Switch branches/tags
Commits on May 10, 2011
  1. Release 0.08192

    frioux committed May 10, 2011
  2. turn 'use vars' into 'our'

    arcanez authored and frioux committed May 9, 2011
    move the $VERSION up to the top of DBIx/ to aid in version detection
    (things like Module::Metadata find the $VERSION string and base it on the last package statement it has seen, not knowing that } 'ends' the package)
Commits on May 9, 2011
Commits on May 7, 2011
Commits on May 3, 2011
  1. Release 0.08191

    frioux committed May 3, 2011
Commits on May 2, 2011
  1. Release 0.08190_01

    frioux committed May 2, 2011
  2. Throw away the error prone _strip_cond_qualifiers

    ribasushi authored and frioux committed May 2, 2011
    Review of 61f031b made it clear that in order to make this work properly
    pretty much the entire sqla logic will need to be replicated. So instead
    the hacky approach was taken to abuse a botched version of SQLA which
    is guaranteed to parse its own command tree correctly :)
  3. Revert the smart-select-ordering introduced in 36fd7f0

    ribasushi authored and frioux committed Feb 15, 2011
    While on its surface an awesome idea, it proced to be problematic:
    users of ->cursor and ->as_query expected that the result will have the same
    column-order as what they put in the arguments. Also users abuse +select as
    there's no tomorrow, which yielded interesting bugs. The current (and I hope
    final) behavior is:
    At the end of a chain, the resulting SQL will contain:
    - a merger of all columns and then all +columns attributes in the
      declared order. If columns is declared as a multiselect hashref
      i.e. { as1 => sel1, as2 => sel2 }, then the selectors for this
      segment are ordered alphabetically. De-duplication is also always
      performed on columns arguments - only the first set of identical
      select/as pairs is preserved
    - a merger of all select/as pairs, strict order, no deduplication
    - a merger of +select/+as pairs, same rules as above
    As a result of this prefetch on un-balanced resultsets now properly throws
    an exception
  4. Channel all DBIHacks methods to the replicated writer

    ribasushi authored and frioux committed Apr 30, 2011
    Note this is a workaround to fix an actual reported bug. The whole system
    reeks of misdesign, and needs to be revisited at some point.
  5. Revert incorrect assumption about non-functional cached pagers 6524522

    ribasushi committed Apr 30, 2011
    While on the surface it seems logical that a cached resultset is inherently
    non-pageable, one has to remember that search chaining with different
    arguments (i.e. different offset) fires a new query, not subject to the
    existing resultset cache.
  6. Use clearer varnames

    ribasushi committed Oct 4, 2010
  7. Cleanup Oracle's 00a2818 / add support for update/delete with blobs i…

    rkitover authored and ribasushi committed May 1, 2011
    …n WHERE
    Make a separate path for as_query, so that the sql does not pass thorugh storage
    multiple times. Also do not mangle the sql unless blob binds are detected.
Commits on Apr 29, 2011
  1. prefetch docs: prefetch works on all rel types

    felliott authored and ribasushi committed Apr 21, 2011
    Update the prefetch docs to state that prefetch works with all
    relationship types, where previously it had said that it works with
    only belongs_to and has_one.
    Also update the subsequent example. The old example was confusing
    because it specified two has_many relationships on the same level,
    something that DBIC emits a warning about.  That example is replaced
    with a CD-centric example that demonstrates how many different
    relationship types can be used together in a prefetch, but omits the
    multiple has_manys.
    Add a warning, example, and explanation of the troubles with using
    multiple has_many relations in a prefetch.
    The old text incorrectly stated that prefetch will override join and
    select attributes. Describe current behavior with examples.
  2. savepoints for SQLite

    rkitover authored and ribasushi committed Feb 25, 2011
  3. Another overhaul of transaction/savepoint handling

    ribasushi committed Apr 26, 2011
    - Move most of the transaction logic back to ::Storage, only leave
      DBI-specific overrides in ::Storage::DBI
    - Fix bug where a nested rollback would destabilize the entire execution
    - Better checks for connectivity/ordering of txn/svp operations
    - Make DBIC::Storage::NESTED_ROLLBACK_EXCEPTION a proper exception subclass
    - Standardize the txn/svp names - renamed the actual workers
      _dbh_begin_work               to _exec_txn_begin
      _dbh_commit                   to _exec_txn_commit
      _dbh_rollback                 to _exec_txn_rollback
      _svp_[begin|release|rollback] to _exec_svp_[begin|release|rollback]
Commits on Apr 22, 2011
Commits on Apr 21, 2011
  1. Drop DBI requirement down to 1.57

    ribasushi committed Apr 21, 2011
    SQLite, MySQL, Pg, Oracle, ODBC and Sybase all tested to work
Commits on Apr 19, 2011
  1. Fix $rs->next when using HRI with right-join-side data injected into …

    Alexander Kuznetsov authored and ribasushi committed Mar 10, 2011
    …main obj
Commits on Apr 16, 2011
  1. Add retrieve_on_insert column_info flag, to autoretrieve RDBMS-side d…

    rbo authored and ribasushi committed Mar 10, 2011
    Also move the UniqueIdentifier code from insert() to _prefetch_autovalues()
    where it makes more logical sense
Commits on Apr 14, 2011
  1. fix some spelling mistakes

    arcanez committed Apr 14, 2011