1.66 2014/08/20
- No changes since 1.65_02
1.65_02
- Stop unilaterally disabling the "UTF8" flag before executing queries
- Make ->Fields case-sensitive in the column names it returns, as well
as in the table name it takes.
1.65_01
- Make ->Fields only lookup information on the table requested, not all
fields, for performance. It also is now case-sensitive in table name.
- Omit calls to ->Fields entirely for PostgreSQL 9.1 and above
Do it for all packages that
* mention perl, or
* have a directory name starting with p5-*, or
* depend on a package starting with p5-
like last time, for 5.18, where this didn't lead to complaints.
Let me know if you have any this time.
1.65 Wed Jul 3 16:40:48 PDT 2013
Ruslan Zakirov:
* Bug fix for DateTimeInterval extraction on Pg
1.64 Mon Jul 1 10:20:43 PDT 2013
No changes since 1.63_03. Simply a non-dev release of everything since 1.63.
1.63_03 Fri Jun 14 13:11:01 PDT 2013
Ruslan Zakirov:
* warn when rollback and commit are mixed
* Handle->NullsOrder
* skip timezone tests on SQLite when tzinfo is not there
* skip tests if mysql can not do timezones
* DISTINCT argument in Join method
* DISTINCT argument in Join and NewAlias
Thomas Sibley:
* Reset the iterator position whenever a search is run
* Return the correct record from ->Last instead of the first record
* Document the caveat of using GotoItem with a non-zero N
1.63_02 Wed Apr 17 18:47:31 MSK 2013
* _Set now can take undef as argument to mean default or NULL.
Still may result in error if default is not defined and no_nulls
is true for the column. If old behaviour is required set
$record->{'no_undefs_in_set'} to true value.
* FUNCTION argument is now allowed in Limit. Code to combine
FUNCTION, ALIAS and FIELD was refactored and unified in one
place - CombineFunctionWithField method. Used in Column, GroupBy
and Limit. This change should be backwards compatible.
* Handle->DateTimeIntervalFunction
1.63_01 Wed Mar 27 13:02:35 PDT 2013
* IN and NOT IN operators in ->Limit method
* Add an AdditionalColumn method to collections
* Add an AS parameter to Column method in collections
* Consistent query generation by sorting hash keys/values
1.63 Fri Sep 14 2012 01:19:38 GMT+0400 (MSK)
* joins_are_distinct hint to indicate that distinct is not
required for the current set of joins.
1.62 Mon Mar 26 09:31:05 UTC 2012
* Bind values were ignored in SimpleUpdateFromSelect
a) refer 'perl' in their Makefile, or
b) have a directory name of p5-*, or
c) have any dependency on any p5-* package
Like last time, where this caused no complaints.
Changes from previous:
1.61 Fri Sep 16 15:47:50 MSD 2011
* New methods in Handle for mass changes from select statements:
InsertFromSelect, DeleteFromSelect and SimpleUpdateFromSelect
* New methods in Handle for generation of date time related SQL
1.60 Thu Sep 15 01:01:15 MSD 2011
* custom BuildDSN for Oracle
** Database is treated as SID if SID is not provided
** Build 'dbi:Oracle:<SID>' instead of 'dbi:Oracle:sid=<SID>'
* changes in DBIx::SearchBuilder->Column method
** complete documentation
** support for empty FIELD argument
** column naming fix when explicit ALIAS => 'main' passed
1.59 Fri Nov 19 13:45:01 MSK 2010
* DBIx::SearchBuilder->DistinctFieldValues method
1.58 Wed Oct 20 02:17:37 MSD 2010
* SIGNATURE fix
* delete obsolete cvs metadata from a module
1.57 Mon Sep 4 21:21:57 UTC 2010
* INCOMPATIBLE CHANGE: NextPage and PrevPage were adding rows from
the previous page. Jesse claims that when he wrote this code, he
was 20 years old and it seemed like a good idea at the time.
