but puts the changelog in a more data driven format, so that we
can begin linking version directives with them as well
as with migration documents, once those are also moved into sphinx.
- build a new Sphinx extension that allows dialect info
to be entered as directives which is then rendered consistently
throughout all dialect/dbapi sections
- break out the "empty_strings" requirement for oracle test
- begin consolidating docs for dialects to be more self contained
- add a separate section for "external" dialects
- not sure how we're going to go with this yet.
that works in terms of the "impl" type by default.
This is a behavioral change for those TypeDecorator
classes that specify a custom __init__ method; those
types will need to re-define __repr__() if they need
__repr__() to provide a faithful constructor representation.
[ticket:2594]
"tuple" rows that contain
types which aren't hashable, by setting the flag
"hashable=False" on the corresponding TypeEngine object
in use. Custom types that return unhashable types
(typically lists) can set this flag to False.
[ticket:2592]
- [bug] Applying a column expression to a select
statement using a label with or without other
modifying constructs will no longer "target" that
expression to the underlying Column; this affects
ORM operations that rely upon Column targeting
in order to retrieve results. That is, a query
like query(User.id, User.id.label('foo')) will now
track the value of each "User.id" expression separately
instead of munging them together. It is not expected
that any users will be impacted by this; however,
a usage that uses select() in conjunction with
query.from_statement() and attempts to load fully
composed ORM entities may not function as expected
if the select() named Column objects with arbitrary
.label() names, as these will no longer target to
the Column objects mapped by that entity.
[ticket:2591]
to help with generative building. Also slight adjustment
regarding how SS "correlates" columns; the new methodology
no longer applies meaning to the underlying
Table column being selected. This improves
some fairly esoteric situations, and the logic
that was there didn't seem to have any purpose.
- [feature] Some support for auto-rendering of a
relationship join condition based on the mapped
attribute, with usage of core SQL constructs.
E.g. select([SomeClass]).where(SomeClass.somerelationship)
would render SELECT from "someclass" and use the
primaryjoin of "somerelationship" as the WHERE
clause. This changes the previous meaning
of "SomeClass.somerelationship" when used in a
core SQL context; previously, it would "resolve"
to the parent selectable, which wasn't generally
useful. Related to [ticket:2245].
mapped classes allows that unmapped classes
can be specified for instance- and mapper-events.
The established events will be automatically
set up on subclasses of that class when the
propagate=True flag is passed, and the
events will be set up for that class itself
if and when it is ultimately mapped.
[ticket:2585]
- [bug] The instrumentation events class_instrument(),
class_uninstrument(), and attribute_instrument()
will now fire off only for descendant classes
of the class assigned to listen(). Previously,
an event listener would be assigned to listen
for all classes in all cases regardless of the
"target" argument passed. [ticket:2590]
String types. When present, renders as
COLLATE <collation>. This to support the
COLLATE keyword now supported by several
databases including MySQL, SQLite, and Postgresql.
[ticket:2276]
- [change] The Text() type renders the length
given to it, if a length was specified.
accepts "fst" argument which is the new
"fractional seconds" specifier for recent
MySQL versions. The datatype will interpret
a microseconds portion received from the driver,
however note that at this time most/all MySQL
DBAPIs do not support returning this value.
[ticket:2534]
- attempted to modernize the types tests in test_mysql a little, though has a long
way to go
API to better support highly specialized
systems such as the Akiban database, including
more hooks to allow an execution context to
access type processors.
state post-flush due to event listeners;
any states that are marked as "dirty" from an
attribute perspective, usually via column-attribute
set events within after_insert(), after_update(),
etc., will get the "history" flag reset
in all cases, instead of only those instances
that were part of the flush. This has the effect
that this "dirty" state doesn't carry over
after the flush and won't result in UPDATE
statements. A warning is emitted to this
effect; the set_committed_state()
method can be used to assign attributes on objects
without producing history events. [ticket:2582]
when unsupported methods are used inside the
"execute" portion of the flush. These are
the familiar methods add(), delete(), etc.
as well as collection and related-object
manipulations, as called within mapper-level
flush events
like after_insert(), after_update(), etc.
It's been prominently documented for a long
time that SQLAlchemy cannot guarantee
results when the Session is manipulated within
the execution of the flush plan,
however users are still doing it, so now
there's a warning. Maybe someday the Session
will be enhanced to support these operations
inside of the flush, but for now, results
can't be guaranteed.
a deleted object in the identity map with another
object of the same primary key would raise a
"conflicting state" error on rollback(),
if the replaced primary key were established either
via non-unitofwork-established INSERT statement
or by primary key switch of another instance.
