1.1 Changelog¶
1.1.18¶
Released: March 6, 2018postgresql¶
mysql¶
MySQL dialects now query the server version using
SELECT @@version
explicitly to the server to ensure we are getting the correct version information back. Proxy servers like MaxScale interfere with the value that is passed to the DBAPI’s connection.server_version value so this is no longer reliable.References: #4205
1.1.17¶
Released: February 22, 2018Repaired regression caused in 1.2.3 and 1.1.16 regarding association proxy objects, revising the approach to #4185 when calculating the «owning class» of an association proxy to default to choosing the current class if the proxy object is not directly associated with a mapped class, such as a mixin.
References: #4185
1.1.16¶
Released: February 16, 2018orm¶
Fixed issue in post_update feature where an UPDATE is emitted when the parent object has been deleted but the dependent object is not. This issue has existed for a long time however since 1.2 now asserts rows matched for post_update, this was raising an error.
References: #4187
Fixed regression caused by fix for issue #4116 affecting versions 1.2.2 as well as 1.1.15, which had the effect of mis-calculation of the «owning class» of an
AssociationProxy
as theNoneType
class in some declarative mixin/inheritance situations as well as if the association proxy were accessed off of an un-mapped class. The «figure out the owner» logic has been replaced by an in-depth routine that searches through the complete mapper hierarchy assigned to the class or subclass to determine the correct (we hope) match; will not assign the owner if no match is found. An exception is now raised if the proxy is used against an un-mapped instance.References: #4185
Fixed bug where an object that is expunged during a rollback of a nested or subtransaction which also had its primary key mutated would not be correctly removed from the session, causing subsequent issues in using the session.
References: #4151
sql¶
Added
nullsfirst()
andnullslast()
as top level imports in thesqlalchemy.
andsqlalchemy.sql.
namespace. Pull request courtesy Lele Gaifax.Fixed bug in
Insert.values()
where using the «multi-values» format in combination withColumn
objects as keys rather than strings would fail. Pull request courtesy Aubrey Stark-Toller.References: #4162
postgresql¶
Added «SSL SYSCALL error: Operation timed out» to the list of messages that trigger a «disconnect» scenario for the psycopg2 driver. Pull request courtesy André Cruz.
Added «TRUNCATE» to the list of keywords accepted by the PostgreSQL dialect as an «autocommit»-triggering keyword. Pull request courtesy Jacob Hayes.
mysql¶
Fixed bug where the MySQL «concat» and «match» operators failed to propagate kwargs to the left and right expressions, causing compiler options such as «literal_binds» to fail.
References: #4136
misc¶
Fixed a fairly serious connection pool bug where a connection that is acquired after being refreshed as a result of a user-defined
DisconnectionError
or due to the 1.2-released «pre_ping» feature would not be correctly reset if the connection were returned to the pool by weakref cleanup (e.g. the front-facing object is garbage collected); the weakref would still refer to the previously invalidated DBAPI connection which would have the reset operation erroneously called upon it instead. This would lead to stack traces in the logs and a connection being checked into the pool without being reset, which can cause locking issues.References: #4184
1.1.15¶
Released: November 3, 2017orm¶
Fixed bug where the association proxy would inadvertently link itself to an
AliasedClass
object if it were called first with theAliasedClass
as a parent, causing errors upon subsequent usage.References: #4116
Fixed bug where ORM relationship would warn against conflicting sync targets (e.g. two relationships would both write to the same column) for sibling classes in an inheritance hierarchy, where the two relationships would never actually conflict during writes.
References: #4078
Fixed bug where correlated select used against single-table inheritance entity would fail to render correctly in the outer query, due to adjustment for single inheritance discriminator criteria inappropriately re-applying the criteria to the outer query.
References: #4103
orm declarative¶
Fixed bug where a descriptor that is elsewhere a mapped column or relationship within a hierarchy based on
AbstractConcreteBase
would be referred towards during a refresh operation, causing an error as the attribute is not mapped as a mapper property. A similar issue can arise for other attributes like the «type» column added byAbstractConcreteBase
if the class fails to include «concrete=True» in its mapper, however the check here should also prevent that scenario from causing a problem.References: #4124
sql¶
Fixed bug where
__repr__
ofColumnDefault
would fail if the argument were a tuple. Pull request courtesy Nicolas Caniart.References: #4126
Fixed bug where the recently added
ColumnOperators.any_()
andColumnOperators.all_()
methods didn’t work when called as methods, as opposed to using the standalone functionsany_()
andall_()
. Also added documentation examples for these relatively unintuitive SQL operators.References: #4093
postgresql¶
Made further fixes to the
ARRAY
class in conjunction with COLLATE, as the fix made in #4006 failed to accommodate for a multidimensional array.References: #4006
Fixed bug in
array_agg
function where passing an argument that is already of typeARRAY
, such as a PostgreSQLarray
construct, would produce aValueError
, due to the function attempting to nest the arrays.References: #4107
Fixed bug in PostgreSQL
Insert.on_conflict_do_update()
which would prevent the insert statement from being used as a CTE, e.g. viaInsert.cte()
, within another statement.References: #4074
mysql¶
Warning emitted when MariaDB 10.2.8 or earlier in the 10.2 series is detected as there are major issues with CHECK constraints within these versions that were resolved as of 10.2.9.
Note that this changelog message was NOT released with SQLAlchemy 1.2.0b3 and was added retroactively.
References: #4097
MySQL 5.7.20 now warns for use of the @tx_isolation variable; a version check is now performed and uses @transaction_isolation instead to prevent this warning.
References: #4120
Fixed issue where CURRENT_TIMESTAMP would not reflect correctly in the MariaDB 10.2 series due to a syntax change, where the function is now represented as
current_timestamp()
.References: #4096
MariaDB 10.2 now supports CHECK constraints (warning: use version 10.2.9 or greater due to upstream issues noted in #4097). Reflection now takes these CHECK constraints into account when they are present in the
SHOW CREATE TABLE
output.References: #4098
sqlite¶
Fixed bug where SQLite CHECK constraint reflection would fail if the referenced table were in a remote schema, e.g. on SQLite a remote database referred to by ATTACH.
References: #4099
mssql¶
Added a full range of «connection closed» exception codes to the PyODBC dialect for SQL Server, including „08S01“, „01002“, „08003“, „08007“, „08S02“, „08001“, „HYT00“, „HY010“. Previously, only „08S01“ was covered.
References: #4095
1.1.14¶
Released: September 5, 2017orm¶
Fixed bug in
Session.merge()
following along similar lines as that of #4030, where an internal check for a target object in the identity map could lead to an error if it were to be garbage collected immediately before the merge routine actually retrieves the object.References: #4069
Fixed bug where an
undefer_group()
option would not be recognized if it extended from a relationship that was loading using joined eager loading. Additionally, as the bug led to excess work being performed, Python function call counts are also improved by 20% within the initial calculation of result set columns, complementing the joined eager load improvements of #3915.References: #4048
Fixed race condition in ORM identity map which would cause objects to be inappropriately removed during a load operation, causing duplicate object identities to occur, particularly under joined eager loading which involves deduplication of objects. The issue is specific to garbage collection of weak references and is observed only under the PyPy interpreter.
