enhancements where index reflection on Postgresql versions specific
to only the 8.1, 8.2 series again
broke, surrounding the ever problematic int2vector type. While
int2vector supports array operations as of 8.1, apparently it only
supports CAST to a varchar as of 8.3.
fix#3000
implementation allows an event handler to redefine the specific mechanics
by which an arbitrary dialect invokes execute() or executemany() on a
DBAPI cursor. The new events, at this point semi-public and experimental,
are in support of some upcoming transaction-related extensions.
after one or more :class:`.Connection` objects have been created
(such as by an orm :class:`.Session` or via explicit connect)
and the listener will pick up events from those connections.
Previously, performance concerns pushed the event transfer from
:class:`.Engine` to :class:`.Connection` at init-time only, but
we've inlined a bunch of conditional checks to make this possible
without any additional function calls. fixes#2978
within userland engine.dispose(); as some SQLA tests already failed when the replace step
was removed, due to those conns still being referenced, it's likely this will
create surprises for all those users that incorrectly use dispose()
and it's not really worth dealing with. This doesn't affect the change
we made for ref: #2985.
recycles the connection pool when a "disconnect" condition is detected;
instead of discarding the pool and explicitly closing out connections,
the pool is retained and a "generational" timestamp is updated to
reflect the current time, thereby causing all existing connections
to be recycled when they are next checked out. This greatly simplifies
the recycle process, removes the need for "waking up" connect attempts
waiting on the old pool and eliminates the race condition that many
immediately-discarded "pool" objects could be created during the
recycle operation. fixes#2985
:class:`.DateTime`. As Oracle has no "datetime" type per se,
it instead has only ``DATE``, it is appropriate here that the
``DATE`` type as present in the Oracle dialect be an instance of
:class:`.DateTime`. This issue doesn't change anything as far as
the behavior of the type, as data conversion is handled by the
DBAPI in any case, however the improved subclass layout will help
the use cases of inspecting types for cross-database compatibility.
Also removed uppercase ``DATETIME`` from the Oracle dialect as this
type isn't functional in that context. fixes#2987
removal of the ``setuptools.Feature`` extension from setuptools.
If this keyword isn't present, the setup will still succeed
with setuptools rather than falling back to distutils. C extension
building can be disabled now also by setting the
DISABLE_SQLALCHEMY_CEXT environment variable. This variable works
whether or not setuptools is even available. fixes#2986
- using platform.python_implementation() in setup.py to detect CPython.
I've tested this function on OSX and linux on Python 2.6 through 3.4,
including 3.1, 3.2, 3.3.
Unfortunately, on OSX + 3.2 only, it seems to segfault. I've tried
installing 3.2.5 from the python.org .dmg, building it from source,
and also blew away the whole 3.2 directory, something seems to be wrong
with the "platform" module on that platform only, and there's also no
issue on bugs.python.org; however, I'm going with
it anyway. If someone is using 3.2 on OSX they really should be upgrading.
1. fix the typo in the paragraph, fixes#2998
2. as zope-sqlalchemy only provides transaction integration and not session scoping,
dial back the language here as people are probably using scoped_session with pyramid anyway
3. as I'm going to again start recommending people don't cling to flask-sqlalchemy so hard,
take out the word "strongly" from the recommendation.
4. as flask is the only framework I can think of that actually has an explicit SQLAlchemy
layer that handles setting up scoped_session, take out the word "most", now it's "some web frameworks"
(by which it means "only flask...and flask-sqlalchemy is probably not worth using anyway")
emitted for the "_cursor_execute()" method of :class:`.Connection`;
this is the "quick" executor that is used for things like
when a sequence is executed ahead of an INSERT statement, as well as
for dialect startup checks like unicode returns, charset, etc.
the :meth:`.ConnectionEvents.before_cursor_execute` event was already
invoked here. The "executemany" flag is now always set to False
here, as this event always corresponds to a single execution.
Previously the flag could be True if we were acting on behalf of
an executemany INSERT statement.
against a non-selectable, such as a :func:`.literal_column`, and then
an attempt was made to use :meth:`.Query.join` such that the "left"
side would be determined as ``None`` and then fail. This condition
is now detected explicitly.
If autoloading of a table fails, don't register it in a metadata
instance. It seems that the original behaviour was accidentally
changed in f6198d9abf, restore it.
Closes issue #2988
mark a constraint name as already having had a naming convention applied.
This token will be used by Alembic migrations as of Alembic 0.6.4
in order to render constraints in migration scripts with names marked
as already having been subject to a naming convention.
re: #2991
where the name of a check constraint making use of the
`"%(constraint_name)s"` token would get doubled up for the
constraint generated by a boolean or enum type, and overall
duplicate events would cause the `"%(constraint_name)s"` token
to keep compounding itself.
fixes#2991
or :meth:`.MapperEvents.after_configured` events are applied to a
specific mapper or mapped class, as the events are only invoked
for the :class:`.Mapper` target at the general level.
Mike Bayer