- modify inspection system:
1. raise a new exception for any case where the inspection
context can't be returned. this supersedes the "not mapped"
errors.
2. don't configure mappers on a mapper inspection. this allows
the inspectors to be used during mapper config time. instead,
the mapper configures on "with_polymorphic_selectable" now,
which is needed for all queries
- add a bunch of new "is_XYZ" attributes to inspectors
- finish making the name change of "compile" -> "configure", for some reason
this was only done partially
AbstractConcreteBase helper where the "type"
attribute from the superclass would not
be overridden on the subclass to produce the
"reserved for base" error message, instead placing
a do-nothing attribute there. This was inconsistent
vs. using ConcreteBase as well as all the behavior
of classical concrete mappings, where the "type"
column from the polymorphic base would be explicitly
disabled on subclasses, unless overridden
explicitly.
inheriting mappers to be GC'ed when the
class itself is deferenced. The mapper
must not have its own table (i.e.
single table inh only) without polymorphic
attributes in place.
This allows for the use case of
creating a temporary subclass of a declarative
mapped class, with no table or mapping
directives of its own, to be garbage collected
when dereferenced by a unit test.
[ticket:2526]
in declarative_base() is now a
WeakValueDictionary. So subclasses of
"Base" that are dereferenced will be
garbage collected, *if they are not
referred to by any other mappers/superclass
mappers*. [ticket:2526]
which prevented referring to a CTE in a union
of itself without it being aliased.
CTEs now render uniquely
on name, rendering the outermost CTE of a given
name only - all other references are rendered
just as the name. This even includes other
CTE/SELECTs that refer to different versions
of the same CTE object, such as a SELECT
or a UNION ALL of that SELECT. We are
somewhat loosening the usual link between object
identity and lexical identity in this case.
A true name conflict between two unrelated
CTEs now raises an error.
that an FK dependency between two tables is
not significant during flush if the tables
are related via joined inheritance and the FK
dependency is not part of the inherit_condition,
saves the user a use_alter directive.
[ticket:2527]
- rework bindtemplate system of "numbered" params by applying
the numbers last, as we now need to generate these out of order
in some cases
- add positional assertion to assert_compile
- add new cte_positional collection to track bindparams generated
within cte visits; splice this onto the beginning of self.positiontup
at cte render time, [ticket:2521]
and the entire related system of alternate
class implementation is now moved out
to sqlalchemy.ext.instrumentation. This is
a seldom used system that adds significant
complexity and overhead to the mechanics of
class instrumentation. The new architecture
allows it to remain unused until
InstrumentationManager is actually imported,
at which point it is bootstrapped into
the core.
- break out key mechanics of loading objects
into new "orm.loading" module, removing implementation
details from both mapper.py and query.py. is analogous
to persistence.py
- some other cleanup and old cruft removal
Mike Bayer