=head1 WHAT'S NEW IN TANGRAM
=head1 RELEASE 2.10, 29 Mar 2006
This release is the first step in the major refactoring of Tangram,
which is API-backwards compatible. The differences between 2.09 and
2.10 are limited to namespace re-organisation. The test suite still
succeeds unmodified, but has been tweaked to suppress warnings that
deprecated modules are being used.
=over
=item *
Sync release with 2.09. svk this time made this extremely painful,
but we must forgive it because after all trying to support distributed
source management tool atop of a synchronised versioning filesystem is
pushing shit uphill.
=back
=head2 DEV. RELEASE 2.09_02, 21 Nov 2005
=over
=item *
Sync release with 2.08_06. svk made this far, far, too easy.
=back
=head2 DEV. RELEASE 2.09_01, 20 Oct 2005
=over
=item *
Renamed lots of modules to a grand new layout scheme.
Various classes, such as C<Tangram::Scalar>, C<Tangram::Set>, etc, are
now called L<Tangram::Type::Scalar>, L<Tangram::Type::Coll::Set>, etc.
However, so long as you C<use Tangram::Core>, C<use Tangram> or C<use
Tangram::Compat> before any of these old names are used, then C<@INC>
magic should be able to catch the inclusion and load the correct
module instead.
=item *
Added new C<use Tangram> import arguments C<:core> and
C<:compat_quiet>. to be documented.
=back
=head1 RELEASE 2.09, 29 Mar 2006
=over
=item *
Disable some tests on MySQL, that try to do things far too difficult
for its little brain.
=item *
Suppress a warning in the test suite.
=back
=head2 DEV. RELEASE 2.08_06, 21 Nov 2005
=over
=item *
Test suite changes: add a test for weakref, fix bad prefetch test that
was causing false negatives.
=item *
Make Scalar::Util the hard dependency that it is.
=item *
Fix the default C<get_id> function debug output, related to the test
for weakref. Also fix a bug in C<get_id> that caused bizarre erratic
behaviour in the scariest of circumstances.
=item *
Disabled nested join tests on SQLite.
=back
=head2 DEV. RELEASE 2.08_05, 20 Oct 2005
=over
=item *
Previous aggregation function change broke
C<$storage-E<gt>count($filter)>. Fixed.
=item *
Fixed outer joins again. They've even documented now! C<:D> See
L<Tangram::Storage/select>.
=item *
Fix F<Makefile.PL> so as not to deliver F<Changes.pod>.
=back
=head2 DEV. RELEASE 2.08_04, 13 Oct 2005
=over
=item *
A very nasty but also obscure bug affecting horizontal mapping and
polymorphic retrieval (or aggregation) was reported by Alex.
The fix for it has made remote expressions always include filtering
(via WHERE) on the type column as well as the ID column in most
queries. This should not affect the result of queries; it is just
passing on an assertion in the code base through to the database
back-end. The times it will affect results are when your currently
active schema does not match the one used to create the database and
you were previously relying on side-effects.
Examples of the query differences will follow here before the 2.09
release.
=back
This version has broken outer joins; to be fixed in the next release.
Also, this Changes.pod seems to being delivered to some random place
in your C<@INC>; sorry about that, the next release will exclude it
via the appropriate L<ExtUtils::MakeMaker> voodoo.
=head2 DEV. RELEASE 2.08_03, 31 Aug 2005
=over
=item *
Make F<Makefile.PL> output display modules that failed, rather than
letting C<ExtUtils::MakeMaker> do it for it.
=item *
Make L<Data::Lazy> 0.6 a "hard" requirement in the F<Makefile.PL>.
This is still slightly problematic as only version 0.5 will be
installed via the CPAN shell.
=item *
Support direct use of SQL in order clauses via C<order =E<gt> [ 'SQL'
]> (see L<Tangram::Storage/select>)
=item *
Support the SQLite 2 found in Ubuntu Hoary's C<libdbd-sqlite-perl>
package, which uses a connection string of C<sqlite2>.
