Modified the :class:`.DeclarativeMeta` metaclass to pass ``cls.__dict__``
into the declarative scanning process to look for attributes, rather than
the separate dictionary passed to the type's ``__init__()`` method. This
allows user-defined base classes that add attributes within an
``__init_subclass__()`` to work as expected, as ``__init_subclass__()`` can
only affect the ``cls.__dict__`` itself and not the other dictionary. This
is technically a regression from 1.3 where ``__dict__`` was being used.
Additionally makes the reference between ClassManager and
the declarative configuration object a weak reference, so that it
can be discarded after mappers are set up.
Fixes: #7900
Change-Id: I3c2fd4e227cc1891aa4bb3d7d5b43d5686f9f27c
in this patch the asyncio/events.py module, which
existed only to raise errors when trying to attach event
listeners, is removed, as we were already coding an asyncio-specific
workaround in upstream Pool / Session to raise this error,
just moved the error out to the target and did the same thing
for Engine.
We also add an async_sessionmaker class. The initial rationale
here is because sessionmaker() is hardcoded to Session subclasses,
and there's not a way to get the use case of
sessionmaker(class_=AsyncSession) to type correctly without changing
the sessionmaker() symbol itself to be a function and not a class,
which gets too complicated for what this is. Additionally,
_SessionClassMethods has only three methods on it, one of which
is not usable with asyncio (close_all()), the others
not generally used from the session class.
Change-Id: I064a5fa5d91cc8d5bbe9597437536e37b4e801fe
Also adds some fixes to annotation-based mapping
that have come up, as well as starts to add more
pep-484 test cases
Change-Id: Ia722bbbc7967a11b23b66c8084eb61df9d233fee
our decorator thing generates code in any case,
so point it at the file itself to generate real code
for the blocks rather than doing things dynamically.
this will allow typing tools to have no problem
whatsoever and we also reduce import time overhead.
file size will be a lot bigger though, shrugs.
syntax / dupe method / etc. checking will be accomplished
by our existing linting / typing / formatting tools.
As we are also using "from __future__ import annotations",
we also no longer have to apply quotes to generated
annotations.
Change-Id: I20962cb65bda63ff0fb67357ab346e9b1ef4f108
Have each result subclass be generic to the
kind of row/object it returns. rework things so that
a significant number of "type ignores" can go away
and also allow asyncio to more cleanly proxy
the result objects.
Change-Id: Ia3ddb2cb52f5856839bd8c9c46f0289ab4b10405
it referred towards _columntoproperty refering to
lists of MapperProperty. this comment goes all the
way to pre 0.1 being released. it's likely been
wrong for nearly all that time.
Change-Id: I71234ae58a6253249d92224356e38372e4aff148
I *think* we don't need to tell .coverage to look in lib,
it seems to be covering from whereever successfuly. not
sure, but want to run on jenkins to see how it comes out.
Change-Id: Ib74f427a5b09fcbeba0188dcf02dab27c329a8cd
For third party dialects, repaired a missing requirement for the
``SimpleUpdateDeleteTest`` suite test which was not checking for a working
"rowcount" function on the target dialect.
Fixes: #7919
Change-Id: I2bc68132131eb36c43b8dabec0fac86272e26df5
Added modified ISO-8601 rendering (i.e. ISO-8601 with the T converted to a
space) when using ``literal_binds`` with the SQL compilers provided by the
PostgreSQL, MySQL, MariaDB, MSSQL, Oracle dialects. For Oracle, the ISO
format is wrapped inside of an appropriate TO_DATE() function call.
Previously this rendering was not implemented for dialect-specific
compilation.
Fixes: #5052
Change-Id: I7af15a51fedf5c5a8e76e645f7c3be997ece35f0
As developers are now buying lots of Apple M1
machines, and AFAWK greenlet is still not able
to provide a pre-built wheel, we are going to get a lot
of devs trying to use asyncio on their Apple M1s, in
greater proportions compared to devs running containers
etc. on other various less popular CPU architectures.
Add a new FAQ section for installation, add new
red dragon to the very top of asyncio docs,
add new verbiage, all of which includes
"Apple M1" in bold text, to minimize the chance of anyone
missing this.
Fixes: #7922Fixes: #7714
Change-Id: I106923a2860a4efd77d1b999197be102afc1f73d
Full "RETURNING" support is implemented for the cx_Oracle dialect, meaning
multiple RETURNING rows are now recived for DML statements that produce
more than one row for RETURNING.
cx_Oracle 7 is now the minimum version for cx_Oracle.
Getting Oracle to do multirow returning took about 5 minutes. however,
getting Oracle's RETURNING system to integrate with ORM-enabled
insert, update, delete, is a big deal because that architecture wasn't
really working very robustly, including some recent changes in 1.4
for FromStatement were done in a hurry, so this patch also cleans up
the FromStatement situation and begins to establish it more concretely
as the base for all ReturnsRows / TextClause ORM scenarios.
Fixes: #6245
Change-Id: I2b4e6007affa51ce311d2d5baa3917f356ab961f
Fixed regression caused by 🎫`7823` which impacted the caching
system, such that bound parameters that had been "cloned" within ORM
operations, such as polymorphic loading, would in some cases not acquire
their correct execution-time value leading to incorrect bind values being
rendered.
