_handle_dbapi_exception_noconnection() to only invoke in the case
of raw_connection() in the constructor of Connection. in all other
cases the Connection proceeds with _handle_dbapi_exception() including
revalidate.
to it, so that we say we will do the wrapping just once right here
in _execute_context() / _execute_default(). An adjustment is made
to _handle_dbapi_error() to not assume self.__connection in case
we are already in an invalidated state
further adjustment to
0639c199a5, 41e7253dee, #3266
take effect in all engine connection use cases, including
when user-custom connect routines are used via the
:paramref:`.create_engine.creator` parameter, as well as when
the :class:`.Connection` encounters a connection error on
revalidation.
fixes#3266
and parameters are not displayed if None, reducing confusion for
error messages that weren't related to a statement. The full
module and classname for the DBAPI-level exception is displayed,
making it clear that this is a wrapped DBAPI exception. The
statement and parameters themselves are bounded within a bracketed
sections to better isolate them from the error message and from
each other.
fixes#3172
:paramref:`.create_engine.execution_options` or
:meth:`.Engine.update_execution_options` are not passed to the
special :class:`.Connection` used to initialize the dialect
within the "first connect" event; dialects will usually
perform their own queries in this phase, and none of the
current available options should be applied here. In
particular, the "autocommit" option was causing an attempt to
autocommit within this initial connect which would fail with
an AttributeError due to the non-standard state of the
:class:`.Connection`.
fixes#3200
when you call :meth:`.Connection.connect`, would not share transaction
status with the parent. The architecture of branching has been tweaked
a bit so that the branched connection defers to the parent for
all transactional status and operations.
fixes#3190
when you call :meth:`.Connection.connect`, would not share invalidation
status with the parent. The architecture of branching has been tweaked
a bit so that the branched connection defers to the parent for
all invalidation status and operations.
fixes#3215
we need to be in transactions (tested elsewhere) and we need to
emit the correct FOR UPDATE strings (tested elsewhere). There's nothing
in SQLA to be tested as far as validating that for update causes exceptions
or not, and these tests frequently fail as they are timing sensitive.
as that of 🎫`3199`, when the ``named=True`` parameter
would be used. Some events would fail to register, and others
would not invoke the event arguments correctly, generally in the
case of when an event was "wrapped" for adaption in some other way.
The "named" mechanics have been rearranged to not interfere with
the argument signature expected by internal wrapper functions.
fixes#3197
reports a column that isn't found to be present in the table,
a warning is emitted and the column is skipped. This can occur
for some special system column situations as has been observed
with Oracle. fixes#3180
debug logging message would not emit if the logging were set up using
``logging.setLevel()``, rather than using the ``echo_pool`` flag.
Tests to assert this logging have been added. This is a
regression that was introduced in 0.9.0.
fixes#3168
for an INSERT or UPDATE are now sorted when they contribute towards
the "compiled cache" cache key. These keys were previously not
deterministically ordered, meaning the same statement could be
cached multiple times on equivalent keys, costing both in terms of
memory as well as performance.
fixes#3165
In test/engine/test_transaction/test_two_phase_recover(), a COMMIT
PREPARED is issued while in a transaction. This causes an error, and
a prepared transaction is left hanging around which causes
the subsequent test to hang. I've altered the test to execute the
offending query with autocommit=true, then when it gets to the COMMIT
PRPARED it can go ahead.
There's another complication for pg8000 because its tpc_recover() method
started a transaction if one wasn't already in progress. I've decided
that this is incorrect behaviour and so from pg8000-1.9.13 this method
never starts or stops a transaction.
non-standard DBAPI exceptions, such as the psycopg2
TransactionRollbackError. These exceptions will now be raised
using the closest available subclass in ``sqlalchemy.exc``, in the
case of TransactionRollbackError, ``sqlalchemy.exc.OperationalError``.
fixes#3075
events from firing for those statements which it uses internally
to detect if a table exists or not. This is achieved using an
execution option ``skip_user_error_events`` that disables the handle
error event for the scope of that execution. In this way, user code
that rewrites exceptions doesn't need to worry about the MySQL
dialect or other dialects that occasionally need to catch
SQLAlchemy specific exceptions.
in all cases unconditionally, the number of use cases that go beyond what
dbapi_error() is expecting has gone too far for an 0.9 release.
Additionally, the number of things we'd like to track is really a lot
more than the five arguments here, and ExecutionContext is really not
suitable as totally public API for this. So restore dbapi_error
to its old version, deprecate, and build out handle_error instead.
This is a lot more extensible and doesn't get in the way of anything
compatibility-wise.
have been enhanced such that the function handler is now capable
of raising or returning a new exception object, which will replace
the exception normally being thrown by SQLAlchemy.
fixes#3076
- Fixed bug which would occur if a DBAPI exception
occurs when the engine first connects and does its initial checks,
and the exception is not a disconnect exception, yet the cursor
raises an error when we try to close it. In this case the real
exception would be quashed as we tried to log the cursor close
exception via the connection pool and failed, as we were trying
to access the pool's logger in a way that is inappropriate
in this very specific scenario. fixes#3063
1. make sure pool._invalidate() sets the timestamp up before
invalidating the target connection. we can otherwise show how the
conn.invalidate() + pool._invalidate() can lead to an extra connection
being made.
2. to help with that, soften up the check on connection.invalidate()
when connection is already closed. a warning is fine here
3. add a mutex to test_max_overflow() when we connect, because the way
we're using mock depends on an iterator, that needs to be synchronized
Mike Bayer