root/trunk/Changes

# $Id$

Revision history for Data::ObjectDriver
0.07
    - When an object is changed, delete the cache instead of updating it.
      It is a bit more expensive, but should safer.
    - Fixed a uuv warning thanks to Kazuhiro Shibuya
    - Fixed a issue in the test suite for newer version of SQLite
      reported by Emmanuel Seyman
      http://rt.cpan.org/Ticket/Display.html?id=45186

0.06  2009.01.28
    - Added peek_next() method to ResultSet, q.v.
    - Localized creation of D::OD::Iterator object. Thanks to Hirotaka Ogawa
      for the patch.
    - Fixed compilation error with Perl 5.10. Thanks to smpeters for the patch.
    - Added a new $object->uncache_object as a mirror of cache_object(), which
      purge one object from the cache layer, for the cases where you want a
      manual control over it.
    - Added a "distinct" method to D::OD::SQL that forces the DISTINCT keyword
      in the generated SQL statement. Thanks to John Berthels for the patch.
    - Added a "window_size" argument for the search() method of the caching
      layer to constrain the number of objects loaded from the database for
      large or unbounded searches. 
    - Added a "comment" argument to search parameter allowing the SQL
      queries to be watermarked with SQL comments.
    - Added a "object_is_stored" method on DOD objects, which returns true until
      the object has been saved in the persistent store.
    - Added a "pk_str" method on base objects has a nice shortcut for printing
      the primary key of an object. 
    - Added a "reuse_dbh" option to D::OD::D::DBI, if enabled it caches and reuses
      $dbh using the dsn as the key.
    - Exposed the transaction mechanism built in the drivers at the object levels:
      D::OD::BO->begin_work now starts a global transaction across all drivers
      ending with a rollback or a commit on the same class.
    - Fix problem with prepare_cache in DBD::SQLite
    - Fix PerlCritic tests
      http://rt.cpan.org/Ticket/Display.html?id=37197
    - Fix problems under 5.10
      http://rt.cpan.org/Ticket/Display.html?id=30941
    - Fix test failures under Win32
      http://rt.cpan.org/Ticket/Display.html?id=24480
    - Pg *can't* handle REPLACE
      https://rt.cpan.org/Ticket/Display.html?id=38840
    - Fixed an issue where Pg would look into the wrong sequence if DBI has a
      'prefix' configured (used in a undocumented TheSchwartz feature)
      https://rt.cpan.org/Ticket/Display.html?id=41880
    - Added Oracle support, courtesy of Xiaoou Wu (Oracle)
      https://rt.cpan.org/Ticket/Display.html?id=41929
    - Added an "add_index_hint" method to D::OD::SQL to allow specifying
      a "USE INDEX" hint.
    - Added an experimental GearmanDBI driver that provides query coalescing
      using Gearman workers (to sit in front of a direct DBI driver).

0.05  2008.02.24
    - Added a new Data::ObjectDriver::ResultSet abstraction for building
      result sets with lazy-loading of the actual results. This allows for
      passing around a representation of a full result set (no limit, no
      offset, etc), and allowing callers to modify the set as needed.
    - search() now returns a subref blessed into the new D::OD::Iterator class.
      It's backwards-compatible (you can still call $iter->()), but it now
      supports $iter->next() as well.
    - Added a D::OD::SQL::add_complex_where method, for creating more complex
      WHERE clauses with boolean operations.
    - Added instrumentation/profiling for the memcached, Apache, and RAM
      caching drivers.
    - Improved "remove" support in the experimental Multiplexer driver.
    - Fixed an ordering bug with BaseCache->update: the cache is now updated
      after the fallback (a persistent store, usually) is updated, to prevent
      the cache being updated but the backend erroring out.
    - Let DSNs start with "DBI:" instead of only "dbi:"
    - Fix a bug where the iterator version of search() (search() called
      in scalar context) wasn't calling finish() on $sth. It was generating
      warnings on certain circumstances.
    - Fixed a circular reference when using has_a.

