are now exercised within the normal unit test suite in both Python
2 and Python 3.
- remove the old testdocs.py runner and replace with test/base/test_tutorials.py
- use pytest's unicode fixer so that we can test for unicode strings
in both py2k/3k
- use py3k format overall for prints, exceptions
- add other fixers to guarantee deterministic results
- add skips and ellipses to outputs that aren't worth matching
the :class:`.Select` construct would have its ``._raw_columns``
collection mutated in-place when compiling the :class:`.Insert`
construct, when the target :class:`.Table` has Python-side defaults.
The :class:`.Select` construct would compile standalone with the
erroneous column present subsequent to compilation of the
:class:`.Insert`, and the the :class:`.Insert` statement itself would
fail on a second compile attempt due to duplicate bound parameters.
fixes#3603
of the :class:`.mysql.DATETIME`, :class:`.mysql.TIMESTAMP` and
:class:`.mysql.TIME` types would be incorrectly placed into the
``timestamp`` attribute, which is unused by MySQL, instead of the
``fsp`` attribute.
fixes#3602
primary key before emitting an INSERT, and merges distinct objects with
duplicate primary keys together as they are encountered, which is
essentially semi-deterministic at best. This behavior
matches what happens already with persistent objects.
fixes#3601
such as a CHECK constraint would render an erroneous comma in the
definition; this scenario can occur such as with a Postgresql
INHERITS table that has no columns of its own.
fixes#3598
returns an invalid date or time format from a DATE or TIME
column, rather than failing with a NoneType error. Pull request
courtesy Ed Avis.
pullreq github:206
directly or within lazy loads, didn't consider the mapper's "get clause"
as part of the cache key, causing bound parameter mismatches if the
clause got re-generated. This clause is cached by mappers
on the fly but in highly concurrent scenarios may be generated more
than once when first accessed.
fixes#3597
statement. This feature is available by passing the
:paramref:`~.sqlalchemy.sql.expression.update.preserve_parameter_order`
flag either to the core :class:`.Update` construct or alternatively
adding it to the :paramref:`.Query.update.update_args` dictionary at
the ORM-level, also passing the parameters themselves as a list of 2-tuples.
Thanks to Gorka Eguileor for implementation and tests.
adapted from pullreq github:200
versions 0.8.0 and 0.8.1, due 🎫`2714`. The case where
joined eager loading needs to join out over a subclass-bound
relationship when "with_polymorphic" were also used would fail
to join from the correct entity.
fixes#3593
Fixed bug where in Py2K a unicode literal would not be accepted as the
string name of a class or other argument within declarative using
:func:`.backref` on :func:`.relationship`.
amends commit e6f67f4805
For example, this query:
SELECT s1.users.name FROM s1.users FOR UPDATE OF s1.users
should actually be:
SELECT s1.users.name FROM s1.users FOR UPDATE OF users
fixes#3573
limit/offset criteria that forces a subquery b. the relationship
uses "secondary" c. the primaryjoin of the relationship refers to
a column that is either not part of the primary key, or is a PK
col in a joined-inheritance subclass table that is under a different
attribute name than the parent table's primary key column d. the
query defers the columns that are present in the primaryjoin, typically
via not being included in load_only(); the necessary column(s) would
not be present in the subquery and produce invalid SQL.
fixes#3592
scope of a :meth:`.Session.flush` operation that's raising an
exception, as has been observed in some MySQL SAVEPOINT cases, prevents
the original database exception from being observed when it was
emitted during flush, but only on Py2K because Py2K does not support
exception chaining; on Py3K the originating exception is chained. As
a workaround, a warning is emitted in this specific case showing at
least the string message of the original database error before we
proceed to raise the rollback-originating exception.
fixes#2696
to return ``datetime.timedelta`` in the same way as that of
:obj:`.types.Interval.python_type`, rather than raising
``NotImplementedError``.
fixes#3571
(cherry picked from commit 29d6f6e19b)
relationship flag; this flag *does* have an effect when the baked
lazy loader plugin has been invoked. clarify the intent of this
flag as an "opt out" but only has an effect when the baked system
is loaded anyway. fixes#3572
for UPDATE statements in the ORM (e.g. :ref:`feature_updatemany`)
would break on Postgresql and other RETURNING backends
when using server-side version generation
schemes, as the server side value is retrieved via RETURNING which
is not supported with executemany.
fixes#3556
:class:`.AssociationProxy` constructor, to suit the
:attr:`.AssociationProxy.info` accessor that was added in
🎫`2971`. This is possible because :class:`.AssociationProxy`
is constructed explicitly, unlike a hybrid which is constructed
implicitly via the decorator syntax.
fixes#3551
Mike Bayer