References: #4068
Fixed bug in
Session.merge()
where objects in a collection that had the primary key attribute set toNone
for a key that is typically autoincrementing would be considered to be a database-persisted key for part of the internal deduplication process, causing only one object to actually be inserted in the database.References: #4056
An
InvalidRequestError
is raised when asynonym()
is used against an attribute that is not against aMapperProperty
, such as an association proxy. Previously, a recursion overflow would occur trying to locate non-existent attributes.References: #4067
sql¶
Altered the range specification for window functions to allow for two of the same PRECEDING or FOLLOWING keywords in a range by allowing for the left side of the range to be positive and for the right to be negative, e.g. (1, 3) is «1 FOLLOWING AND 3 FOLLOWING».
References: #4053
1.1.13¶
Released: August 3, 2017oracle¶
Fixed performance regression caused by the fix for #3937 where cx_Oracle as of version 5.3 dropped the
.UNICODE
symbol from its namespace, which was interpreted as cx_Oracle’s «WITH_UNICODE» mode being turned on unconditionally, which invokes functions on the SQLAlchemy side which convert all strings to unicode unconditionally and causing a performance impact. In fact, per cx_Oracle’s author the «WITH_UNICODE» mode has been removed entirely as of 5.1, so the expensive unicode conversion functions are no longer necessary and are disabled if cx_Oracle 5.1 or greater is detected under Python 2. The warning against «WITH_UNICODE» mode that was removed under #3937 is also restored.This change is also backported to: 1.0.19
References: #4035
1.1.12¶
Released: July 24, 2017orm¶
Fixed regression from 1.1.11 where adding additional non-entity columns to a query that includes an entity with subqueryload relationships would fail, due to an inspection added in 1.1.11 as a result of #4011.
References: #4033
Fixed bug involving JSON NULL evaluation logic added in 1.1 as part of #3514 where the logic would not accommodate ORM mapped attributes named differently from the
Column
that was mapped.References: #4031
Added
KeyError
checks to all methods withinWeakInstanceDict
where a check forkey in dict
is followed by indexed access to that key, to guard against a race against garbage collection that under load can remove the key from the dict after the code assumes its present, leading to very infrequentKeyError
raises.References: #4030
oracle¶
Added new keywords
Sequence.cache
andSequence.order
toSequence
, to allow rendering of the CACHE parameter understood by Oracle and PostgreSQL, and the ORDER parameter understood by Oracle. Pull request courtesy David Moore.
tests¶
Fixed issue in testing fixtures which was incompatible with a change made as of Python 3.6.2 involving context managers.
This change is also backported to: 1.0.18
References: #4034
1.1.11¶
Released: Monday, June 19, 2017orm¶
Fixed issue with subquery eagerloading which continues on from the series of issues fixed in #2699, #3106, #3893 involving that the «subquery» contains the correct FROM clause when beginning from a joined inheritance subclass and then subquery eager loading onto a relationship from the base class, while the query also includes criteria against the subclass. The fix in the previous tickets did not accommodate for additional subqueryload operations loading more deeply from the first level, so the fix has been further generalized.
References: #4011
sql¶
Fixed AttributeError which would occur in
WithinGroup
construct during an iteration of the structure.References: #4012
postgresql¶
Continuing with the fix that correctly handles PostgreSQL version string «10devel» released in 1.1.8, an additional regexp bump to handle version strings of the form «10beta1». While PostgreSQL now offers better ways to get this information, we are sticking w/ the regexp at least through 1.1.x for the least amount of risk to compatibility w/ older or alternate PostgreSQL databases.
References: #4005
Fixed bug where using
ARRAY
with a string type that features a collation would fail to produce the correct syntax within CREATE TABLE.References: #4006
mysql¶
MySQL 5.7 has introduced permission limiting for the «SHOW VARIABLES» command; the MySQL dialect will now handle when SHOW returns no row, in particular for the initial fetch of SQL_MODE, and will emit a warning that user permissions should be modified to allow the row to be present.
References: #4007
mssql¶
Fixed bug where SQL Server transaction isolation must be fetched from a different view when using Azure data warehouse, the query is now attempted against both views and then a NotImplemented is raised unconditionally if failure continues to provide the best resiliency against future arbitrary API changes in new SQL Server versions.
References: #3994
Added a placeholder type
XML
to the SQL Server dialect, so that a reflected table which includes this type can be re-rendered as a CREATE TABLE. The type has no special round-trip behavior nor does it currently support additional qualifying arguments.References: #3973
oracle¶
Support for two-phase transactions has been removed entirely for cx_Oracle when version 6.0b1 or later of the DBAPI is in use. The two- phase feature historically has never been usable under cx_Oracle 5.x in any case, and cx_Oracle 6.x has removed the connection-level «twophase» flag upon which this feature relied.
References: #3997
1.1.10¶
Released: Friday, May 19, 2017orm¶
Fixed bug where a cascade such as «delete-orphan» (but others as well) would fail to locate an object linked to a relationship that itself is local to a subclass in an inheritance relationship, thus causing the operation to not take place.
References: #3986
schema¶
An
ArgumentError
is now raised if aForeignKeyConstraint
object is created with a mismatched number of «local» and «remote» columns, which otherwise causes the internal state of the constraint to be incorrect. Note that this also impacts the condition where a dialect’s reflection process produces a mismatched set of columns for a foreign key constraint.References: #3949
postgresql¶
Added «autocommit» support for GRANT, REVOKE keywords. Pull request courtesy Jacob Hayes.
mysql¶
Removed an ancient and unnecessary intercept of the UTC_TIMESTAMP MySQL function, which was getting in the way of using it with a parameter.
References: #3966
Fixed bug in MySQL dialect regarding rendering of table options in conjunction with PARTITION options when rendering CREATE TABLE. The PARTITION related options need to follow the table options, whereas previously this ordering was not enforced.
References: #3961
oracle¶
Fixed bug in cx_Oracle dialect where version string parsing would fail for cx_Oracle version 6.0b1 due to the «b» character. Version string parsing is now via a regexp rather than a simple split.
References: #3975
misc¶
Protected against testing «None» as a class in the case where declarative classes are being garbage collected and new automap prepare() operations are taking place concurrently, very infrequently hitting a weakref that has not been fully acted upon after gc.
References: #3980
1.1.9¶
Released: April 4, 2017sql¶
Fixed regression released in 1.1.5 due to #3859 where adjustments to the «right-hand-side» evaluation of an expression based on
Variant
to honor the underlying type’s «right-hand-side» rules caused theVariant
type to be inappropriately lost, in those cases when we do want the left-hand side type to be transferred directly to the right hand side so that bind-level rules can be applied to the expression’s argument.References: #3952
Changed the mechanics of
ResultProxy
to unconditionally delay the «autoclose» step until theConnection
is done with the object; in the case where PostgreSQL ON CONFLICT with RETURNING returns no rows, autoclose was occurring in this previously non-existent use case, causing the usual autocommit behavior that occurs unconditionally upon INSERT/UPDATE/DELETE to fail.References: #3955
misc¶
1.1.8¶
Released: March 31, 2017postgresql¶
Added support for parsing the PostgreSQL version string for a development version like «PostgreSQL 10devel». Pull request courtesy Sean McCully.
misc¶
Fixed bug in
sqlalchemy.ext.mutable
where theMutable.as_mutable()
method would not track a type that had been copied usingTypeEngine.copy()
. This became more of a regression in 1.1 compared to 1.0 because theTypeDecorator
class is now a subclass ofSchemaEventTarget
, which among other things indicates to the parentColumn
that the type should be copied when theColumn
is. These copies are common when using declarative with mixins or abstract classes.References: #3950
Added support for bound parameters, e.g. those normally set up via
Query.params()
, to theResult.count()
method. Previously, support for parameters were omitted. Pull request courtesy Pat Deegan.
