1.4.33

Released: March 31, 2022

orm

• [orm] [usecase] Added _orm.with_polymorphic.adapt_on_names to the _orm.with_polymorphic() function, which allows a polymorphic load (typically with concrete mapping) to be stated against an alternative selectable that will adapt to the original mapped selectable on column names alone.

  References: #7805

• [orm] [usecase] Added new attributes UpdateBase.returning_column_descriptions and UpdateBase.entity_description to allow for inspection of ORM attributes and entities that are installed as part of an Insert, Update, or Delete construct. The Select.column_descriptions accessor is also now implemented for Core-only selectables.

  References: #7861

• [orm] [performance] [bug] Improvements in memory usage by the ORM, removing a significant set of intermediary expression objects that are typically stored when a copy of an expression object is created. These clones have been greatly reduced, reducing the number of total expression objects stored in memory by ORM mappings by about 30%.

  References: #7823

• [orm] [bug] [regression] Fixed regression in "dynamic" loader strategy where the _orm.Query.filter_by() method would not be given an appropriate entity to filter from, in the case where a "secondary" table were present in the relationship being queried and the mapping were against something complex such as a "with polymorphic".

  References: #7868

• [orm] [bug] Fixed bug where _orm.composite() attributes would not work in conjunction with the _orm.selectin_polymorphic() loader strategy for joined table inheritance.

  References: #7801

• [orm] [bug] Fixed issue where the _orm.selectin_polymorphic() loader option would not work with joined inheritance mappers that don't have a fixed "polymorphic_on" column. Additionally added test support for a wider variety of usage patterns with this construct.

  References: #7799

• [orm] [bug] Fixed bug in _orm.with_loader_criteria() function where loader criteria would not be applied to a joined eager load that were invoked within the scope of a refresh operation for the parent object.

  References: #7862

• [orm] [bug] Fixed issue where the _orm.Mapper would reduce a user-defined _orm.Mapper.primary_key argument too aggressively, in the case of mapping to a UNION where for some of the SELECT entries, two columns are essentially equivalent, but in another, they are not, such as in a recursive CTE. The logic here has been changed to accept a given user-defined PK as given, where columns will be related to the mapped selectable but no longer "reduced" as this heuristic can't accommodate for all situations.

  References: #7842

engine

• [engine] [usecase] Added new parameter Engine.dispose.close, defaulting to True. When False, the engine disposal does not touch the connections in the old pool at all, simply dropping the pool and replacing it. This use case is so that when the original pool is transferred from a parent process, the parent process may continue to use those connections.

  References: #7815, #7877

• [engine] [bug] Further clarified connection-level logging to indicate the BEGIN, ROLLBACK and COMMIT log messages do not actually indicate a real transaction when the AUTOCOMMIT isolation level is in use; messaging has been extended to include the BEGIN message itself, and the messaging has also been fixed to accommodate when the Engine level create_engine.isolation_level parameter was used directly.

  References: #7853

sql

• [sql] [usecase] Added new parameter FunctionElement.table_valued.joins_implicitly, for the FunctionElement.table_valued() construct. This parameter indicates that the given table-valued function implicitly joins to the table it refers towards, essentially disabling the "from linting" feature, i.e. the "cartesian product" warning, from taking effect due to the presence of this parameter. May be used for functions such as func.json_each().

  References: #7845

• [sql] [bug] The bindparam.literal_execute parameter now takes part of the cache generation of a bindparam(), since it changes the sql string generated by the compiler. Previously the correct bind values were used, but the literal_execute would be ignored on subsequent executions of the same query.

  References: #7876

• [sql] [bug] [regression] Fixed regression caused by #7760 where the new capabilities of TextualSelect were not fully implemented within the compiler properly, leading to issues with composed INSERT constructs such as "INSERT FROM SELECT" and "INSERT...ON CONFLICT" when combined with CTE and textual statements.

  References: #7798

schema

• [schema] [usecase] Added support so that the Table.to_metadata.referred_schema_fn callable passed to Table.to_metadata() may return the value BLANK_SCHEMA to indicate that the referenced foreign key should be reset to None. The RETAIN_SCHEMA symbol may also be returned from this function to indicate "no change", which will behave the same as None currently does which also indicates no change.

  References: #7860

sqlite

• [sqlite] [bug] [reflection] Fixed bug where the name of CHECK constraints under SQLite would not be reflected if the name were created using quotes, as is the case when the name uses mixed case or special characters.

  References: #5463

mssql

• [mssql] [bug] [regression] Fixed regression caused by #7160 where FK reflection in conjunction with a low compatibility level setting (compatibility level 80: SQL Server 2000) causes an "Ambiguous column name" error. Patch courtesy @Lin-Your.

  References: #7812

misc

• [bug] [ext] Improved the error message that's raised for the case where the association_proxy() construct attempts to access a target attribute at the class level, and this access fails. The particular use case here is when proxying to a hybrid attribute that does not include a working class-level implementation.

  References: #7827