* When logging queries, include full stack trace
* support $sb->NewAlias( 'table' => 'LEFT' );
* allow join to depend on nothing
* catch cases when there are more closing parens then should be
* Oracle: Use ROW_NUMBER() to propagate row ordering from inside the DISTINCT
* Various performance improvements through small internal refactorings
* Implemented 'sub Fields' on Oracle
* unify case insensitive characters to avoid using LOWER() in some
cases
* We now RedoSearch when RowsPerPage is changed
* No longer RedoSearch if FirstRow is called, but is not actually changed
* Document all paging functions and test them
* handle LOWER() in redundant LEFT joins optimizer,
for Oracle and may be Pg
* Make debugging problems easier by passing errors back
https://rt.cpan.org/Ticket/Display.html?id=55203
* fix Record->PrimaryKeys, field names in values hash are lc'ed
https://rt.cpan.org/Ticket/Display.html?id=18280
* doc updates and cleanups
to trigger/signal a rebuild for the transition 5.10.1 -> 5.12.1.
The list of packages is computed by finding all packages which end
up having either of PERL5_USE_PACKLIST, BUILDLINK_API_DEPENDS.perl,
or PERL5_PACKLIST defined in their make setup (tested via
"make show-vars VARNAMES=..."), minus the packages updated after
the perl package update.
sno@ was right after all, obache@ kindly asked and he@ led the
way. Thanks!
- Updating package for p5 module DBIx::SearchBuilder from 1.55 to 1.56
- Removing core dependency to File::Temp
- Setting Module::Install as module type
Upstream changes:
1.56 Fri Jul 17 02:05:32 MSD 2009
* Don't use LOWER/ILIKE with dates, heuristic is used, but shouldn't
harm other things
* Don't apply DISTINCT on queries with group by, COUNT(DISTINCT x) is
different and covered in Column method
- Updating package for p5 module DBIx::SearchBuilder from 1.54 to 1.55
- Setting gnu-gpl-v2 as license
- reordering dependencies
Upstream changes:
1.55 Thu May 7 19:44:47 MSD 2009
* Put test suite SQLite databases inside of tempdirs so they get
garbage collected properly. Thanks to Andreas Koenig
[rt.cpan.org #41322]
* Allow ->Join to pre-existing collection object
* Imlement and test SB::Handle::Fields
* Pg can not guaranty order in the following queries:
SELECT ... FROM (SELECT... ORDER BY ...)
we use them to build distinct sets with ordering by columns
in joined tables. Switched to group by instead of sub-selects.
to trigger/signal a rebuild for the transition 5.8.8 -> 5.10.0.
The list of packages is computed by finding all packages which end
up having either of PERL5_USE_PACKLIST, BUILDLINK_API_DEPENDS.perl,
or PERL5_PACKLIST defined in their make setup (tested via
"make show-vars VARNAMES=...").
- verify the dependencies added: p5-File-Temp, p5-Encode
ChangeLog:
1.54 Wed Jul 9 09:34:25 EDT 2008
When aborting transactions, we need to flush our cache,
because SQLite is reusing the primary id for later inserts and the cache
can otherwise become inconsistent.
1.53 Tue Apr 2 03:06:56 UTC 2008
* Fix mysql version check in DistinctQuery function
* Fix order by outer column on Oracle
* Improve tests
1.52 Tue Apr 1 00:48:56 UTC 2008
* Fix order by outer column on SQLite, mysql, adjust Pg. Add test that
cover this.
