0.26 29. Sept 2004

    - fix "Can't upgrade that kind of scalar", so it now runs with Perl5.8.x

    - remove backticks around table names returned at least from newer versions
      of DBI::mysql and DBD::ODBC


0.25 30 Dec. 2003
    
    2003-12-07  U-MOKSHA\metaperl  <metaperl@urth.org>

	* Recordset.pm : added support for direct Query input.

    2003-12-02  U-MOKSHA\metaperl  <metaperl@urth.org>	
	
	* test.pl (DoTest): all tests now pass successfully for Postgres
	and SQLite
	
	
    2003-11-26  U-MOKSHA\metaperl  <metaperl@urth.org>

	* Recordset.pm (BuildFields): when building the FROM clause for
	LEFT JOINs Recordset placed parentheses around it, like this:

	SELECT
	dbixrs1.id,dbixrs1.name,dbixrs1.value1,dbixrs1.addon,dbixrs2.value2 FROM (dbixrs1 left join dbixrs2 on dbixrs1.id = dbixrs2.id) WHERE   dbixrs1.id IN (2, 5, 10); 

	but this caused a syntax error in sqlite, so a new leftjoin type
	of 4 was introduced to support LEFT JOIN with no parentheses
	around the join
	
	* added support to DBIx::Database so that it would accept a DBI
	database handle. the entire test suite for DBD::Pg now passes.

	* added the '$makesql' flag to Search() so that generated SQL can
	be returned instead of executed. It will remain undocumented until
	it is supported for Update(), Delete(), and Insert(). It was added
	in order that the SQL generated for the test "Order, Group,
	Append" would pass on Postgres which finds the generated SQL
	invalid. 
	

  - Make sure filters given by name override filters given by type.

0.24 10. July 2001

  - Added Code to DBIx::Database to Create/Modify/Drop tables
    See CreateTables/DropTables
  - *fieldname can take an array ref, to specify different
    operators for multiple values which should compared to the same
    field. This is handy for selecting a range.
  - Filters are correctly apply if a arrayref with multiple value
    are passed to a SELECT.
  - Reset error code and string in DBIx::Database -> new
  - If an array of values id passed in for one field and the operator
    is '=' now the IN sql operator is used, instead of a set of '='.
  - Set Postgres type 1005 to not numeric. Spotted by Michael Maruka.
  - Ignore errors in ListFields when retrieving metadata for a table.
  - Statement handle is closed as soon as possible, to avoid out of
    cursors situations.
  - new parameter $expr allow to group multiple sub expressions in 
    a sql where, therefore allowing more complex conditions.
  - Added new parameter !MergeFunc which allow to specify a function
    that is called, when multiple records with the same key are found
    in a DBIx::Recordset::Hash object.
  - Added some code to handle table- and fieldnames which include
    spaces and special charaters. This is still experimental.
  - Fixed problem with !TableFilter setup
  - Set correct brackets when creating an left outer join. This is necessary
    for some database to join more the two tables.

0.23  22. Sept 2000

  - separated DBIx base classes in different files, so they are visible and
    searchable by CPAN

0.22  22. Sept 2000

  - Fixed a problem that !PrimKey was not correctly set when
    TableAttr('!Serial') was set.
  - call $sth -> finish when attempt to read after the end of
    a fetched set of records to avoid problems with to many
    open cursors.
  - backlinks of automaticly deteced links are now prefixed
    with a star ('*') instead of one dash ('-') to avoid conflicts. 
    Also the prefix__ is used in backlinks now. Also if there is a
    prefix like foo__ it is included in the backlink.
    NOTE: This may break some software, but this step
    is neccessary for uniqueness. Otherwise there are situations
    where it can't be predict, which is the destination of 
    a given link.
  - Fix incorrect SQL expression when an empty array is given as
    parameter to an WHERE expression.
  - Added method Reset to set the current record to the initial
    state, so a call to Next returns the first record.
  - If DBIx::Database -> new cannot handle a table it will only log
    a warning, but won't die anymore.
  - Added new Parameter !PreFetch and !Expires which can be used
    when tieing a hash, which will prefetch the data from the
    database and storing it in memory, so many accesses to these
    keys will be much faster. It is also possible to give an
    expiration time (or function) after which the data is 
    refetched.