0.04  2007.05.02
    - Fixed a bug where single-PK classes were not returning the
      objects correctly sorted during a lookup_multi.
    - Added support for MySQL 'REPLACE INTO' syntax with a new
      $obj->replace() method.
    - Added a new trigger 'post_inflate'.
    - Fixed a minor issue (warning) with no_changed_flags in column_func()
    - Added has_a() construct to build linking methods between classes.
    - remove() returns number of affected rows, with DBI semantics.
    - Bulk inserting of data can now be done with the new bulk_insert()
      Class method.  It uses Postgres' COPY command or MySQL's 
      multi-value inserts to load data quickly.
    - The new() constructor for objects now accepts column name/value
      pairs which are passed to a new init() method.
    - The new init() method can be called on any object to set
      many parameters in one call.
    - This init() method can also be overridden, allowing for custom
      initialization code.
    - Added parens around terms within complex SQL conditionals, to allow
      even more complex conditions to be created.
    - Made the second argument to D::OD::SQL::add_select optional. It defaults
      to the value of the first argument (the column name).
    - Pass along $orig_obj (original object) when calling post_save/post_update
      triggers, even when the object hasn't changed.
    - A non-numeric value in a LIMIT now causes an exception.
    - Fixed a bug where calling SQL->add_join twice would create an invalid
      SQL statement.
    - More documentation!
    - Added more DOD::DBD::* options: sql_for_unixtime, can_delete_with_limit,
      is_case_insensitive, can_replace, sql_class.
    - Added an experimental Multiplexer class to direct writes to multiple
      backend drivers.
    - Added a generic end_query method, analogous to start_query, which is
      called after each query. Useful for profiling etc.
    - Text::SimpleTable is now loaded dynamically so that it's no longer
      a requirement for D::OD.

0.03  2006.08.05
    - Added an inflate and deflate mechanism to memcached caching. When objects
      are stored in memcached, they are now deflated to a hash reference
      containing only the column values; retrieving the object from memcached
      automatically inflates the object to the full representation. Classes
      can override inflate and deflate to store additional information in the
      memcached representation that's kept automatically up-to-date.
    - Added a SimplePartition driver, which helps to make partitioning, well,
      simpler. Still to come: documentation and a tutorial on partitioning.
    - Many, many bug fixes and improvements to the caching drivers.
    - Added detection of changed columns, such that only columns that have
      been changed are updated in an UPDATE statement.
    - Added a clear_cache method to the D::O::D::Cache::RAM class.
    - Added cross-DBD error handling, which maps local error codes to error
      constants in Data::ObjectDriver::Errors. The list of supported errors
      is pretty miniscule thus far (just one), but will be expanded as needed.
    - Added support for query profiling (Data::ObjectDriver::Profiler), which
      counts queries, calculates frequent queries, and can produce reports.
    - Added support for optional table prefixes, which simplifies setting up
      identical schemas in the same database.
    - Added an optional $terms argument to D::O::D::DBI->update, which can
      add additional terms to the UPDATE statement besides just the PK.
    - Added a D::O::D::DBI->begin_work method, and improved the commit and
      rollback methods.
    - Added a D::O::D::DBI->last_error method.
    - Added support for multiple JOIN clauses with a new D::O::D::SQL->add_join
      method.
    - Multiple OR values are now contained in an IN (...) clause rather than
      many joined OR clauses.
    - Added a for_update option to search, which allows constructing a
      SELECT ... FOR UPDATE query.
    - D::O::D::BaseObject->column is now removed and replaced with a
      column_func method, which returns a subroutine reference used to
      initialize the dynamically-created methods for each column. This allows
      some optimizations.

0.02  2006.02.21
    - Added Data::ObjectDriver::BaseView, a base class for creating "views"
      across multiple tables, or involving more complex aggregate queries.
    - Added trigger/callback support for common operations. See the
      Data::ObjectDriver documentation for more details.
    - Added GROUP BY support in Data::ObjectDriver::SQL.
    - Data::ObjectDriver::BaseCache->search now uses lookup_multi to do a
      very fast PK lookup, which will hit the cache first before the
      backend DB.
    - Fixed bugs with BLOB columns in SQLite driver.
    - Added connect_options option to Data::ObjectDriver::Driver::DBI, for
      passing in custom options for a DBI->connect call.
    - Data::ObjectDriver::BaseObject->remove now works as a class method.
    - Added Data::ObjectDriver::BaseObject->primary_key_tuple for retrieving
      the primary key value(s) for an object.
    - Added Data::ObjectDriver::BaseObject->refresh to reload an object from
      the database.
    - Added support for HAVING clauses in Data::ObjectDriver::SQL. For views
      that are not attached to a particular datasource, any terms passed in
      to the query will automatically be turned into HAVING clauses.
    - Improved the lookup_multi method for all BaseCache subclasses: we now
      allow the subclass to look up multiple values in the cache and return
      any already-cached items, then make a list of the remaining IDs and
      send them to fallback->lookup_multi.
    - Driver::DBI->lookup_multi will now use an OR clause to look up
      multiple values in one query.
    - Added lots of test cases.
    - Pod fix (Thanks to Koichi Taniguchi)

0.01  2005.09.23
    - Initial distribution.