1.51 Tue Jan 15 22:53:56 UTC 2008
* Fix CountAll method when paging is enabled and data is in memory already
1.50 Fri Nov 23 23:24:00 UTC 2007
* Oracle: Don't DISTINCT query when there is a group by clause
* Fix a problem when we have more then two collections in a union
and some of them are empty
1.49 Sat Jul 7 18:45:41 EDT 2007
* Fix a CPAN signature issue
1.47 Sun Mar 4 03:30:00 UTC 2007
* Do the search in unions only when we must do them, not on every
call to the Next method
* Don't index ex/ dir to avoid complains by the indexer of PAUSE/CPAN
1.46 Sun Feb 25 19:04:00 UTC 2007
* when doing a union, we need to actually search, rather than just
doing a count
* add support for testing with Oracle backend
* Use CROSS JOIN instead of ',' as SQL parsers in Pg and some mysql
are buggy and cannot parse "FROM X, Y JOIN Z ON Z.f = X.f"
* deprecate DEBUG method, it's still there but produce warning
* fix CleanSlate method that was missing several keys
* fix a long standing bug we had, we didn't write depends_on data about
a join, so we could build queries with incorrect parens around join
conditions
* fix default values for ALIAS1 argument in the Join method, istead of
defaulting FIELD1 to 'main' value
* fix a TODO test
* internal refactoring of a storage for query's conditions,
instead of building query strings right after the limit
or join, we now build a perl structure
* don't clone attributes that don't exists in the Clone method
* we use Encode module without perl version check for a long time, so
we can get rid of all checks for the version and load the module
at compile time everywhere we need it
* implement MayBeNull method in the handler that checks if applied
conditions allow NULLs in the result set
* implement cascaded LEFT JOINs optimization
* additional tests for CleanSlate and Clone methods, ENTRY_AGGREGATOR
argument, different types of joins and LEFT JOIN optimizer
developer is officially maintaining the package.
The rationale for changing this from "tech-pkg" to "pkgsrc-users" is
that it implies that any user can try to maintain the package (by
submitting patches to the mailing list). Since the folks most likely
to care about the package are the folks that want to use it or are
already using it, this would leverage the energy of users who aren't
developers.
1.38 Thu Dec 29 03:17:54 EST 2005
* Released 1.37 dev series
1.37_01 Thu Dec 8 15:56:50 EST 2005
* Switched Postgres sequence lookups to use CURRVAL, rather than OIDs
1.36 Fri Dec 2 18:04:21 EST 2005
* Change to how we resolve virtual columns to deal with a
"no such attribute" bug in RT
1.35 Wed Nov 2 22:36:02 EST 2005
* Doc fixes and OrderBy cleanup from ruslan
1.34 Wed Nov 2 22:26:15 EST 2005
* Clone support from Ruslan
1.33 Thu Sep 22 14:27:46 EDT 2005
* Better SQL statement logging from alex
1.32 Thu Sep 1 06:52:42 EDT 2005
* DBD::SQLite is necessary for the test suite to run correctl
1.31 Fri Jul 29 12:47:25 EDT 2005
* Updated MANIFEST to fix a build issue -
Thanks to Andy Lester and David Glasser
1.30 Thu Jul 28 10:17:27 EDT 2005
* Removed {{{ and }}} fold markers. Patch from Ruslan
1.30_03 Thu Jun 9 01:35:49 EDT 2005
* Significant new tests from Ruslan Zakirov and Dave Glasser
* You no longer need to explicitly bless a DBIx::SearchBuilder::Handle subclass
* Start of a major overhaul of the subclass API for DBIx::SearchBuilder::Record objects.
A new "schema" method will define the data in _ClassAccessible and also generate database
schema using DBIx::DBSchema.
Fixes from Ruslan:
* for numeric types, make the empty check be "null or 0", not "null or ''"
* New search tests from ruslan
* added an init_data method to t/utils.pl
* CleanSlate doesnt init show_rows
* CleanSlate doesnt clean _{open|close}_parens
* get rid of stupid ifs in CleanSlate
* get rid of evals in _DoSearch and _DoCount, use Handle methods to control DBI error handling
* rewrite LoadByPrimaryKeys args handling to consistent with other Load* methods
* report error when PK filed is missing in LoadByPrimaryKeys
* fix warning in __Set methods when newvalue is undef
* small code cleanups
* test coverage grows from 75.2% to 84.7% for Record.pm
1.30_02 Sun May 22 15:21:19 EDT 2005
- Lots of patches from Ruslan:
First and main change is using of `goto &$AUTOLOAD` syntax, that helps
avoid code duplication and hides AUTOLOAD sub from stack trace. I think
this also would help implement CompileAllAutoSubs method easier.