[ticket:2583]
- rewrite --dropfirst to be more industrial strength, includes views
- fix order_by="foreign_key" to maintain the same ordering as
metadata.sorted_tables. Not ideal that this was the other way throughout
0.7 but this is still a little-used method, in contrast to metadata.sorted_tables.
moved to its own project on Bitbucket,
taking advantage of the new SQLAlchemy
dialect compliance suite. The dialect is
still in very rough shape and probably not
ready for general use yet, however
it does have *extremely* rudimental
functionality now.
would double up columns if the same constraint/table
existed in multiple schemas.
- force returns_rows to False for inserts where we know rows shouldnt be returned;
allows post_exec() to use the cursor without issue
passing an empty list for either partition_by
or order_by, as opposed to None, would fail
to generate correctly.
Courtesy Gunnlaugur Por Briem.
[ticket:2574]
subquery within an ORDER BY would fail to render correctly
if the stament also used LIMIT/OFFSET, due to mis-rendering
within the ROW_NUMBER() OVER clause. Fix courtesy
sayap [ticket:2538]
will now schema-qualify the name of the index
to be that of the parent table. Previously this
name was omitted which apparently creates the
index in the default schema, rather than that
of the table.
- add some failure cases
- [bug] Firebird now uses strict "ansi bind rules"
so that bound parameters don't render in the
columns clause of a statement - they render
literally instead.
- [bug] Support for passing datetime as date when
using the DateTime type with Firebird; other
dialects support this.
are now returned in the order in which the constraint
itself defines them, rather than how the table
orders them. Courtesy Gunnlaugur Por Briem.
[ticket:2531].
is replaced, such as via extend_existing,
the "auto increment" column used by insert()
constructs is reset. Previously it would
remain referring to the previous primary
key column. [ticket:2525]
convert non-string values to string, to accommodate
old SQLite versions that don't deliver
default info as a string. [ticket:2265]
- factor sqlite column reflection to be like we did for postgresql,
in a separate method.
"concat" and "match" operators to be the same as
that of "is", "like", and others; this helps with
parenthesization rendering when used in conjunction
with "IS". [ticket:2564]
that occurs within Session.commit(), such that the
extra state added by an after_flush() or
after_flush_postexec() hook is also flushed in a
subsequent flush, before the "commit" completes.
Subsequent calls to flush() will continue until
the after_flush hooks stop adding new state.
An "overflow" counter of 100 is also in place,
in the event of a broken after_flush() hook
adding new content each time. [ticket:2566]
and after_transaction_end
allows tracking of new SessionTransaction objects.
If the object is inspected, can be used to determine
when a session first becomes active and when
it deactivates.
to work around a SQLite issue that itself was
"fixed" as of sqlite 3.6.14, regarding quotes
surrounding a table name when using
the "foreign_key_list" pragma. The fix has been
adjusted to not interfere with quotes that
are *actually in the name* of a column or table,
to as much a degree as possible; sqlite still
doesn't return the correct result for foreign_key_list()
if the target table actually has quotes surrounding
its name, as *part* of its name (i.e. """mytable""").
[ticket:2568]
between a @declared_attr Column and a
directly-defined Column on a mixin. In both
cases, the Column will be applied to the
declared class' table, but not to that of a
joined inheritance subclass. Previously,
the directly-defined Column would be placed
on both the base and the sub table, which isn't
typically what's desired. [ticket:2565]
will now be produced via the func.* accessor
as well, as users naturally try to access these
names from func.* they might as well do
what's expected, even though the returned
object is not a FunctionElement.
[ticket:2562]
CREATE TABLE that provides access to the render for each
Column individually, by constructing a @compiles
function against the new schema.CreateColumn
construct. [ticket:2463]
setinputsizes() set can be customized by sending
a list of string DBAPI type names to exclude.
This list was previously fixed. The list also
now defaults to STRING, UNICODE, removing
CLOB, NCLOB from the list. [ticket:2469]
result-row targeting. It should be possible
to use a select() statement with string
based columns in it, that is
select(['id', 'name']).select_from('mytable'),
and have this statement be targetable by
Column objects with those names; this is the
mechanism by which
query(MyClass).from_statement(some_statement)
works. At some point the specific case of
using select(['id']), which is equivalent to
select([literal_column('id')]), stopped working
here, so this has been re-instated and of
course tested. [ticket:2558]
contains() operators to do a better job with
negation (NOT LIKE), and also to assemble them
at compilation time so that their rendered SQL
can be altered, such as in the case for Firebird
STARTING WITH [ticket:2470]
- [feature] firebird - The "startswith()" operator renders
as "STARTING WITH", "~startswith()" renders
as "NOT STARTING WITH", using FB's more efficient
operator. [ticket:2470]
Mike Bayer