affected multiple connections, each
connection would individually invoke a new
disposal of the pool, even though only
one disposal is needed. [ticket:2522]
a disconnect detect + dispose that occurs
when the QueuePool has threads waiting
for connections would leave those
threads waiting for the duration of
the timeout on the old pool. The fix
now notifies those waiters with a special
exception case and has them move onto
the new pool. This fix may or may
not be ported to 0.7. [ticket:2522]
last_updated_params(), last_inserted_params(),
postfetch_cols(), prefetch_cols() all
assert that the given statement is a compiled
construct, and is an insert() or update()
statement as is appropriate, else
raise InvalidRequestError. [ticket:2498]
- ResultProxy.last_inserted_ids is removed,
replaced by inserted_primary_key.
string positional parameters passed to
engine/connection execute() would fail to be
interpreted correctly, due to __iter__
being present on Py3K string.
[ticket:2503]. Also in 0.7.8.
reflection" system has been moved into the
declarative extension itself, using the
new DeferredReflection class. This
class is now tested with both single
and joined table inheritance use cases.
[ticket:2485]
- [bug] The autoload_replace flag on Table,
when False, will cause any reflected foreign key
constraints which refer to already-declared
columns to be skipped, assuming that the
in-Python declared column will take over
the task of specifying in-Python ForeignKey
or ForeignKeyConstraint declarations.
deprecated; use Inspector.get_pk_constraint().
Courtesy Diana Clarke. [ticket:2422]
- restored default get_primary_keys()/get_pk_constraint() wrapper
to help maintain compatibility with third party dialects
created against 0.6 or 0.7
accepts the Inspector object as the first
argument, preceding "table". Code which
uses the 0.7 version of this very new
event will need modification to add the
"inspector" object as the first argument.
[ticket:2418]
option for connections. If no parameters
are present, will pass the statement
as cursor.execute(statement), thereby invoking
the DBAPIs behavior when no parameter collection
is present; for psycopg2 and mysql-python, this
means not interpreting % signs in the string.
This only occurs with this option, and not
just if the param list is blank, as otherwise
this would produce inconsistent behavior
of SQL expressions that normally escape percent
signs (and while compiling, can't know ahead of
time if parameters will be present in
some cases). [ticket:2407]
to create_engine, allows control over
"connection return" behavior. Also added
new arguments 'rollback', 'commit', None
to pool.reset_on_return to allow more control
over connection return activity. [ticket:2378]
when set to False on Table, the Table can be autoloaded
without existing columns being replaced. Allows
more flexible chains of Table construction/reflection
to be constructed, including that it helps with
combining Declarative with table reflection.
See the new example on the wiki. [ticket:2356]
- [bug] Improved the API for add_column() such that
if the same column is added to its own table,
an error is not raised and the constraints
don't get doubled up. Also helps with some
reflection/declarative patterns. [ticket:2356]
object named "<a>_<b>" which matched a column
labeled as "<tablename>_<colname>" could match
inappropriately when targeting in a result
set row. [ticket:2377]
- requires that we change the tuple format in RowProxy.
Makes an improvement to the cases tested against
an unpickled RowProxy as well though doesn't solve the
problem there entirely.
or invalid string is passed to any of the
date/time processors used by SQLite, including
C and Python versions. [ticket:2382]
- changed the import model of processors.py so that we can
get at the pure python versions and C versions simultaneously
in tests.
list of index names and the names of columns
within those indexes. [ticket:2269]
- rewrite unicode reflection test to be of more general use on
broken backends
would throw an error on an invalidated
connection if the transaction were a
two-phase or savepoint transaction.
For plain transactions, rollback() is a no-op
if the connection is invalidated, so while
it wasn't 100% clear if it should be a no-op,
at least now the interface is consistent.
[ticket:2317]
now allows for the reflection process to take
effect for a Table object that's already been
defined; when autoload=True and extend_existing=True
are both set, the full set of columns will be
reflected from the Table which will then
*overwrite* those columns already present,
rather than no activity occurring. Columns that
are present directly in the autoload run
will be used as always, however.
[ticket:1410]
option to Inspector.get_table_names
wasn't implementing the sort properly, replaced
with the existing sort algorithm
- clean up metadata usage in reflection tests
- MetaData() accepts "schema" and "quote_schema"
arguments, which will be applied to the same-named
arguments of a Table
or Sequence which leaves these at their default
of ``None``.
- Sequence accepts "quote_schema" argument
- tometadata() for Table will use the "schema"
of the incoming MetaData for the new Table
if the schema argument is explicitly "None"
- Added CreateSchema and DropSchema DDL
constructs - these accept just the string
name of a schema and a "quote" flag.
- When using default "schema" with MetaData,
ForeignKey will also assume the "default" schema
when locating remote table. This allows the "schema"
argument on MetaData to be applied to any
set of Table objects that otherwise don't have
a "schema".
- a "has_schema" method has been implemented
on dialect, but only works on Postgresql so far.
Courtesy Manlio Perillo, [ticket:1679]
"return unicode detection" step within connect,
allows databases that crash on NVARCHAR to
continue initializing, assuming no NVARCHAR
type implemented. [ticket:2299]
long lists of bound parameter sets will be
compressed with an informative indicator
of the compression taking place. Exception
messages use the same improved formatting.
[ticket:2243]
Mike Bayer