- It's also one of the steps to better tests coverage.
- Test coverage for Record.pm grows from 66% to 75.2%.
- _LoadFromSQL never reported error when PK fields are missed. Fixed.
- fetchrow_hashref dies only when RaiseErrors is true, because we can
control this from Handle obj so we should die according to
$Handle->RaiseErrors property. Fixed.
- When RaiseErrors is "false" then fetchrow_hashref returns undef and we
should check $sth->err(see `perldoc DBI`). Fixed.
- After call to fetchrow we should clean "fetched" internal hash and fill
it only when we return successful result. Fixed.
- If SimpleQuery fails, _LoadFromSQL method doesn't return any error
message. Fixed.
1.30_01 Mon May 16 21:37:03 BST 2005
- Patches from Ruslan to switch to using 'capitalization.pm' for our regular_case subroutine aliases
1.27 Sun May 8 22:49:30 EDT 2005
- Added supoprt for functions containing "?" to represent the parameter
in ->Column()
- Added better support for functional columns in search listings and
group by clauses
1.26 Sun Apr 17 19:22:23 EDT 2005
- Added support for expression based left joins
1.25 Sat Apr 9 12:33:30 EDT 2005
- Backed out a change introduced in 1.23 that caused table and column names
to be quoted, causing Postgres to flip out.
1.24 Wed Apr 6 22:54:37 EDT 2005
- Added a new "SearchBuilder::Unique" module for uniquifying search results
1.23
- Now use DBI->quote_identifier to quote column and table names (Ruslan)
- Test suite updates (Ruslan)
1.22 Mon Jan 24 07:42:46 EST 2005
- Require encode since we require encode.
1.21
- Oracle LOB handling caused us to corrupt item values on update.
- Just before inserting things into the database, turn off their utf8 flag.
The flag didn't have any positve impact _and_ it can take down
recent DBD::Oracle releases. (This is a new failure in DBD::Oracle 1.16)
1.20 Tue Jan 18 08:24:15 EST 2005
- Minor test suite fixes from Ruslan.
1.19 Sat Jan 8 18:22:59 EST 2005
- Performing a search multiple times could result in multiple copies of
records in a collection. Uncovered thanks to Kevin Chen and Alex Vandiver.
1.18
- Release the changes from 1.17
1.17_03
- Properly mark BLOB columns in UPDATE calls. (DBD::Oracle 1.16 broke without this)
1.17_02
- Better handling of empty values for SB::Record::_Accessible. ( --Ruslan)
1.17_01
- More record tests from Ruz
1.16 Thu Dec 9 23:49:39 EST 2004
- Fixed a bug in D::SB::R::Cachable that could cause it to load the wrong row from the cache if you were loading
by alternate keys and had since changed one of the attributes of a previous row. This was unmasked by a
bug that Ruslan Zakirov found in RT 3.3's custom field handling
module directory has changed (eg. "darwin-2level" vs.
"darwin-thread-multi-2level").
binary packages of perl modules need to be distinguishable between
being built against threaded perl and unthreaded perl, so bump the
PKGREVISION of all perl module packages and introduce
BUILDLINK_RECOMMENDED for perl as perl>=5.8.5nb5 so the correct
dependencies are registered and the binary packages are distinct.
addresses PR pkg/28619 from H. Todd Fujinaka.
Add HOMEPAGE pointing into search.cpan.org.
Add DEPENDS on p5-Cache-Simple-TimedExpiry, as that appears to be required
to have "make test" complete.
Change log:
1.15 Sat Nov 27 13:09:56 EST 2004
- Fix a testsuite bug when DBD::SQLite isn't there
1.14
- Silenced warnings about uninitialized warnings when inserting null cols
into the database.
- Started adding lowercase method name aliases
- Minor refactoring of 'id' method for a stupid, tiny perf improvement
- Refactoring of DBIx::SearchBuilder::Record::Cachable for performance
improvement
- Added a FlushCache method to DBIx::SearchBuilder::Record::Cachable.