1.1.7¶
Released: March 27, 2017orm¶
An
aliased()
construct can now be passed to theQuery.select_entity_from()
method. Entities will be pulled from the selectable represented by thealiased()
construct. This allows special options foraliased()
such asaliased.adapt_on_names
to be used in conjunction withQuery.select_entity_from()
.References: #3933
Fixed a race condition which could occur under threaded environments as a result of the caching added via #3915. An internal collection of
Column
objects could be regenerated on an alias object inappropriately, confusing a joined eager loader when it attempts to render SQL and collect results and resulting in an attribute error. The collection is now generated up front before the alias object is cached and shared among threads.References: #3947
engine¶
Added an exception handler that will warn for the «cause» exception on Py2K when the «autorollback» feature of
Connection
itself raises an exception. In Py3K, the two exceptions are naturally reported by the interpreter as one occurring during the handling of the other. This is continuing with the series of changes for rollback failure handling that were last visited as part of #2696 in 1.0.12.References: #3946
sql¶
Added support for the
Variant
and theSchemaType
objects to be compatible with each other. That is, a variant can be created against a type likeEnum
, and the instructions to create constraints and/or database-specific type objects will propagate correctly as per the variant’s dialect mapping.References: #2892
Fixed bug in compiler where the string identifier of a savepoint would be cached in the identifier quoting dictionary; as these identifiers are arbitrary, a small memory leak could occur if a single
Connection
had an unbounded number of savepoints used, as well as if the savepoint clause constructs were used directly with an unbounded umber of savepoint names. The memory leak does not impact the vast majority of cases as normally theConnection
, which renders savepoint names with a simple counter starting at «1», is used on a per-transaction or per-fixed-number-of-transactions basis before being discarded.References: #3931
Fixed bug in new «schema translate» feature where the translated schema name would be invoked in terms of an alias name when rendered along with a column expression; occurred only when the source translate name was «None». The «schema translate» feature now only takes effect for
SchemaItem
andSchemaType
subclasses, that is, objects that correspond to a DDL-creatable structure in a database.References: #3924
oracle¶
A fix to cx_Oracle’s WITH_UNICODE mode which was uncovered by the fact that cx_Oracle 5.3 now seems to hardcode this flag on in the build; an internal method that uses this mode wasn’t using the correct signature.
This change is also backported to: 1.0.18
References: #3937
1.1.6¶
Released: February 28, 2017orm¶
Addressed some long unattended performance concerns within the joined eager loader query construction system that have accumulated since earlier versions as a result of increased abstraction. The use of ad- hoc
AliasedClass
objects per query, which produces lots of column lookup overhead each time, has been replaced with a cached approach that makes use of a small pool ofAliasedClass
objects that are reused between invocations of joined eager loading. Some mechanics involving eager join path construction have also been optimized. Callcounts for an end-to-end query construction + single row fetch test with a worst-case joined loader scenario have been reduced by about 60% vs. 1.1.5 and 42% vs. that of 0.8.6.References: #3915
Fixed a major inefficiency in the «eager_defaults» feature whereby an unnecessary SELECT would be emitted for column values where the ORM had explicitly inserted NULL, corresponding to attributes that were unset on the object but did not have any server default specified, as well as expired attributes on update that nevertheless had no server onupdate set up. As these columns are not part of the RETURNING that eager_defaults tries to use, they should not be post-SELECTed either.
References: #3909
Fixed two closely related bugs involving the mapper eager_defaults flag in conjunction with single-table inheritance; one where the eager defaults logic would inadvertently try to access a column that’s part of the mapper’s «exclude_properties» list (used by Declarative with single table inheritance) during the eager defaults fetch, and the other where the full load of the row in order to fetch the defaults would fail to use the correct inheriting mapper.
References: #3908
Fixed bug first introduced in 0.9.7 as a result of #3106 which would cause an incorrect query in some forms of multi-level subqueryload against aliased entities, with an unnecessary extra FROM entity in the innermost subquery.
References: #3893
orm declarative¶
Fixed bug where the «automatic exclude» feature of declarative that ensures a column local to a single table inheritance subclass does not appear as an attribute on other derivations of the base would not take effect for multiple levels of subclassing from the base.
References: #3895
sql¶
Fixed bug whereby the
DDLEvents.column_reflect()
event would not allow a non-textual expression to be passed as the value of the «default» for the new column, such as aFetchedValue
object to indicate a generic triggered default or atext()
construct. Clarified the documentation in this regard as well.References: #3905
postgresql¶
Added regular expressions for the «IMPORT FOREIGN SCHEMA», «REFRESH MATERIALIZED VIEW» PostgreSQL statements so that they autocommit when invoked via a connection or engine without an explicit transaction. Pull requests courtesy Frazer McLean and Paweł Stiasny.
References: #3804
Fixed bug in PostgreSQL
ExcludeConstraint
where the «whereclause» and «using» parameters would not be copied during an operation likeTable.tometadata()
.References: #3900
mysql¶
Added new MySQL 8.0 reserved words to the MySQL dialect for proper quoting. Pull request courtesy Hanno Schlichting.
mssql¶
Added a version check to the «get_isolation_level» feature, which is invoked upon first connect, so that it skips for SQL Server version 2000, as the necessary system view is not available prior to SQL Server 2005.
References: #3898
misc¶
1.1.5¶
Released: January 17, 2017orm¶
Fixed bug involving joined eager loading against multiple entities when polymorphic inheritance is also in use which would throw «„NoneType“ object has no attribute „isa“». The issue was introduced by the fix for #3611.
This change is also backported to: 1.0.17
References: #3884
Fixed bug in subquery loading where an object encountered as an «existing» row, e.g. already loaded from a different path in the same query, would not invoke subquery loaders for unloaded attributes that specified this loading. This issue is in the same area as that of #3431, #3811 which involved similar issues with joined loading.
References: #3854
The
Session.no_autoflush
context manager now ensures that the autoflush flag is reset within a «finally» block, so that if an exception is raised within the block, the state still resets appropriately. Pull request courtesy Emin Arakelian.Fixed bug where the single-table inheritance query criteria would not be inserted into the query in the case that the
Bundle
construct were used as the selection criteria.References: #3874
Fixed bug related to #3177, where a UNION or other set operation emitted by a
Query
would apply «single-inheritance» criteria to the outside of the union (also referencing the wrong selectable), even though this criteria is now expected to be already present on the inside subqueries. The single-inheritance criteria is now omitted once union() or another set operation is called againstQuery
in the same way asQuery.from_self()
.References: #3856
examples¶
Fixed two issues with the versioned_history example, one is that the history table now gets autoincrement=False to avoid 1.1’s new errors regarding composite primary keys with autoincrement; the other is that the sqlite_autoincrement flag is now used to ensure on SQLite, unique identifiers are used for the lifespan of a table even if some rows are deleted. Pull request courtesy Carlos García Montoro.
