  additional dependencies over 1.0.5.

- API change: parameter to setNextQuery in SQLWarning got const qualifier.
  SQLWarning itself is now an interface with protected destructor.


GA  1.0.5 - 2009-04-20

- Changed the interface of sql::ConnectionMetaData, sql::ResultSetMetaData
  and sql::ParameterMetaData to have a protected destructor. In this way the
  client code doesn't need, and won't be able, to destruct the metadata
  objects returned by the connector. The connector will handle their
  destruction. This enables statements like :
  connection->getMetaData->getSchema();
  without the result of leaking memory because we lost the pointer returned
  by getMetaData(). (Lawrin, Andrey)

- Large overhaul of the code to improve the memory management to not leak in
  exceptional situations. Big improvement compared to Beta1. (Andrey)

- Fixed the interface of sql::Driver and sql::Connection so they accept the
  options map by alias instead of by value. (Andrey)

- Changed the return type of sql::SQLException::getSQLState() from
  std::string to const char * to be consistent with std::exception::what().
  (Andrey)

- Implemented getResultSetType() and setResultSetType() for Statement. Used
  are TYPE_FORWARD_ONLY, which means unbuffered result set and
  TYPE_SCROLL_INSENSITIVE, which means buffered result set. (Andrey)

- Implemented getResultSetType() for PreparedStatement. The setter is not
  implemented because currently PreparedStatement can't do refetching and
  storing the result means the bind buffers will be correct. (Andrey)

- Added "defaultStatementResultType" to MySQL_Connection::setClientOption() as
  an option. Also the method now returns `sql::Connection *`. (Andrey)

- Added Result::getType() and implemented it in the three result set classes.
  (Andrey)

- Enabled tracing functionality when building with VC8 and up (VS2005 and up).
  (Andrey)

- Added better support for named pipes, on Windows. Use pipe:// and add the
  path to the pipe. Shared memory connections are currently not supported.
  (Andrey)

- Fixed a bug in MySQL_Connection::setSessionVariable() which led to exception
  being thrown. (Andrey)


Beta   1.0.4 - 2009-03-31

- Prepared support for upcoming Connector/C. (Georg)

- Added Windows installer. (Georg)

- Bumping up CMake minimum version requirement from 2.4.2 to 2.6.2.
  We need the latest version for Windows. (Lawrin)

- Added "metadataUseInfoSchema" to connection propery map which allows you
  to control the use of the INFORMATION_SCHEMA for meta data. (Andrey)

- Fixed a bug in all implementations of ResultSet::relative() which was giving
  wrong return value although positioning was working correctly. (Andrey)

- Fixed a leak in MySQL_PreparedResultSet when the result was containing a BLOB
  column. (Andrey)

- Implemented MySQL_ConnectionMetaData::supportsConvert(from, to). (Andrey)

- Introduced sql::DataType::YEAR to complement MySQL's YEAR type. (Andrey)

- Introduced PreparedStatement::getMetaData(). (Andrey)

- Introduced ResultSetMetaData::isZerofill(), which is not in the JDBC
  specification. (Andrey)

- Fixed all implementations of ResultSet::isNull() to check whether the current
  position is on a real row, not isBeforeFirst() nor isAfterLast(), like all
  getXXX methods do. (Andrey)

- Implementation for MySQL_DatabaseMetaData::getProcedures() when
  INFORMATION_SCHEMA is asked not to be used. (Andrey)

- Removed MySQL_DatabaseMetaData::getProcedureColumns() from the interface.
  Until now it was returning always an empty result set. Full implementation
  will be added at a later stage. (Andrey)

- Changed a bunch of methods of DatabaseMetaData()::getXXX, which returned `int`
  to return `unsigned int` because it makes more sense. (Andrey)


Alpha  1.0.3 - 2009-03-03

- Added new tests at test/unit/classes. Those tests are mostly about
  code coverage. Most of the actual functionality of the driver is tested
  by the tests found at test/CJUnitPort. (Ulf)

- New data types added to the list returned by DatabaseMetaData::getTypeInfo();
  FLOAT UNSIGED, DECIMAL UNSIGNED, DOUBLE UNSIGNED. Those tests may not
  be in the JDBC specification. However, due to the change you should be able
  to look up every type and type name returned by, for example,
  ResultSetMetaData::getColumnTypeName(). (Andrey)

- MySQL_Driver::getPatchVersion introducted. (Andrey)

- Major performance improvements due to new buffered resultset implementation
  by Andrey. (Ulf)

- Addition of test/unit/README with instructions for writing bug/regression
  tests. (Ulf)