0.21  21. Juni 2000

  - Fixed problem with wrong StartRecordNo, which will also cause
    problems with prev and next. Spotted by Alexander Siegel.
  - Fixed problem that FETCHSIZE also returns one to much.
    Spotted by Robert.
  - Fixed the $last attribute of Search. Spotted by Roman Maeder.
  - Fixed a bug in FETCHSIZE spotted by Robert.
  - Allow different Filters for \\field and field in select.
  - Fixed a problem with bind_param and LONGVARCHAR together
    with DBD::ODBC
  - Applied patch from Rob McMillin to the documentaion, which
    corrects my bad english.
  - DBIx::Recordset now supports serials fields. It is able
    to insert a value from a sequence into a record upon insert
    and return the value (for databases which supports sequences)
    or simply return the value of last serial field (for databases
    which does support serial/autoincrment fields)
  - For DBMS with does not support sequences, there are two new
    classes (DBIx::Recordset::DBSeq/FileSeq) which emulates sequences
    in a DB table or the filesystem (see !SeqClass parameter)
  - Execute now retrievs always the correct record after an INSERT
    when a serial fields is specified. (and serial fields are supported
    by the DBMS)
  - Documented the $where parameter, which allows to pass a literal
    SQL WHERE expresseion. (See SYNOPSIS)
  - Added the $values parameter, so you can now pass parameters which
    should be bind to the placeholders in the expression given 
    with $where. This allows more complex WHERE expressions to be setup.
  - Wrote a short SYNOSIS sections, which gives some examples to make
    it easier to start with DBIx::Recordset.      
  - Multiple values for a single field could now be also passed as
    array ref.
  - !TableFilter parameter is now always used as prefix, so only
    thoses tables which starts with the prefix given in !TableFilter
    are deteced by DBIx::Database.
  - Removed the error message that zero record are updated instead
    of one, because this were sometimes not correct. Spotted by
    Roman Maeder.
  - Fixed a problem which occurs when you access a linked table
    for the first time (e.g. $set{-to}{foo}). Spotted by
    Andrej Mikus.
  - Removed a lot of -w warnings.
    