- Started to flesh out a...test suite
- SearchBuilder now truncates strings before inserting them into character
types in the database as mysql generally does. Additionally, it truncates
things at utf8 character boundaries...as mysql does not.
- Fix for an undefined record cache warning on load from Autrijus Tang
- Major documentation cleanups --Simon Cavalletto
- A few tweaks to the ::Record class to eliminate the
hard-coding of the name of the id column --Simon Cavalletto
1.12
- Better error handling for some query build failure cases
- Corrected query builder for SQLite
- More refactoring.
1.11
- When loading an object whose "id" has been altered, as in
the case of RT's "Merge" functionality, the wrong object was
returned by the caching layer. Special casing for the "id"
method was removed.
1.10
- Identical to 1.10_05
1.10_05
- Reworked the _Accessible mechanism in DBIx::SearchBuilder::Record to
remove a horribly crufty old caching mechanism that created a copy
of the accessible hash for each and every object instantiated,
sometimes quite slowly.
1.10_04 Mon Aug 30 17:33:18 EDT 2004
A query builder fix for an issue that bit RT2:
Unsatisfied dependency chain in Joins Users_2 at /usr/local/share/perl/5.8.3/DBIx/SearchBuilder/Handle.pm line 965, line 69.
Stack:
[/usr/local/share/perl/5.8.3/DBIx/SearchBuilder/Handle.pm:965]
[/usr/local/share/perl/5.8.3/DBIx/SearchBuilder.pm:326]
[/usr/local/share/perl/5.8.3/DBIx/SearchBuilder.pm:119]
[/usr/local/share/perl/5.8.3/DBIx/SearchBuilder.pm:410]
1.10_03 Mon Aug 30 14:31:10 EDT 2004
- Cache Sanity fixes from Autrijus Tang <autrijus@autrijus.org>
1.10_02 Thu Aug 26 13:31:13 EDT 2004
1.10_01 Thu Aug 26 00:08:31 EDT 2004
- Reimplemented DBIx::SearchBuilder:::Record::Cachable
to use Cache::Simple::TimedExpiry. This should make it faster and more
memory efficient.
1.02_03 Thu Jul 22 13:29:17 EDT 2004
- Additional bullet proofing for joins.
Now we default to ALIAS1 being "main" (cubic@acronis.ru)
1.02_02 Tue Jul 20 13:06:06 EDT 2004
- Fixed a join bug that mostly manifests as a 'Dependency chain'
error on RT2.
1.02_01 Wed Jul 7 12:28:08 EDT 2004
- magic _Object instantiation from cubic@acronis.ru
- make SB::_Handle settable directly (cubic@acronis.ru)
- document the above
1.01 Sun Jun 27 23:35:46 EDT 2004
- Releasing 1.00_06 as stable
1.00_06 - Pg/Oracle: Don't attempt to do case insensitive comparisons on
integer values.
1.00_05 - Force utf8 flag on when doing searches for utf8 data; this
is a workaround for DBDs that don't do it themselves.
1.00_04 - Move Postgres specific join behaviour to the superclass so
everyone gets the benefit.
1.00_03 - Remove "AS" from table name aliases on joins, since Oracle
doesn't like em.
1.00_02 - Slightly cleaner code in SearchBuilder->GotoPage
1.00_01 - Better handling of case insensitive comparisons on Postgres
- Proper support for query paging on SQLite
Patch provided by Dieter Roelants in PR pkg/25132
0.99 - Bundled changes from 0.98* and released production version
Removed duplicate code in cache expiry routines
Experimental SearchBuilder::Union collection object.
Released at the YAPC::Taipei::22004 Release Party
0.98_04 - New mysql/oracle "Join" code that allows more complex bundling of
joins from Linda and Robert
0.98_03 - New test infrastructure from Andy Lester
0.98_02 - Better handling of != clauses on Postgres
0.97_02 - Support for "Group By" clauses.
Support for delayed load of certain columns from Autrijus Tang.
0.97_01 - Oracle doesn't support binary-safe clobs in a reasonable manner.