- Experimental support for STLPort. This feature may be removed again at any
  time later without prior warning! Check cmake -L for configuration
  instructions. (Andrey)

- Fixed a bug in MySQL_PreparedResultSet::getString(). Returned string had real
  data but the length was random. Now, the string is initialized with correct
  length and thus is binary safe. (Andrey)

- Added properties-enabled (map of key->value) methods for connecting, which
  add many connect options. (Andrey)
  -- Driver::connect( map )
  -- Connection::Connection( map )

- New BLOB implementation. Got rid of sql::Blob in favor of std::istream. C++'s
  IOStream library is very powerful, similar to PHP's streams. It makes no sense
  to reinvent the wheel. For example one can pass a std::istringstream object
  to setBlob() if the data is in memory, or just open a file std::fstream and
  let it stream to the DB, or write own stream. Similar will be true for getBlob()
  where we can just copy data, if buffered result set, or stream, if we implement
  it. (Andrey)

- Implemented ResultSet::getBlob() which returns std::stream. (Andrey)

- Fixed MySQL_DatabaseMetaData::getTablePrivileges() to work correctly. Test cases
  added in the first unit testing framework. (Andrey)

- Implemented MySQL_Connection::setSessionVariable() for setting variables like
  sql_mode. (Andrey)

- Implemented MySQL_DatabaseMetaData::getColumnPrivileges(). (Andrey)

- cppconn/datatype.h changed and used again. Reimplemented the type subsystem to
  be more usable - more types for binary and non-binary strings. (Andrey)

- Implementation for MySQL_DatabaseMetaData::getImportedKeys() for
  MySQL versions before 5.1.16 using SHOW, and above using INFORMATION_SCHEMA.
  (Andrey)

- Implemented MySQL_ConnectionMetaData::getProcedureColumns(). (Andrey)

- make package_source packs now with bzip2. (Andrey)

- Re-added getTypeInfo() with information about all types supported by MySQL and
  the sql::DataType. (Andrey)

- Exchanged the implementation of MySQL_ConstructedResultSet to use more efficient
  non O(n) but O(1) access method. This should improve the speed with which
  the metadata result sets are used. Also, there is less copy during the
  construction of the result set, which means that all result sets returned from
  the metadata functions will be faster. (Andrey)

- Introduced, internally, sql::mysql::MyVal which has implicit constructors used
  in mysql_metadata.cpp to create result sets with more native data instead of
  always string (varchar). (Andrey)

- Renamed ResultSet::getLong() to ResultSet::getInt64(). resultset.h includes
  typdefs for Windows to be able to use int64_t. (Andrey)

- Introduced ResultSet::getUInt() and ResultSet::getUInt64(). (Andrey)

- Corrected handling of unsigned server types. Now returning correct values.
  (Andrey)

- Fixed handling of numeric columns in ResultSetMetaData::isCaseSensitive to
  return false. (Andrey)

- Better implementation for ResultSetMetaData::isReadOnly. Values generated
  from views are read-only. Seems that these generated values don't have `db`
  in MYSQL_FIELD set, while all normal columns do have. (Andrey)

- Implemented MySQL_DatabaseMetaData::getExportedKeys(). (Andrey)

- Implemented MySQL_DatabaseMetaData::getCrossReference(). (Andrey)


Alpha  1.0.2 - 2008-12-19

- Adding test/unit as a basis for general unit tests based on the new
  test framework, see test/unit/example for basic usage examples (Ulf)

- Fixed MySQL_PreparedStatement::setBlob() to really work. In the tests
  there is a simple example of a class implementing sql::Blob. (Andrey)

- Addition of a new unit test framework for JDBC compliance and
  regression testing. (Lawrin)

- Implemented MySQL_ResultSetMetaData::getPrecision() and
  MySQL_Prepared_ResultSetMetaData::getPrecision(),
  updating example. (Andrey)

- Fixing bug in FLOAT handling. (Andrey)

- Fixing bug in getString():
  getString() is binary safe now (Andrey),
  new example. (Ulf)

- Fixing bugs in MySQL_PreparedStatements:
  setBigInt() and setDatetime() have decremented the internal column
  index before forwarding the request. This resulted in double-decrement
  and wrong internal column index. Typical error message:
  setString() ... invalid "parameterIndex" (Ulf)

- Adding PHP script examples/cpp_trace_analyzer.php to filter the
  output of the debug trace. Please see the inline comments for documentation.
  This script is unsupported! We do no promise to maintain it. (Ulf)