=back
=head2 DEV. RELEASE 2.08_02, 16 Aug 2005
=over
=item *
Fix L<Data::Lazy> dependency version number
=item *
Allow one-column L<Tangram::Ref>
=item *
add C<$storage-E<gt>make_oid()> (See L<Tangram::Storage/make_oid>)
=item *
Various documentation fixes from Andres Kievsky
=item *
Make test suite fail gracefully (skip all) if disabled
=back
=head2 DEV. RELEASE 2.08_01, 29 Jun 2005
=over
=item *
Remove redundant confusing copyright messages; the whole work is
copyrighted, such notices should either be complete or absent IMHO.
=item *
Fix link to Tangram site and mailing list
=item *
Merge in a whole lotta Oracle fixes
=back
=head1 RELEASE 2.08, March 2005
This release was C<2.07_07> with a couple of minor bug fixes.
=over
=item *
L<Tangram::Dump> now checks (via C<$storage-E<gt>id_maybe_insert()>)
whether objects that it is saving should be inserted to the DB first,
which allows the practice illustrated by the F<t/musicstore/> test
suite to work.
=back
=head2 DEV. RELEASE 2.07_07, 24 Dec 2004
=over
=item *
added support for multiple test suites, each with their own schema, to
end the ballooning schema madness that was affecting the springfield
test suite.
=item *
reworked the way the DBMS date formatting worked again, adding
L<Tangram::CookedDateTime>
=item *
added support for L<Time::Piece> and L<DateTime>.
=item *
much work on database portability. The L<Tangram::Pg> back-end now
works very well, and I have also ported to SQLite (see
L<Tangram::SQLite>).
=item *
much more documentation, including L<Tangram::Intro>
=item *
reworked the test suite to have multiple sections
=item *
new test section for Object Persistathon (F<t/musicstore/> schema)
=back
=head2 DEV. RELEASE 2.07_06, 10 Oct 2004
=over
=item *
Support for SQL sequences. See L<Tangram::Schema>.
=back
=head2 DEV. RELEASE 2.07_05, 2.07_06
=over
=item *
lots of changes. full list to follow once I get the time to look at
the revision history :-}
=back
=head2 DEV. RELEASE 2.07_04, 20 May 2004
=over
=item *
Various POD modifications, including the main Tangram page, the
addition of the stub L<Tangram::Complicity> and the removal of bogus
C<Tangram::Dialect>
=item *
Enhanced TRACE messages
=item *
Added support for Oracle via L<Tangram::Oracle>
=item *
Added support for cursors that select no objects (just retrieve extra
columns, eg for aggregate selects)
=item *
`Fixed' F<t/datetime.t> so that it doesn't do PostgreSQL/MySQL
specific tests unless it is running on those databases
=item *
L<Tangram::Storage> and L<Tangram::Relational::Engine> now hold a
handle to the L<Tangram::Relational> object that created them,
allowing for various DBMS dependant hooks (used to rewrite column
types, format dates, etc)
=item *
Extra sanity checking in the core
=item *
The id_col feature in L<Tangram::Schema> now actually works :-).
Previously, there were many places where C<.id> was hard-coded.
=item *
Added C<-E<gt>in()> method, a L<Tangram::Expr> object that allows
direct use of the SQL C<IN ()> construct, rather than just via
C<includes_or()>, or with collections using joins.
=item *
L<Tangram::DMDateTime> now expects database to understand the ISO-8601
date format, but the MySQL driver has been enhanced to convert
ISO-8601 to MySQL format.
=item *
New method C<Tangram::Storage::unload_all> (see
L<Tangram::Storage/unload_all>)
=back
=head2 DEV. RELEASE 2.07_03, 17 May 2004
=over
=item *
Added L<Tangram::IDBIF> to combine multiple object fields into one
perl_dump style field (including a "catch-all" facility) Test suite
for this isn't complete yet.
=back
Most other changes from Kurt Stephens L<ks.perl@kurtstephens.com>, with
many thanks:
=over
=item *
Do not C<$self-E<gt>{db}-E<gt>disconnect> in
C<Tangram::Storage::disconnect()> and C<::DESTROY()>, if
C<$self-E<gt>{db} eq $opts-E<gt>{dbh}> in
C<Tangram::Storage-E<gt>connect($schema, $cs, $user, $pw, { 'dbh'
=E<gt> ... } )>; See C<$self-E<gt>{db_owned}>.