References: #3872
engine¶
The «extend_existing» option of
Table
reflection would cause indexes and constraints to be doubled up in the case that the parameter were used withMetaData.reflect()
(as the automap extension does) due to tables being reflected both within the foreign key path as well as directly. A new de-duplicating set is passed through within theMetaData.reflect()
sequence to prevent double reflection in this way.References: #3861
sql¶
Fixed bug originally introduced in 0.9 via #1068 where order_by(<some Label()>) would order by the label name based on name alone, that is, even if the labeled expression were not at all the same expression otherwise present, implicitly or explicitly, in the selectable. The logic that orders by label now ensures that the labeled expression is related to the one that resolves to that name before ordering by the label name; additionally, the name has to resolve to an actual label explicit in the expression elsewhere, not just a column name. This logic is carefully kept separate from the order by(textual name) feature that has a slightly different purpose.
References: #3882
Fixed 1.1 regression where «import *» would not work for sqlalchemy.sql.expression, due to mis-spelled
any_
andall_
functions.References: #3878
The engine URL embedded in the exception for «could not reflect» in
MetaData.reflect()
now conceals the password; also the__repr__
forTLEngine
now acts like that ofEngine
, concealing the URL password. Pull request courtesy Valery Yundin.Fixed issue in
Variant
where the «right hand coercion» logic, inherited fromTypeDecorator
, would coerce the right-hand side into theVariant
itself, rather than what the default type for theVariant
would do. In the case ofVariant
, we want the type to act mostly like the base type so the default logic ofTypeDecorator
is now overridden to fall back to the underlying wrapped type’s logic. Is mostly relevant for JSON at the moment.References: #3859
Fixed bug where literal_binds compiler flag was not honored by the
Insert
construct for the «multiple values» feature; the subsequent values are now rendered as literals.References: #3880
postgresql¶
Fixed bug in new «ON CONFLICT DO UPDATE» feature where the «set» values for the UPDATE clause would not be subject to type-level processing, as normally takes effect to handle both user-defined type level conversions as well as dialect-required conversions, such as those required for JSON datatypes. Additionally, clarified that the keys in the
set_
dictionary should match the «key» of the column, if distinct from the column name. A warning is emitted for remaining column names that don’t match column keys; for compatibility reasons, these are emitted as they were previously.References: #3888
The
TIME
andTIMESTAMP
datatypes now support a setting of zero for «precision»; previously a zero would be ignored. Pull request courtesy Ionuț Ciocîrlan.
mysql¶
Added a new parameter
mysql_prefix
supported by theIndex
construct, allows specification of MySQL-specific prefixes such as «FULLTEXT». Pull request courtesy Joseph Schorr.The MySQL dialect now will not warn when a reflected column has a «COMMENT» keyword on it, but note however the comment is not yet reflected; this is on the roadmap for a future release. Pull request courtesy Lele Long.
References: #3867
mssql¶
Fixed bug where SQL Server dialects would attempt to select the last row identity for an INSERT from SELECT, failing in the case when the SELECT has no rows. For such a statement, the inline flag is set to True indicating no last primary key should be fetched.
References: #3876
oracle¶
Fixed bug where an INSERT from SELECT where the source table contains an autoincrementing Sequence would fail to compile correctly.
References: #3877
Fixed bug where the «COMPRESSION» keyword was used in the ALL_TABLES query on Oracle 9.2; even though Oracle docs state table compression was introduced in 9i, the actual column is not present until 10.1.
References: #3875
firebird¶
Ported the fix for Oracle quoted-lowercase names to Firebird, so that a table name that is quoted as lower case can be reflected properly including when the table name comes from the get_table_names() inspection function.
References: #3548
misc¶
Fixed Python 3.6 DeprecationWarnings related to escaped strings without the „r“ modifier, and added test coverage for Python 3.6.
This change is also backported to: 1.0.17
References: #3886
1.1.4¶
Released: November 15, 2016orm¶
Fixed bug in
Session.bulk_update_mappings()
where an alternate-named primary key attribute would not track properly into the UPDATE statement.This change is also backported to: 1.0.16
References: #3849
Fixed bug in
Session.bulk_save()
where an UPDATE would not function correctly in conjunction with a mapping that implements a version id counter.This change is also backported to: 1.0.16
References: #3781
Fixed bug where the
Mapper.attrs
,Mapper.all_orm_descriptors
and other derived attributes would fail to refresh when mapper properties or other ORM constructs were added to the mapper/class after these accessors were first called.This change is also backported to: 1.0.16
References: #3778
Fixed regression in collections due to #3457 whereby deserialize during pickle or deepcopy would fail to establish all attributes of an ORM collection, causing further mutation operations to fail.
References: #3852
Fixed long-standing bug where the «noload» relationship loading strategy would cause backrefs and/or back_populates options to be ignored.
References: #3845
engine¶
Removed long-broken «default_schema_name()» method from
Connection
. This method was left over from a very old version and was non-working (e.g. would raise). Pull request courtesy Benjamin Dopplinger.
sql¶
postgresql¶
Fixed regression caused by the fix in #3807 (version 1.1.0) where we ensured that the tablename was qualified in the WHERE clause of the DO UPDATE portion of PostgreSQL’s ON CONFLICT, however you cannot put the table name in the WHERE clause in the actual ON CONFLICT itself. This was an incorrect assumption, so that portion of the change in #3807 is rolled back.
mysql¶
Added support for server side cursors to the mysqlclient and pymysql dialects. This feature is available via the
Connection.execution_options.stream_results
flag as well as theserver_side_cursors=True
dialect argument in the same way that it has been for psycopg2 on PostgreSQL. Pull request courtesy Roman Podoliaka.MySQL’s native ENUM type supports any non-valid value being sent, and in response will return a blank string. A hardcoded rule to check for «is returning the blank string» has been added to the MySQL implementation for ENUM so that this blank string is returned to the application rather than being rejected as a non-valid value. Note that if your MySQL enum is linking values to objects, you still get the blank string back.
References: #3841
sqlite¶
Added quotes to the PRAGMA directives in the pysqlcipher dialect to support additional cipher arguments appropriately. Pull request courtesy Kevin Jurczyk.
Added an optional import for the pysqlcipher3 DBAPI when using the pysqlcipher dialect. This package will attempt to be imported if the Python-2 only pysqlcipher DBAPI is non-present. Pull request courtesy Kevin Jurczyk.
mssql¶
Fixed bug in pyodbc dialect (as well as in the mostly non-working adodbapi dialect) whereby a semicolon present in the password or username fields could be interpreted as a separator for another token; the values are now quoted when semicolons are present.