- Fixed bugs in MySQL_DatabaseMetaData :
  All supportsCatalogXXXXX methods were returning `true` and
  all supportSchemaXXXX methods false, which is not as it should be.
  Now it is reversed, to be consistent with the rest. (Andrey)

- Implemented MySQL_PreparedStatement::clearParameters(). (Andrey)

- Fixed a bug in MySQL_ConnectionMetaData::getColumns() which was
  performing a cartesian product of the columns in the table times
  the columns matching columnNamePattern. example/connection_meta_schemaobj.cpp
  extended to cover the function. (Andrey)

- Fixed lame bug in MySQL_ConnectionMetaData::getIndexInfo() which
  did not work because the schema name wasn't included in the query sent
  to the server. (Andrey)

- Implemented MySQL_PreparedStatement::setNull(). (Andrey)

- Reverted implementation of MySQL_DatabaseMetaData::getTypeInfo().
  Now unimplemented. In addition, removed cppconn/datatype.h for now
  till we havea proper implementation of the types.

- DATE, DATETIME and TIME are now being handled when calling
  MySQL_PreparedResultSet::
  -- getString()
  -- getDouble()
  -- getInt()
  -- getLong()
  -- getBoolean()

- Fixed MySQL_PreparedStatementResultSet::getDouble() to return proper
  value when the underlying type is MYSQL_TYPE_FLOAT. (Andrey)

- Changed ResultSetMetaData::
  -- getColumnDisplaySize()
  -- getPrecision()
  -- getScale()
  to return unsigned int instead of signed int. (Andrey)

- Implemented getScale(), getPrecision() and getColumnDisplaySize() for
  MySQL_ResultSetMetaData and MySQL_Prepared_ResultSetMetaData. (Andrey)


Alpha  1.0.1 - 2008-12-01

- New directory layout

- MySQL Workbench 5.1 using Connector/C++ for its database connectivity.

- Addition of Connector/J tests converted

- Changing sql::DbcException to implement the interface of JDBC's SQLException.
  And renamed it to sql::SQLException.

- Renamed sql::DbcInvalidArgument to sql::InvalidArgumentException

- Renamed sql::DbcMethodNotImplemented to sql::MethodNotImplementedException

- All tests changed to create TAP compliant output

- Introduction of experimental CPack support, see make help

- Metadata: switching to column names "TABLE_CAT" (was: TABLE_CATALOG) and
  "TABLE_SCHEM" (was: TABLE_SCHEMA) for JDBC compliance

- ConnectionMetaData::getImportedKeys():
  PKTABLE_CATALOG -> PKTABLE_CAT, PKTABLE_SCHEMA -> PKTABLE_SCHEM,
  FKTABLE_CATALOG -> FKTABLE_CAT, FKTABLE_SCHEMA -> FKTABLE_SCHEM

- ConnectionMetaData::getPrimaryKeys():
  COLUMN -> COLUMN_NAME, SEQUENCE -> KEY_SEQ, INDEX_NAME -> PK_NAME

- ConnectionMetaData::getProcedures:
  PROCEDURE_SCHEMA -> PROCEDURE_SCHEM

- ConnectionMetaData::getTables:
  TABLE_COMMENT -> REMARKS

- All examples can be given optional connection parameters on the command
  line, for example
  `examples/connect tcp://host:port user pass database`
  or
  `examples/connect unix:///path/to/mysql.sock user pass database`

- Adding experimental GCov support, cmake -DMYSQLCPPCONN_GCOV_ENABLE:BOOL=1

- ConnectionMetaData::getCatalogTerm() returns n/a, there is no counterpart
  to catalog in Connector/C++

- Addition of ConnectionMetaData::getSchemas() and  Connection::setSchema().
  None of them is in the JDBC standard

- Driver Manager removed

- `(n)make install` works.  You can change the default installation path.
  Read carefully the messages after executing cmake. Installed are the static
  and the dynamic version of the library, libmysqlcppconn as well as the
  generic interface cppconn + 3 MySQL specific headers
  -- mysql_driver.h (if you want to get your connections from the driver
     instead of instantiating a MySQL_Connection object. This makes your code
pretty portable against the common interface)
  -- mysql_connection.h - If you intend to link directly to the MySQL_Connection
     class and use its specifics not found in sql::Connection
  However, you can make your application fully abstract by not using the two
  headers above but the generic headers.

- sql::mysql::MySQL_SQLException is gone. There is no distinction between
  server and client (= Connector) caused errors based on the type of the
  exception. However, you can still check the error code to figure out the
  reason.


Preview 1.0.0  - 2008-08-05

- First public release