0.20  6. Jan 2000

  - Fixed a problem with perl 5.004_04. Spotted by Paul J. Schinder.
  - Added entry for DBD::Informix in Compat.pm send by Kelly Peet
  - Added outer join syntax for DBD::Informix
  - undef in !Filter passed to Recordset constructor overrides
    filter-function for TableAttr.
  - Fixed a problem with PrevNextForm and MoreRecord spotted by
    Richard Chen and David M. Davission.
  - Fixed some problems with eg/search.pl with help from
    Richard Chen.
  - Added date and time datatype to PostgreSQL quoting with help
    from James Bishop.
  - Added LastError function which returns the last error message
    and code. This solves problems where the DBI object is already
    out of scope and the $DBI::errstr isn't available anymore.
  - Now all setup parameters (the ones starting with !), can also
    be preset via DBIx::Database -> TableAttr.
  - Fix a problem, that an error that sets $@ inside of DESTROY
    will clear out an earlier error message. Based on a patch from
    Dean Brettle.
  - !Filter can take a thrid argument, which will cause
    DBIx::Recordset to always execute the Input Filter
  - Fixed a problem inside Flush, which will report sometime
    'Not a scalar reference..'
  - DBIx::Recordset handle uppercase field and tablesnames correctly
    now. This is important to work with Oracle.
  - New Parameter !TableFilter allows you to specify only a subset
    of tables from the database/tablespace to work with 
  - !TableFilter is taken in account as prefix when searching
    for links
  - Added NeedNullInCreate for ODBC to work with MS SQL Server
  - Recordset Object didn't get the DBHandle from Database object if
    specified as !DataSource. Now it does. (Database object must
    be created with !KeepOpen)
  - !Filter can take rqINSERT or rqUPDATE as Argument after subref, 
    which causes the Filter to always execute on INSERT and/or UPDATE,
    regardless, if there is a value for the field or not.
  - Added methods TableName, TableNameWithoutFilter, TableFilter and
    PrimKey.
  - Support multiple Database object with different !TableFilter.
  - !LongNames is now also supported for SELECTs which select not all,
    but a specified number of fields (i.e. fields are ne '*').
  - DBIx::Database TableLink method, returns all Links for a given 
    table when no linkname is given.
  - Added more tests to ensure that new feature do not break old ones.
  - removed call to non existing method $dbh->begin and call 
    $dbh->commit and $dbh->rollback only when AutoCommit is off, so
    the DBIx::Recordset transaction methods, can be called regardless, if
    the DB supports transactions or not. Patch from Alexander Smishlajev.
  - make clean removes test.log and make realclean remove privious
    configuration. Patch from Alexander Smishlajev.
  - !DataSource can be a DBI database handle. Patch from Alexander Smishlajev.
  - New method Dirty check if there is at leat one dirty row in a recordset.
    Patch from Alexander Smishlajev.
  - DROP test tables at end of test. Patch from Alexander Smishlajev.
  - DBIx::Database now makes an connect to the database if KeepOpen is set
    and the object is taken from an already stored one.
  - Flush will call finish on open statement handles to avoid statement
    handles that stay open to long.
  - Add new parameter !DoOnConnect, which lets you specify any SQL statement
    that should be execute after each connect.
  - PrevNextForm could now take a hash ref as parameter and
    addtionaly generate first, last and goto buttons
  - Documentation updates.
  - Added patch from Klaus Reger to avoid warning about undefined value
    in DESTROY.
  - Added more datatypes to quote for Pg and Oracle. Patch from Klaus Reger.
  - Undef values in where expression are transfored to "is null" operators,
    also when using placeholders.
  - DBIx::Recordset now uses always bind_param, to avoid problems with quoting
    and to speed up things. This means DBD drivers which does not support
    placesholdes will not work anymore with DBIx::Recordset, but all 
    DBD drivers I know, supports now placesholder, so there should be no problem.
    This also remove an potetial insecure eval form the BuildWhere function.
  - DBIx::Recordset::Flush will update all dirty rows, also if some of them
    can't be written to the 
    database. Patch from Alexander Smishlajev.
  - DBIx::Recordset::Row::Flush make sure that exactly one row is updated. Based
    on a patch from Alexander Smishlajev.
  - By setting $FetchsizeWarn to zero, you can now make DBIx::Recordset return
    the value of $sth -> rows, when the FETCHSIZE method is called by Perl.
    This feature is still experimental, because I don't what different DBMS
    returns in $sth -> rows, after an SELECT. Patch from Robert.
  - By setting $PreserveCase, you can cause DBIx::Recordset to _not_ translate all
    names to lowercase. Per default DBIx::Recordset translates all names (tablenames/
    fieldnames) to lowercase to let application work with different DBMS.
  - Updated docs of DBIx::Compat.
  - If the DBMS supports it, DBIx::Recordset will use the LIMIT/OFFSET keywords if you
    only want to fetch a limited number of row from the database (for parameters 
    $start and $max). Patch from Robert.
  - Added an indroduction to DBIx::Recordset (based on my talk on the 1999 Perl conference)
    which is viewable with perldoc Intrors.pod


0.19  30 Mar 1999

  - changed order of group by and order by in select. Spotted by
    Roman Maeder.
  - DBIx::Recordset automaticly tried to determinated links between
    tables (see also DBIx::Database)
  - DBIx::Recordset will only insert a new record in the DB if
    a record contains any data. This avoids inserting a record
    that is only created when try to read an non existent record.   
  - Added new attribute !WriteMode which lets you specify which
    write operation to the db are permited and which are not.
  - Added new attribute '!LinkName' which will automaticly select
    additional fields from links tables (DBMS must support LEFT JOIN)
  - Fixed a problem that occurs in cleanup when you use muliple
    nested links. (while (($k, $v) = each (%h)) is not reentrant!!)
  - Execute will not do a Search after INSERT/UPDATE/DELETE when an
    error has occured, so the error message is not overwritten
  - use croak instead of die
  - SQLDelete accepts empty WHERE, which will delete the whole table
  - Added new object DBIx::Database which will parse the metadata of
    the database and tried to automaticly determinate links between tables.
    All the information is stored for later access by DBIx::Recordset
    objects. This makes setup of a DBIx::Recordset object much faster,
    especially when you use it in the startup file of your web server.
    Also this gives you the possibility to speficy table attributes
    only once at startup time.
  - New method Links return the links of a DBIx::Recordset object
  - New method Link4Field return the link of the specified field if any.
  - tied hash now also support DELETE and CLEAR operation on a table.
  - Insert now skips undef values, because there is no need to insert
    NULL values.
  - Update now set undef to NULL, even for database that does not use
    placeholders.
  - Added Filters. Filters allows you to specify an input and/or an
    output function which transforms the correspondig fields before
    input/output. This could for example be used to transform database
    specific datatypes (e.g. date) to and from a common format, or just
    to a human readable format.
  - Fieldnames which are given to Update and Insert maybe prefixed with a
    \ to avoid any transformation. Suggested by Frank Ridderbusch.
  - Update and Insert converts numeric values to vaild numeric format, e.g.
    'foo' will be converted to 0 when inserting in an integer field. Use
    '\name' => 'foo' to avoid this conversion.
  - 'null' and 'not null' now allowed as unary operators in where expression.
    Suggested by Malcolm Cook.
  - An value of undef is now allowed as key for the hash access. This is
    usefull when your primary key is an autoincrement value. Then you can
    add data with undef as key and call Flush to write out the record.
  - automaticly finish a statement handle, when all data is fetched to
    avoid unneccesaary open statement handles.
  - Make all the new features are working with DBD::Orcale. Many thanks
    to John Tobey for his help.
  - Adapted Compat.pm entries for DBD::Sybase. Help and testing from
    Malcolm Cook and Aaron Ross.
  - $DBIx::Recordset::Debug now allows more values, so you can better
    control what's being logged.
  - Added !LongNames parameter which forces the hash keys to the form
    table.field.
  - Added methods for begin/commit/rollback to make sure DBIx::Recordset
    internal data are correctly updated. Suggested by John Tobey.