0.96 - Releasing 0.96_01 as usable
0.96_01 - Fix a couple of spurious warnings in Record::Cachable
Records loaded from multiple-record searches were never cached
correctly
0.95_03 - Allow case-insensitive loading by columns in SearchBuilder::Record
- Record::LoadByCols now lets you specify operator and values
0.95_01
- Removed historical escaping for non-ascii searche queries
0.94- - Fix for multiple handles in one app from Autrijus Tang
0.93
- Added ODBC database driver from Autrijus Tang
- Added the ability to sort on functions of columns from Autrijus Tang
- Improved case-insensitve searching behavior for PostgreSQL
- Added support for multiple handles in one app from Autrijus Tang (#4167)
- Added initial Informix database driver from Oliver Tappe
0.92 Sept 4, 2003
- Fixed a bug that caused certain types of pre-canned table aliases to fail to work on join
0.90 Aug 8, 2003
- Disable Class::ReturnValue's stack trace feature as it interacted poorly with a stack containing lots of data
0.89_02 July 19, 2003
- Patch from Grant DeGraw to allow ordering by multiple columns.
0.89_01 July 18 2003
- Patch from Brook for:
- better oracle support
- remove "SELECT DISTINCT" when it's not necessary
0.88 June 23 2003
- More correct generation of "Distinct" keyword on counts for queries with left joins
0.87 June 16 2003
- Changed DBIx::SB::Record::Cachable to expire cached object when a "deeper" method call changes their values
0.86 June 7 2003
- Doing conditional connections was failing on postgres, because the handle was defined, but not connected
0.85 June 7 2003
- Stan's destroy fix was actually badly breaking RT
- It's now an optional parameter.
- Provide support for blowing away nested transactions
that aren't yet committed
- Fixed how values of returned hashes are downcased.
- Should be a minor perf improvement
- Moved Stan's destryo fix to the right file
- Better oracle support for unique ids on indexes
- DESTROY fix
- make function naming for autoloaded functions a bit more flexible
- Query builder changes to improve some join performance
- Fixes to a tight loop for cache expiry
- Patches for Oracle BLOB support from Brook Schofield
- Rebuilt Postgres query generator
- Select Distinct altered to support oracle
- Count method enhanced to ignore "LIMIT"s
- LIMIT behaviour changed to be handle specific
- ReadableAttributes and WritableAttributes added as methods to Record.pm
- SB->Count should return no results unless the search is limited
- Eliminate a warning on empty searches
- No longer attempt to cache (and fail) objects that haven't been
database-loaded
Extra checking for cache misses in DBIx::SearchBuilder::Record::Cachable
The start of support for checking database version, so that we can do
version-specific SQL
utf-8 safe searching
arbitrarily complex grouping clauses.
complex query generation.
Adding support for mysqlPP
Fixed a caching bug that caused multiple copies of an object in memory to not
be kept in sync
Fixed bug in setting a column to the value of an SQL statement.
Better support for Postgres 7.2 and transactions.
Support for Class::ReturnValue to channel errors up when expected
Dependency on Class::ReturnValue
Minor cleanups and refactorings to allow percolation of errors on create
The automatic truncation in gensolpkg doesn't work for packages which
have the same package name for the first 5-6 chars.
e.g. amanda-server and amanda-client would be named amanda and amanda.
Now, we add a SVR4_PKGNAME and use amacl for amanda-client and amase for
amanda-server.
All svr4 packages also have a vendor tag, so we have to reserve some chars
for this tag, which is normaly 3 or 4 chars. Thats why we can only use 6
or 5 chars for SVR4_PKGNAME. I used 5 for all the packages, to give the
vendor tag enough room.
All p5-* packages and a few other packages have now a SVR4_PKGNAME.
- SearchBuilder.pm - refactored to allow LEFT joins.
- SearchBuilder::Record::Cachable now constructs cache keys in a way
that doesn't lose when records in different tables have the same keys.
- Added DBIx::SearchBuilder::Record::Cachable from <mhat@netlag.com>
- Changed SearchBuilder->Count to do the right thing if no
query has been performed