=item *
Handle objects that lazily initialize L<Tangram::AbstractSet>
L<Set::Object> fields. See UMMF.
=item *
Added L<Tangram::IntrRef> ("Intrusive Ref") for 1-to-1 bi-directional
Association navigation using only one column in one table. Derived
from L<Tangram::IntrSet>. I need help with this one. :)
=item *
Added C<@sums = $storage-E<gt>sum([ $expr1, $expr2 ], ...)> support.
=item *
Added support for reused tables for mapping both sides of many-to-many
UML Associations to the same table with L<Tangram::Set> and
L<Tangram::Array>. See UMMF's F<example/ex1>.
=item *
Added support for timestamping all tables through
C<$schema-E<gt>{sql}{timestamp_all_tables}>. See
L<Tangram::Schema/Global Properties>.
=item *
Prevent additional records being inserted to the Tangram control table
on C<-E<gt>redeploy>.
=item *
Added support for C<GROUP BY> in L<Tangram::Cursor>.
=item *
Added C<COUNT($remote)> and C<SUM($filter_expr)> support.
=back
Editorial release changes:
=over
=item *
Started using C<_NN> version style, so CPAN marks it as a development
release
=item *
added brief test suite for aggregate functions
=back
=head2 DEV. RELEASE 2.07.2, 6 Jan 2004
=over
=item *
L<Tangram::IntrHash> prefetch test added
=item *
Documentation enhancements
=item *
L<Tangram::Dump> class to `fix' persistent references in data
structures prior to insertion, so that :
=item *
L<Tangram::PerlDump> now can contain persistent objects safely
=item *
and hey, why not write a L<Tangram::Storable>
=item *
and a L<Tangram::YAML> (not working, waiting for a CPAN bug to be fixed)
=back
=head2 DEV. RELEASE 2.07.1, 29 Dec 2003
=over
=item *
L<Tangram::IntrHash> implemented at last :-). C<Tangram::IxHash> and
C<Tangram::IntrIxHash> are next :->
=item *
verbose error message in L<Tangram::Coll> when demand paging is
happening on stale tied collections (eg, erroneously created via
dclone())
=item *
fix for Postgres sub-select detection (Pedro Alves)
=item *
C<make_1st_id_in_tx> now only called for 1st insert (Chengzhi Liang)
=back
=head1 RELEASE 2.07 8 Nov 2003
=over
=item *
Fixed bug with C<Tangram::Intr*> and erasing objects that referred to
a class which didn't use their class name as their table name [closes:
2650]
=item *
Added the ability to use C<limit =E<gt> [ N, Y ]> in
C<Tangram::Storage-E<gt>select> statements (see
L<Tangram::Storage/select>)
=item *
Added C<Tangram::Coll::includes_or>; see L<Tangram::Expr>
=item *
Enhancement of error message from L<Tangram::Relational::Engine>
[related to 2631]
=item *
Support for using table types via C<$schema-E<gt>{sql}{table_type}>
=item *
Support for sub-selects with MySQL, for the one test that uses them.
Also C<$storage-E<gt>{no_subselects}> is auto-detected.
=item *
Completed C<$storage-E<gt>prefetch()> function for all working types,
new test F<t/prefetch.t>
=item *
If you have NULL values in array or hash slots in your database, then
it won't cause a hard error, except possibly with prefetch() [related
to 2631]
=item *
Various small fixes to bizarre internal error messages [related to
2631]
=item *
Minor documentation fixes
=item *
Exceptions on various DB failures are now thrown when
C<DBI::execute()> is called internally, rather than being largely
ignored.