This change is also backported to: 1.0.16
References: #3762
1.1.3¶
Released: October 27, 2016orm¶
Fixed regression caused by #2677 whereby calling
Session.delete()
on an object that was already flushed as deleted in that session would fail to set up the object in the identity map (or reject the object), causing flush errors as the object were in a state not accommodated by the unit of work. The pre-1.1 behavior in this case has been restored, which is that the object is put back into the identity map so that the DELETE statement will be attempted again, which emits a warning that the number of expected rows was not matched (unless the row were restored outside of the session).References: #3839
Fixed regression where some
Query
methods likeQuery.update()
and others would fail if theQuery
were against a series of mapped columns, rather than the mapped entity as a whole.References: #3836
sql¶
Fixed bug involving new value translation and validation feature in
Enum
whereby using the enum object in a string concatenation would maintain theEnum
type as the type of the expression overall, producing missing lookups. A string concatenation against anEnum
-typed column now usesString
as the datatype of the expression itself.References: #3833
Fixed regression which occurred as a side effect of #2919, which in the less typical case of a user-defined
TypeDecorator
that was also itself an instance ofSchemaType
(rather than the implementation being such) would cause the column attachment events to be skipped for the type itself.References: #3832
postgresql¶
PostgreSQL table reflection will ensure that the
Column.autoincrement
flag is set to False when reflecting a primary key column that is not of anInteger
datatype, even if the default is related to an integer-generating sequence. This can happen if a column is created as SERIAL and the datatype is changed. The autoincrement flag can only be True if the datatype is of integer affinity in the 1.1 series.References: #3835
1.1.2¶
Released: October 17, 2016orm¶
Fixed bug involving the rule to disable a joined collection eager loader on the other side of a many-to-one lazy loader, first added in #1495, where the rule would fail if the parent object had some other lazyloader-bound query options associated with it.
References: #3824
Fixed self-referential entity, deferred column loading issue in a similar style as that of #3431, #3811 where an entity is present in multiple positions within the row due to self-referential eager loading; when the deferred loader only applies to one of the paths, the «present» column loader will now override the deferred non- load for that entity regardless of row ordering.
References: #3822
sql¶
Fixed a regression caused by a newly added function that performs the «wrap callable» function of sql
DefaultGenerator
objects, an attribute error raised for__module__
when the default callable was afunctools.partial
or other object that doesn’t have a__module__
attribute.References: #3823
Fixed regression in
Enum
type where event handlers were not transferred in the case of the type object being copied, due to a conflicting copy() method added as part of #3250. This copy occurs normally in situations when the column is copied, such as in tometadata() or when using declarative mixins with columns. The event handler not being present would impact the constraint being created for a non-native enumerated type, but more critically the ENUM object on the PostgreSQL backend.References: #3827
postgresql¶
Changed the naming convention used when generating bound parameters for a multi-VALUES insert statement, so that the numbered parameter names don’t conflict with the anonymized parameters of a WHERE clause, as is now common in a PostgreSQL ON CONFLICT construct.
References: #3828
1.1.1¶
Released: October 7, 2016mssql¶
misc¶
Changed the CompileError raised when various primary key missing situations are detected to a warning. The statement is again passed to the database where it will fail and the DBAPI error (usually IntegrityError) raises as usual.
См.также
The .autoincrement directive is no longer implicitly enabled for a composite primary key column
References: #3216
1.1.0¶
Released: October 5, 2016orm¶
Enhanced the new «raise» lazy loader strategy to also include a «raise_on_sql» variant, available both via
relationship.lazy
as well asraiseload()
. This variant only raises if the lazy load would actually emit SQL, vs. raising if the lazy loader mechanism is invoked at all.References: #3812
The
Query.group_by()
method now resets the group by collection if an argument ofNone
is passed, in the same way thatQuery.order_by()
has worked for a long time. Pull request courtesy Iuri Diniz.Passing False to
Query.order_by()
in order to cancel all order by’s is deprecated; there is no longer any difference between calling this method with False or with None.Fixed bug where joined eager loading would fail for a polymorphically- loaded mapper, where the polymorphic_on was set to an un-mapped expression such as a CASE expression.
This change is also backported to: 1.0.16
References: #3800
Fixed bug where the ArgumentError raised for an invalid bind sent to a Session via
Session.bind_mapper()
,Session.bind_table()
, or the constructor would fail to be correctly raised.This change is also backported to: 1.0.16
References: #3798
Fixed bug in subquery eager loading where a subqueryload of an «of_type()» object linked to a second subqueryload of a plain mapped class, or a longer chain of several «of_type()» attributes, would fail to link the joins correctly.
This change is also backported to: 1.0.15
ORM attributes can now be assigned any object that is has a
__clause_element__()
attribute, which will result in inline SQL the way anyClauseElement
class does. This covers other mapped attributes not otherwise transformed by further expression constructs.References: #3802
Made an adjustment to the bug fix first introduced in [ticket:3431] that involves an object appearing in multiple contexts in a single result set, such that an eager loader that would set the related object value to be None will still fire off, thus satisfying the load of that attribute. Previously, the adjustment only honored a non-None value arriving for an eagerly loaded attribute in a secondary row.
References: #3811
Fixed bug in new
SessionEvents.persistent_to_deleted()
event where the target object could be garbage collected before the event is fired off.References: #3808
The primaryjoin of a
relationship()
construct can now include abindparam()
object that includes a callable function to generate values. Previously, the lazy loader strategy would be incompatible with this use, and additionally would fail to correctly detect if the «use_get» criteria should be used if the primary key were involved with the bound parameter.References: #3767
An UPDATE emitted from the ORM flush process can now accommodate a SQL expression element for a column within the primary key of an object, if the target database supports RETURNING in order to provide the new value, or if the PK value is set «to itself» for the purposes of bumping some other trigger / onupdate on the column.
References: #3801
Fixed bug where the «simple many-to-one» condition that allows lazy loading to use get() from identity map would fail to be invoked if the primaryjoin of the relationship had multiple clauses separated by AND which were not in the same order as that of the primary key columns being compared in each clause. This ordering difference occurs for a composite foreign key where the table-bound columns on the referencing side were not in the same order in the .c collection as the primary key columns on the referenced side….which in turn occurs a lot if one is using declarative mixins and/or declared_attr to set up columns.
References: #3788
An exception is raised when two
@validates
decorators on a mapping make use of the same name. Only one validator of a certain name at a time is supported, there’s no mechanism to chain these together, as the order of the validators at the level of function decorator can’t be made deterministic.References: #3776
Mapper errors raised during
configure_mappers()
now explicitly include the name of the originating mapper in the exception message to help in those situations where the wrapped exception does not itself include the source mapper. Pull request courtesy John Perkins.
orm declarative¶
Constructing a declarative base class that inherits from another class will also inherit its docstring. This means
as_declarative()
acts more like a normal class decorator.
sql¶
Fixed bug in
Table
where the internal method_reset_exported()
would corrupt the state of the object. This method is intended for selectable objects and is called by the ORM in some cases; an erroneous mapper configuration would could lead the ORM to call this on aTable
object.This change is also backported to: 1.0.15
References: #3755
Execution options can now be propagated from within a statement at compile time to the outermost statement, so that if an embedded element wants to set «autocommit» to be True for example, it can propagate this to the enclosing statement. Currently, this feature is enabled for a DML-oriented CTE embedded inside of a SELECT statement, e.g. INSERT/UPDATE/DELETE inside of SELECT.
References: #3805
A string sent as a column default via the
Column.server_default
parameter is now escaped for quotes.References: #3809
Added compiler-level flags used by PostgreSQL to place additional parenthesis than would normally be generated by precedence rules around operations involving JSON, HSTORE indexing operators as well as within their operands since it has been observed that PostgreSQL’s precedence rules for at least the HSTORE indexing operator is not consistent between 9.4 and 9.5.