Fixes: #7903
Change-Id: I61c802749b859bebeb127d24e66d6e77d13ce57a
the pep484 task becomes more intense as there is mounting
pressure to come up with a consistency in how data moves
from end-user to instance variable.
current thinking is coming into:
1. there are _typing._XYZArgument objects that represent "what the
user sent"
2. there's the roles, which represent a kind of "filter" for different
kinds of objects. These are mostly important as the argument
we pass to coerce().
3. there's the thing that coerce() returns, which should be what the
construct uses as its internal representation of the thing.
This is _typing._XYZElement.
but there's some controversy over whether or
not we should pass actual ClauseElements around by their role
or not. I think we shouldn't at the moment, but this makes the
"role-ness" of something a little less portable. Like, we have
to set DMLTableRole for TableClause, Join, and Alias, but then
also we have to repeat those three types in order to set up
_DMLTableElement.
Other change introduced here, there was a deannotate=True
for the left/right of a sql.join(). All tests pass without that.
I'd rather not have that there as if we have a join(A, B) where
A, B are mapped classes, we want them inside of the _annotations.
The rationale seems to be performance, but this performance can
be illustrated to be on the compile side which we hope is cached
in the normal case.
CTEs now accommodate for text selects including recursive.
Get typing to accommodate "util.preloaded" cleanly; add "preloaded"
as a real module. This seemed like we would have needed
pep562 `__getattr__()` but we don't, just set names in
globals() as we import them.
References: #6810
Change-Id: I34d17f617de2fe2c086fc556bd55748dc782faf0
SQLite datetime, date, and time datatypes now use Python standard lib
``fromisoformat()`` methods in order to parse incoming datetime, date, and
time string values. This improves performance vs. the previous regular
expression-based approach, and also automatically accommodates for datetime
and time formats that contain either a six-digit "microseconds" format or a
three-digit "milliseconds" format.
Fixes: #7029
Change-Id: I67aab4fe5ee3055e5996050cf4564981413cc221
Fixed bug in newly implemented
:paramref:`.FunctionElement.table_valued.joins_implicitly` feature where
the parameter would not automatically propagate from the original
:class:`.TableValuedAlias` object to the secondary object produced when
calling upon :meth:`.TableValuedAlias.render_derived` or
:meth:`.TableValuedAlias.alias`.
Additionally repaired these issues in :class:`.TableValuedAlias`:
* repaired a potential memory issue which could occur when
repeatedly calling :meth:`.TableValuedAlias.render_derived` against
successive copies of the same object (for .alias(), we currently
have to still continue chaining from the previous element. not sure
if this can be improved but this is standard behavior for .alias()
elsewhere)
* repaired issue where the individual element types would be lost when
calling upon :meth:`.TableValuedAlias.render_derived` or
:meth:`.TableValuedAlias.alias`.
Fixes: #7890
Change-Id: Ie5120c7ff1e5c1bba5aaf77c782a51c637860208
this second example is not part of the doctest steps,
clarify that it's not part of code examples to be present
in execution steps. Add an extra registry +
declarative base on top
so that even if someone does run it, the Base will have
been reset and the examples will continue to work
(noting that column order in statements may change, but
probably nothing else).
Fixes: #7891
Change-Id: Icb1ba310230841e502185d9d0cadd3c18d467292
Scaled back a fix made for 🎫`6581` where "executemany values" mode
for psycopg2 were disabled for all "ON CONFLICT" styles of INSERT, to
not apply to the "ON CONFLICT DO NOTHING" clause, which does not include
any parameters and is safe for "executemany values" mode. "ON CONFLICT
DO UPDATE" is still blocked from "executemany values" as there may
be additional parameters in the DO UPDATE clause that cannot be batched
(which is the original issue fixed by 🎫`6581`).
Fixes: #7880
Change-Id: Id3e23a0c6699333409a50148fa8923cb8e564bdc
Fixed regression caused by 🎫`7861` where invoking an
:class:`.Insert` construct which contained ORM entities via
:meth:`_orm.Session.execute` would fail.
Fixes: #7878
Change-Id: Icc4d8028249cc417f504fdd3e31e206b5bbc89f8
Added new parameter :paramref:`.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.
Fixes: #7877
Change-Id: I88b0808442381ba5e50674787cdb64f0e77d8b54
The ``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.
Fixes: #7876
Change-Id: I6bf887f1a2fe31f9d0ab68f5b4ff315004d006b2
hitting DML which is causing us to open up the
ColumnCollection structure a bit, as we do put anonymous
column expressions with None here. However, we still want
Table /TableClause to have named column collections that
don't return None, so parametrize the "key" in this
collection also.
* rename some "immutable" elements to "readonly". we change
the contents of immutablecolumncollection underneath, so it's
not "immutable"
Change-Id: I2593995a4e5c6eae874bed5bf76117198be8ae97
if these files are stale, imports fail and the build
silently excludes docstrings. also , typing / docstrings
are generally in the _py version of things, so just
don't include cyexts in doc builds.
Change-Id: Ibec927c7646d4b870a84a297d1b199ccecf3afef
Mike Bayer