0.18  3 Nov 1998 
  - Added !TabJoin parameter to support LEFT/RIGTH/INNER JOIN
    syntax. Spotted by Roman Maeder.
  - Added $group parameter to support GROUP BY
  - Added $append parameter to append arbitrary data to the end
    of an SELECT statement.
  - Added !IgnoreEmpty which gives you the possiblity to 
    specify how DBIx::Recordset handles fields with undef value,
    or empty strings. This may helpfull in a CGI enironement.
    NOTE: with a value of 2 DBIx::Recordset behaves like versions
    before 0.18
  - fixed problem with handling of undef and zero values.
    undef is now treated as SQL NULL and zero is handled
    correctly. See also !IgnoreEmpty.
  - Added entry for DBD::Oracle from Malcolm Cook
  - fixed a bug in PrevNextForm spotted by David Crook
  - The first Next after a sucessfull search returns now
    the first row and not the second one, spotted by Malcom Cook.
  - fixed a bug so that now linked tables a correctly flushed if modified

0.17  8 Oct 1998
  - DBIx::Recordset now can handle subtables. This means if you have
    one tables which has field which is an key for another table
    you are able to access the second table via $set{first}{subfield}
    DBIx::Recordset handles this by creating a special link field,
    which is actually a new Recordset object.
    Look at "Working with multiple tables" in the documentation.    
  - You can now call the method add to get an empty new record and
    then simply put data in the current record 
  - Metadata is cached, that means if you open the same table a
    second time DBIx::Recordset does not query the driver again
    for the metadata, but uses the cached values instead.
  - The Flush methods returns now undef on error, otherwise a true
    value
  - Adapted the data types for newer mSQL drivers in DBIx::Compat
  - Added documentation about working with mulitple tables
  - Added documentation about debugging and DBIx::Recordset logfile
  - DBD::CSV works together with DBIx::Recordset

0.16  26 Aug 1998
  - Reworked memory management
    1.) DBIx::Recordset will be correctly destroyed and the statement handles 
        released, if the object goes out of scope. (You don't need 
        DBIx::Recordset::Undef anymore to destroy an object)
    2.) Worked around a problem that perl will reference an object when a
        reference to it is return from the each operator. This causes reference count
        of the object not going to zero and therfore the object was not destroyed
        when it goes out of scope.

  - Ignore DBD::File for tests because it's just a base class
  - Added support for drivers which does not have field types 
    ($sth -> {TYPE})
  - Fixed return values from Select/Update/Insert/Search/Execute so
    they return undef on error. Spotted by Gary Ashton-Jones.
  - Documentation update

0.15  31 Jul 1998
  - Apdapt STORE method of tied array to perl5.005. DBIx::Recordset works
    now correctly with perl5.004 and perl5.005
  - Update can now handle updates of the primary keys correctly
  - Insert can be done by calling Setup and then write values to the
    array
  - Update works now without an primary key
  - Adapted Compat.pm to the new values in $sth -> {TYPE} of DBD::mSQL
    in 1.19_19 and higher. Spotted by Ray Zimmerman,

0.14  1 May 1998  
  - first public release