References: #3806
The
BaseException
exception class is now intercepted by the exception-handling routines ofConnection
, and includes handling by theConnectionEvents.handle_error()
event. TheConnection
is now invalidated by default in the case of a system level exception that is not a subclass ofException
, includingKeyboardInterrupt
and the greenletGreenletExit
class, to prevent further operations from occurring upon a database connection that is in an unknown and possibly corrupted state. The MySQL drivers are most targeted by this change however the change is across all DBAPIs.References: #3803
The «eq» and «ne» operators are no longer part of the list of «associative» operators, while they remain considered to be «commutative». This allows an expression like
(x == y) == z
to be maintained at the SQL level with parenthesis. Pull request courtesy John Passaro.References: #3799
Stringify of expression with unnamed
Column
objects, as occurs in lots of situations including ORM error reporting, will now render the name in string context as «<name unknown>» rather than raising a compile error.References: #3789
Raise a more descriptive exception / message when ClauseElement or non-SQLAlchemy objects that are not «executable» are erroneously passed to
.execute()
; a new exception ObjectNotExecutableError is raised consistently in all cases.References: #3786
Fixed regression in JSON datatypes where the «literal processor» for a JSON index value would not be invoked. The native String and Integer datatypes are now called upon from within the JSONIndexType and JSONPathType. This is applied to the generic, PostgreSQL, and MySQL JSON types and also has a dependency on #3766.
References: #3765
Fixed bug where
Index
would fail to extract columns from compound SQL expressions if those SQL expressions were wrapped inside of an ORM-style__clause_element__()
construct. This bug exists in 1.0.x as well, however in 1.1 is more noticeable as hybrid_property @expression now returns a wrapped element.References: #3763
postgresql¶
An adjustment to ON CONFLICT such that the «inserted_primary_key» logic is able to accommodate the case where there’s no INSERT or UPDATE and there’s no net change. The value comes out as None in this case, rather than failing on an exception.
References: #3813
Fixed issue in new PG «on conflict» construct where columns including those of the «excluded» namespace would not be table-qualified in the WHERE clauses in the statement.
References: #3807
mysql¶
Added support for parsing MySQL/Connector boolean and integer arguments within the URL query string: connection_timeout, connect_timeout, pool_size, get_warnings, raise_on_warnings, raw, consume_results, ssl_verify_cert, force_ipv6, pool_reset_session, compress, allow_local_infile, use_pure.
This change is also backported to: 1.0.15
References: #3787
Fixed bug where the «literal_binds» flag would not be propagated to a CAST expression under MySQL.
References: #3766
mssql¶
Changed the query used to get «default schema name», from one that queries the database principals table to using the «schema_name()» function, as issues have been reported that the former system was unavailable on the Azure Data Warehouse edition. It is hoped that this will finally work across all SQL Server versions and authentication styles.
This change is also backported to: 1.0.16
References: #3810
Updated the server version info scheme for pyodbc to use SQL Server SERVERPROPERTY(), rather than relying upon pyodbc.SQL_DBMS_VER, which continues to be unreliable particularly with FreeTDS.
This change is also backported to: 1.0.16
References: #3814
Added error code 20017 «unexpected EOF from the server» to the list of disconnect exceptions that result in a connection pool reset. Pull request courtesy Ken Robbins.
This change is also backported to: 1.0.16
References: #3791
misc¶
Fixed bug where setting up a single-table inh subclass of a joined-table subclass which included an extra column would corrupt the foreign keys collection of the mapped table, thereby interfering with the initialization of relationships.
This change is also backported to: 1.0.16
References: #3797
1.1.0b3¶
Released: July 26, 2016orm¶
Removed a warning that dates back to 0.4 which emits when a same-named relationship is placed on two mappers that inherits via joined or single table inheritance. The warning does not apply to the current unit of work implementation.
References: #3749
sql¶
Fixed bug in new CTE feature for update/insert/delete stated as a CTE inside of an enclosing statement (typically SELECT) whereby oninsert and onupdate values weren’t called upon for the embedded statement.
References: #3745
Fixed bug in new CTE feature for update/insert/delete whereby an anonymous (e.g. no name passed)
CTE
construct around the statement would fail.References: #3744
postgresql¶
Fixed bug whereby
TypeDecorator
andVariant
types were not deeply inspected enough by the PostgreSQL dialect to determine if SMALLSERIAL or BIGSERIAL needed to be rendered rather than SERIAL.This change is also backported to: 1.0.14
References: #3739
oracle¶
Fixed bug in
Select.with_for_update.of
, where the Oracle «rownum» approach to LIMIT/OFFSET would fail to accommodate for the expressions inside the «OF» clause, which must be stated at the topmost level referring to expression within the subquery. The expressions are now added to the subquery if needed.This change is also backported to: 1.0.14
References: #3741
misc¶
Added a «default» parameter to the new sqlalchemy.ext.indexable extension.
Fixed bug in
sqlalchemy.ext.baked
where the unbaking of a subquery eager loader query would fail due to a variable scoping issue, when multiple subquery loaders were involved. Pull request courtesy Mark Hahnenberg.This change is also backported to: 1.0.15
References: #3743
sqlalchemy.ext.indexable will intercept IndexError as well as KeyError when raising as AttributeError.
1.1.0b2¶
Released: July 1, 2016sql¶
Fixed issue in SQL math negation operator where the type of the expression would no longer be the numeric type of the original. This would cause issues where the type determined result set behaviors.
This change is also backported to: 1.0.14
References: #3735
Fixed bug whereby the
__getstate__
/__setstate__
methods for sqlalchemy.util.Properties were non-working due to the transition in the 1.0 series to__slots__
. The issue potentially impacted some third-party applications. Pull request courtesy Pieter Mulder.This change is also backported to: 1.0.14
References: #3728
The processing performed by the
Boolean
datatype for backends that only feature integer types has been made consistent between the pure Python and C-extension versions, in that the C-extension version will accept any integer value from the database as a boolean, not just zero and one; additionally, non-boolean integer values being sent to the database are coerced to exactly zero or one, instead of being passed as the original integer value.References: #3730
Rolled back the validation rules a bit in
Enum
to allow unknown string values to pass through, unless the flagvalidate_string=True
is passed to the Enum; any other kind of object is still of course rejected. While the immediate use is to allow comparisons to enums with LIKE, the fact that this use exists indicates there may be more unknown-string-comparison use cases than we expected, which hints that perhaps there are some unknown string-INSERT cases too.References: #3725
postgresql¶
Made a slight behavioral change in the
sqlalchemy.ext.compiler
extension, whereby the existing compilation schemes for an established construct would be removed if that construct itself didn’t already have its own dedicated__visit_name__
. This was a rare occurrence in 1.0, however in 1.1ARRAY
subclassesARRAY
and has this behavior. As a result, setting up a compilation handler for another dialect such as SQLite would render the mainARRAY
object no longer compilable.References: #3732
mysql¶
Dialed back the «order the primary key columns per auto-increment» described in No more generation of an implicit KEY for composite primary key w/ AUTO_INCREMENT a bit, so that if the
PrimaryKeyConstraint
is explicitly defined, the order of columns is maintained exactly, allowing control of this behavior when necessary.References: #3726
1.1.0b1¶
Released: June 16, 2016orm¶
A new ORM extension Индексируемый is added, which allows construction of Python attributes which refer to specific elements of «indexed» structures such as arrays and JSON fields. Pull request courtesy Jeong YunWon.
