++ed by:

104 PAUSE user(s)
104 non-PAUSE user(s).

Peter Rabbitson
and 1 contributors




Changes for version 0.08250

  • New Features / Changes
    • Rewrite from scratch the result constructor codepath - many bugfixes and performance improvements (the current codebase is now capable of outperforming both DBIx::DataModel and Rose::DB::Object on some workloads). Some notable benefits:
      • Multiple has_many prefetch
      • Partial prefetch - you now can select only columns you are interested in, while preserving the collapse functionality (collapse is now exposed as a first-class API attribute)
      • Prefetch of resultsets with arbitrary order (RT#54949, RT#74024, RT#74584)
      • Prefetch no longer inserts right-side table order_by clauses (massively helps the deficient MySQL optimizer)
      • Prefetch with limit on right-side ordered resultsets now works correctly (via aggregated grouping)
      • No longer order the insides of a complex prefetch subquery, unless required to satisfy a limit
      • Stop erroneously considering order_by criteria from a join under distinct => 1 (the distinct should apply to the main source only)
    • Massively optimize codepath around ->cursor(), over 10x speedup on some iterating workloads.
    • Support standalone \[ $sql, $value ] in literal SQL with bind specifications: \[ '? + ?', 42, 69 ] is now equivalent to \[ '? + ?', [ {} => 42 ], [ {} => 69 ] ]
    • Changing the result_class of a ResultSet in progress is now explicitly forbidden. The behavior was undefined before, and would result in wildly differing outcomes depending on $rs attributes.
    • Deprecate returning of prefetched 'filter' rels as part of get_columns() and get_inflated_columns() data
    • Invoking get_inflated_columns() no longer fires get_columns() but instead retrieves data from individual non-inflatable columns via get_column()
    • Emit a warning on incorrect use of nullable columns within a primary key
    • Limited checks are performed on whether columns without declared is_nullable => 1 metadata do in fact sometimes fetch NULLs from the database (the check is currently very limited and is performed only on resultset collapse when the alternative is rather worse)
  • Fixes
    • Fix _dbi_attrs_for_bind() being called befor DBI has been loaded (regression in 0.08210)
    • Fix update/delete operations on resultsets *joining* the updated table failing on MySQL. Resolves oversights in the fixes for RT#81378 and RT#81897
    • Fix open cursors silently resetting when inherited across a fork or a thread
    • Properly support "MySQL-style" left-side group_by with prefetch
    • Fix $grouped_rs->get_column($col)->func($func) producing incorrect SQL (RT#81127)
    • Stop Sybase ASE storage from generating invalid SQL in subselects when a limit without offset is encountered
    • Even more robust behavior of GenericSubQuery limit dialect
    • Make sure deployment_statements() and cursor_class() are called on a resolved storage subclass
  • Misc
    • Fix tests failing due to unspecified resultset retrieval order (test suite now will pass with newest SQLite libs)