=item *
Cosmetic Test Suite change
=item *
ID column name wasn't being used for the index in
L<Tangram::Relational::Engine> (Kurt Stephens)
=item *
Lost patches from
2002/06/25 (Kurt Stephens):
Added missing L<Tangram::Expr> and L<Tangram::Filter> operators:
'+', '-', '*', '/',
'&', '|', '!',
'==', '!=', '<', '>', '<=', '>=',
'eq', 'ne', 'lt', 'gt', 'le', 'ge',
'cos', 'sin', 'acos',
=item *
Lost patches from
2002/06/25 (Kurt Stephens):
Added argument swapping for non-commutative operators, like
"<"
=item *
Fixed export problem with 5.6.1 (reported by Kate Pugh)
=item *
C<deep_update> with L<Tangram::Ref> bug fix (reported by Aaron Mackey)
=item *
F<Makefile.PL> should now hopefully be more informative about which
modules are needed for what.
=item *
Removed C<Selfloader> from L<Tangram::Ref> (Kate Pugh)
=back
=head1 RELEASE 2.06 Jul 17 2003
=over
=item *
Added C<$storage-E<gt>oid_isa($oid, "ClassName")> method for
marshalling OIDs
=item *
Fixed minor bugs causing test suite to issue large warnings on
5.8.1-RC2
=back
=head1 RELEASE 2.05 Jul 2003
=over
=item *
L<Tangram::Scalar> and L<Tangram::Ref> column names are now passed
through C<$schema-E<gt>{normalize}-E<gt>()>
=back
=head2 DEV. RELEASE 2.05pre2 Jun 2003
=over
=item *
Minor L<Tangram::Expr> defined vs logically-false fix (Aaron Mackey)
=item *
explicitly select all C<ORDER BY> columns (Aaron Mackey)
=item *
normalise default "C<slot>" and "C<coll>" column names in
C<Tangram::Intr*> types
=back
=head2 DEV. RELEASE 2.05pre1 May 2003
First "not the momma" release.
Patches from lots of people were applied, don't know who 'coz the
honourable grand master JLL applied a lot of them, I'll try to
summarize and attribute them here:
=over
=item *
back-porting pieces for 5.00503 (that was me)
=item *
a more elegant solution for the
L<Tangram::AbstractArray>/L<Tangram::AbstractHash> corner case than
the one I submitted :-)
=item *
code that (possibly) completes L<Tangram::Hash> support (I haven't
written a test case to confirm this yet)
=item *
A fix to L<Tangram::AbstractSet> that again caters to some very
strange corner cases (this was me too)
=item *
Working implementation of the L<Tangram::DMDateTime> type mapping, a
good thing that running a persistent application server offsets the
huge time it takes to load in L<Date::Manip> :-)
=item *
A bug fix to L<Tangram::FlatArray> mapping type that affects deleting
members (don't know whose)
=item *
C<Tangram::IntrSet::get_intrusions> (? can anyone claim or explain
this one?)
=item *
Minor L<Tangram::PerlDump> fix (a simple one)
=item *
Deep, deep magic in L<Tangram::Relational::Engine>; new function
get_heterogeneity (anyone?)
=item *
C<Use of uninitialised variable in Tangram::Relational::Engine>...
=item *
Other large changes to L<Tangram::Schema> (? - get_bases, get_specs,
is_root, etc)
=item *
A problem with merging schemas when you used a `back-ref' to a class
that is later on in the schema (me)
=item *
Scattered showers of changes in L<Tangram::Storage>
(get_insert_statements etc, lots of `off by one' (reference nesting
level that is) changes)
=back
All I can say is that there have been some very smart people looking
at this code and submitting patches.
=head1 HISTORIC Changes
Originally as F<Changes> in the distribution
Revision history for Perl extension Tangram.