См.также
Added new flag
Session.bulk_insert_mappings.render_nulls
which allows an ORM bulk INSERT to occur with NULL values rendered; this bypasses server side defaults, however allows all statements to be formed with the same set of columns, allowing them to be batched. Pull request courtesy Tobias Sauerwein.Added new event
AttributeEvents.init_scalar()
, as well as a new example suite illustrating its use. This event can be used to provide a Core-generated default value to a Python-side attribute before the object is persisted.References: #1311
Added
AutomapBase.prepare.schema
to theAutomapBase.prepare()
method, to indicate which schema tables should be reflected from if not the default schema. Pull request courtesy Josh Marlow.Added new parameter
mapper.passive_deletes
to available mapper options. This allows a DELETE to proceed for a joined-table inheritance mapping against the base table only, while allowing for ON DELETE CASCADE to handle deleting the row from the subclass tables.References: #2349
Calling str() on a core SQL construct has been made more «friendly», when the construct contains non-standard SQL elements such as RETURNING, array index operations, or dialect-specific or custom datatypes. A string is now returned in these cases rendering an approximation of the construct (typically the PostgreSQL-style version of it) rather than raising an error.
References: #3631
The
str()
call forQuery
will now take into account theEngine
to which theSession
is bound, when generating the string form of the SQL, so that the actual SQL that would be emitted to the database is shown, if possible. Previously, only the engine associated with theMetaData
to which the mappings are associated would be used, if present. If no bind can be located either on theSession
or on theMetaData
to which the mappings are associated, then the «default» dialect is used to render the SQL, as was the case previously.References: #3081
The
SessionEvents
suite now includes events to allow unambiguous tracking of all object lifecycle state transitions in terms of theSession
itself, e.g. pending, transient, persistent, detached. The state of the object within each event is also defined.См.также
References: #2677
Added a new session lifecycle state deleted. This new state represents an object that has been deleted from the persistent state and will move to the detached state once the transaction is committed. This resolves the long-standing issue that objects which were deleted existed in a gray area between persistent and detached. The
InstanceState.persistent
accessor will no longer report on a deleted object as persistent; theInstanceState.deleted
accessor will instead be True for these objects, until they become detached.См.также
References: #2677
Added new checks for the common error case of passing mapped classes or mapped instances into contexts where they are interpreted as SQL bound parameters; a new exception is raised for this.
References: #3321
Added new relationship loading strategy
raiseload()
(also accessible vialazy='raise'
). This strategy behaves almost likenoload()
but instead of returningNone
it raises an InvalidRequestError. Pull request courtesy Adrian Moennich.References: #3512
The
Mapper.order_by
parameter is deprecated. This is an old parameter no longer relevant to how SQLAlchemy works, once the Query object was introduced. By deprecating it we establish that we aren’t supporting non-working use cases and that we encourage applications to move off of the use of this parameter.См.также
References: #3394
The
Session.weak_identity_map
parameter is deprecated. See the new recipe at Поведение при обращении к сеансам for an event-based approach to maintaining strong identity map behavior.См.также
References: #2677
Fixed an issue where a many-to-one change of an object from one parent to another could work inconsistently when combined with an un-flushed modification of the foreign key attribute. The attribute move now considers the database-committed value of the foreign key in order to locate the «previous» parent of the object being moved. This allows events to fire off correctly including backref events. Previously, these events would not always fire. Applications which may have relied on the previously broken behavior may be affected.
References: #3708
Fixed bug where deferred columns would inadvertently be set up for database load on the next object-wide unexpire, when the object were merged into the session with
session.merge(obj, load=False)
.References: #3488
Further continuing on the common MySQL exception case of a savepoint being cancelled first covered in #2696, the failure mode in which the
Session
is placed when a SAVEPOINT vanishes before rollback has been improved to allow theSession
to still function outside of that savepoint. It is assumed that the savepoint operation failed and was cancelled.References: #3680
Fixed bug where a newly inserted instance that is rolled back would still potentially cause persistence conflicts on the next transaction, because the instance would not be checked that it was expired. This fix will resolve a large class of cases that erroneously cause the «New instance with identity X conflicts with persistent instance Y» error.
References: #3677
An improvement to the workings of
Query.correlate()
such that when a «polymorphic» entity is used which represents a straight join of several tables, the statement will ensure that all the tables within the join are part of what’s correlating.References: #3662
Fixed bug which would cause an eagerly loaded many-to-one attribute to not be loaded, if the joined eager load were from a row where the same entity were present multiple times, some calling for the attribute to be eagerly loaded and others not. The logic here is revised to take in the attribute even though a different loader path has handled the parent entity already.
References: #3431
A refinement to the logic which adds columns to the resulting SQL when
Query.distinct()
is combined withQuery.order_by()
such that columns which are already present will not be added a second time, even if they are labeled with a different name. Regardless of this change, the extra columns added to the SQL have never been returned in the final result, so this change only impacts the string form of the statement as well as its behavior when used in a Core execution context. Additionally, columns are no longer added when the DISTINCT ON format is used, provided the query is not wrapped inside a subquery due to joined eager loading.References: #3641
Fixed issue where two same-named relationships that refer to a base class and a concrete-inherited subclass would raise an error if those relationships were set up using «backref», while setting up the identical configuration using relationship() instead with the conflicting names would succeed, as is allowed in the case of a concrete mapping.
См.также
Same-named backrefs will not raise an error when applied to concrete inheritance subclasses
References: #3630
The
Session.merge()
method now tracks pending objects by primary key before emitting an INSERT, and merges distinct objects with duplicate primary keys together as they are encountered, which is essentially semi-deterministic at best. This behavior matches what happens already with persistent objects.References: #3601
Fixed bug where the «single table inheritance» criteria would be added onto the end of a query in some inappropriate situations, such as when querying from an exists() of a single-inheritance subclass.