0.01 Sun Mar 21 10:46:45 1999
- original version; created by h2xs 1.18
0.03 Stardate: unknown
- first public alpha release
0.08 Stardate: unknown
- second public alpha release
0.09 Stardate: unknown
- split Tangram.pm into several modules
- replaced @ISA with 'use base'
- replaced default typetags with automounting
- prefetch now takes either class or remote
- implemented prefetch for IntrArray and IntrSet
*::prefetch still need real-life testing,
especially when class(coll) <> class(item)
1.00 Nov 07 1999 - first mature release version
- deploy() now also works directly on DBI handles
- deploy() allows specification of SQL types via the schema hash
- intrusive collections allow element-to-container navigation
- improve speed of read_object
- fixed various bugs
- rewrote part of, and expanded documentation
1.01 Nov 12 1999 - interim version
- added (raw) mapping between Perl strings and date/time SQL types
1.02 Nov 14 1999 - interim version
- added support for aggregated refs, arrays and sets
1.03 Dec 28 1999 - interim version
- added support for customizing object & class id types, etc
- as a consequence Tangram should now work with PostgreSql
- fixed bug: erase() failed in presence of stateless leaf classes
- temporarily removed RawDate and RawTime tests because M$ SQL
cannot handle them; those modules are still present though
1.04 Jan 03 2000 - interim version
- fixed two bugs related to explicit tables
1.05 Jan 03 2000 - interim version
- fixed: Storage::prefetch now works with inherited members
- fixed: _insert() ignored collections in stateless classes
1.06 Feb 29 2000 - higly experimental interim version
- added typechecks on collection members
- started cleaning up autovivification of colls upon save
- documented like()
- documented extended select options retrieve, order & filter
- changed _limit into the Limit in test suite
- added temporary hack in Makefile.PL to support Access
- added mapping for flat arrays
- factorized code common to Array & FlatArray
1.07 Feb 29 2000 - interim version
- adapted test suite for Sybase and other NULL-picky databases
- added experimental support for vendor extensions (Dialect)
- added skeleton Dialect::Sybase with support for datepart
- did extensive rework on Expr instantiation
- began documenting Type
- added Storage::reload()
1.08 Mar 26 2000 - interim version
- batch allocation of object ids
- support for Lobs contributed by marian.kelc@ruhr-uni-bochum.de
- Storage::reset()
1.09 ? - interim version
- String calls DBI for correct quoting, as Marian Kelc suggested
- fixed Array problem when removing elements from middle of collection
- (?)
1.10 Apr 19 2000 - interim version
- changed reset() into unload() and added support for partial unloading
- tx_do() can now propagate arguments to closure/anon-sub;
insert(), update() and erase() pass anon-subs to tx_do to prevent
leaks (thanks Radu Greab)
1.11 May 29 2000 - public version
- ordering result set by descending order is now supported
- die() on attempt to load from an invalid id
- removed silly table id allocation limit (t1000) in Coll.pm
1.13 Jul 11 2000 - interim version
- new FlatHash and DMDateTime mappings contributed by Gabor Herr
<herr@iti.informatik.tu-darmstadt.de>
- some more doc in Type.pod
- date & time doc moved to DateTime.pod
- deploy() puts \n between instructions in trace
1.15 Jul 21 2000 - interim version
- new PerlDump mapping contributed by Gabor Herr
<herr@iti.informatik.tu-darmstadt.de>
1.16 Jul 25 2000 - interim version
- fixed deploy() bug: didn't take col names into account for Array, Set and Hash
- fixed query_expr bug in Scalar and Ref: used field name instead of col name
1.17 Jul 28 2000 - interim version
- Gabor's patches on deep updates
1.18 Jul 28 2000 - stable version
- weakrefs!
- tuples
- mysql tx acquired lock regardless of tx nesting
1.19 Jul 28 2000 - stable version
- improved id allocation for mysql
- bitwise operators for mysql
- manual class id allocation
- install now works in the absence of DBI and Set::Object
- Makefile.PL prerequisites work now
- Guided Tour is now generated & can be run
- added support for LIMIT in selects
- deploy() didn't take explicit oid/cid type into account; fixed
- deploy() wrongly defaulted to NULLable types; fixed
- class id can now be specified
- oid and cid now default to INTEGER
- fixed error in Expr.pod but I don't remember what it was
- minor improvement to Schema.pod
- added Springfield.pm for users to play with
- added support for SELECT SUM
2.00 Feb ?? 2001 - first S.O.L. release
- added copyright notices
- added a copy of the GPL
- implemented inlined exports
- INSERT and UPDATE now make use of prepared statements when available
- re-executable Cursors
- changed database layout, Class Table is gone and 'control' table in there
- object ids are no longer allocated on a class basis
- introduced Tangram::Relational
- dialect are now done via derivation
- stored object id no longer contains type id
- introduced Relational::Engine
- flexible mapping
=cut