References: #3582
Added a new type-level modifier
TypeEngine.evaluates_none()
which indicates to the ORM that a positive set of None should be persisted as the value NULL, instead of omitting the column from the INSERT statement. This feature is used both as part of the implementation for #3514 as well as a standalone feature available on any type.References: #3250
Internal calls to «bookkeeping» functions within
Session.bulk_save_objects()
and related bulk methods have been scaled back to the extent that this functionality is not currently used, e.g. checks for column default values to be fetched after an INSERT or UPDATE statement.References: #3526
Additional fixes have been made regarding the value of
None
in conjunction with the PostgreSQLJSON
type. When theJSON.none_as_null
flag is left at its default value ofFalse
, the ORM will now correctly insert the JSON «„null“» string into the column whenever the value on the ORM object is set to the valueNone
or when the valueNone
is used withSession.bulk_insert_mappings()
, including if the column has a default or server default on it.См.также
JSON «null» is inserted as expected with ORM operations, omitted when not present
New options allowing explicit persistence of NULL over a default
References: #3514
engine¶
Added connection pool events
ConnectionEvents.close()
,ConnectionEvents.detach()
,ConnectionEvents.close_detached()
.All string formatting of bound parameter sets and result rows for logging, exception, and
repr()
purposes now truncate very large scalar values within each collection, including an «N characters truncated» notation, similar to how the display for large multiple-parameter sets are themselves truncated.References: #2837
Multi-tenancy schema translation for
Table
objects is added. This supports the use case of an application that uses the same set ofTable
objects in many schemas, such as schema-per-user. A new execution optionConnection.execution_options.schema_translate_map
is added.References: #2685
Added a new entrypoint system to the engine to allow «plugins» to be stated in the query string for a URL. Custom plugins can be written which will be given the chance up front to alter and/or consume the engine’s URL and keyword arguments, and then at engine create time will be given the engine itself to allow additional modifications or event registration. Plugins are written as a subclass of
CreateEnginePlugin
; see that class for details.References: #3536
sql¶
Added TABLESAMPLE support via the new
FromClause.tablesample()
method and standalone function. Pull request courtesy Ilja Everilä.См.также
References: #3718
Added support for ranges in window functions, using the
over.range_
andover.rows
parameters.References: #3049
Implemented reflection of CHECK constraints for SQLite and PostgreSQL. This is available via the new inspector method
Inspector.get_check_constraints()
as well as when reflectingTable
objects in the form ofCheckConstraint
objects present in the constraints collection. Pull request courtesy Alex Grönholm.Реализовано отражение ограничений CHECK для SQLite и PostgreSQL. Это доступно через новый метод инспектора
ColumnOperators.is_distinct_from()
, а также при отраженииColumnOperators.isnot_distinct_from()
объектов в виде объектов, присутствующих в коллекции ограничений. Pull request любезно предоставлен Алексом Грёнхольмом.Добавлен хук в
DDLCompiler.visit_create_table()
под названиемDDLCompiler.create_table_suffix()
, позволяющий пользовательским диалектам добавлять ключевые слова после предложения «CREATE TABLE». Pull request любезно предоставлен Марком Санданом.Отрицательные целочисленные индексы теперь учитываются в строках, возвращаемых из
ResultProxy
. Pull request любезно предоставлен Emanuele Gaifas.Добавлены
Select.lateral()
и связанные с ними конструкции, позволяющие использовать ключевое слово LATERAL стандарта SQL, которое в настоящее время поддерживается только PostgreSQL.См.также
References: #2857
Добавлена поддержка рендеринга «FULL OUTER JOIN» как в Core, так и в ORM. Pull request любезно предоставлен Стефаном Урбанеком.
References: #1957
Функциональность CTE была расширена для поддержки всего DML, позволяя операторам INSERT, UPDATE и DELETE указывать свое собственное предложение WITH, а также сами эти операторы могут быть выражениями CTE, если они включают предложение RETURNING.
References: #2551
В тип данных
enum.Enum
добавлена поддержка перечислимых классов в стиле PEP-435, а именно классаEnum
в Python 3, а также совместимых библиотек перечислений. Тип данныхEnum
теперь также выполняет проверку входящих значений в Python и добавляет возможность отказаться от создания ограничения CHECKEnum.create_constraint
. Pull request любезно предоставлен Алексом Грёнхольмом.Глубокое усовершенствование недавно добавленного метода
TextClause.columns()
и его взаимодействия с обработкой результирующих строк позволяет теперь позиционно сопоставлять столбцы, передаваемые в метод, со столбцами результата в операторе, а не сопоставлять только по имени. Преимущество этого заключается в том, что при связывании текстового SQL-оператора с ORM или Core табличной моделью не требуется система маркировки или дедупликации общих имен столбцов, что также означает отсутствие необходимости беспокоиться о соответствии имен таблиц столбцам ORM и так далее. Кроме того,ResultProxy
был усовершенствован для сопоставления ключей столбцов и строк со строкой с большей точностью в некоторых случаях.См.также
ResultSet column matching enhancements; positional column setup for textual SQL - обзор функций
TextClause.columns() will match columns positionally, not by name, when passed positionally - замечания по обратной совместимости
References: #3501
Добавлен новый тип в ядро
JSON
. Он является основой типа PostgreSQLJSON
, а также нового типаJSON
, чтобы можно было использовать PG/MySQL-агностичный JSON-столбец. Тип поддерживает базовый поиск по индексу и пути.См.также
References: #3619
Добавлена поддержка функций «набор-агрегат» вида
<function> WITHIN GROUP (ORDER BY <criteria>)
, использующих методFunctionElement.within_group()
. Добавлена серия распространенных функций «набор-агрегат» с возвращаемыми типами, производными от набора. Сюда входят функции типаpercentile_cont
,dense_rank
и другие.References: #1370
Добавлена поддержка стандартной для SQL функции
array_agg
, которая автоматически возвращаетARRAY
правильного типа и поддерживает операции с индексами / срезами, а такжеarray_agg()
, которая возвращаетARRAY
с дополнительными возможностями сравнения. Поскольку массивы на данный момент поддерживаются только на PostgreSQL, фактически работает только на PostgreSQL. Также добавлена новая конструкцияaggregate_order_by
для поддержки расширения PG «ORDER BY».References: #3132
Добавлен новый тип в ядро
ARRAY
. Это основа типа PostgreSQLARRAY
, и теперь он является частью Core, чтобы начать поддержку различных возможностей, поддерживающих массивы по стандарту SQL, включая некоторые функции и возможную поддержку собственных массивов в других базах данных, где есть понятие «массив», таких как DB2 или Oracle. Кроме того, были добавлены новые операторыany_()
иall_()
. Они поддерживают не только конструкции массивов в PostgreSQL, но и подзапросы, которые можно использовать в MySQL (но, к сожалению, не в PostgreSQL).References: #3516
Система, по которой колонка
Column
считается колонкой с «автоинкрементом», была изменена таким образом, что автоинкременты больше не включаются неявно для колонокTable
, имеющих составной первичный ключ. Чтобы обеспечить возможность включения автоинкремента для составного столбца-члена PK и в то же время сохранить давнее поведение SQLAlchemy по включению неявного автоинкремента для единственного целочисленного первичного ключа, к параметруColumn.autoincrement
"auto"
было добавлено третье состояние, которое теперь используется по умолчанию.См.также
The .autoincrement directive is no longer implicitly enabled for a composite primary key column
No more generation of an implicit KEY for composite primary key w/ AUTO_INCREMENT
References: #3216
FromClause.count()
является устаревшей. Эта функция использует произвольный столбец в таблице и не является надежной; для использования в ядре следует предпочестьfunc.count()
.References: #3724
Исправлено утверждение, которое поднималось несколько неуместно, если индекс
Index
был связан сColumn
, который связан сTableClause
в нижнем регистре; ассоциация должна игнорироваться для целей связывания индекса сTable
.References: #3616
Конструкция
type_coerce()
теперь является полноценным элементом выражения Core, который подвергается поздней оценке во время компиляции. Ранее эта функция была только функцией преобразования, которая обрабатывала различные входы выражения, возвращая либоLabel
ориентированного на столбец выражения, либо копию данного объектаBindParameter
, что, в частности, не позволяло логически поддерживать операцию, когда преобразование выражения на уровне ORM преобразовывало столбец в связанный параметр (например, для ленивой загрузки).References: #3531
Расширитель типа
TypeDecorator
теперь будет работать совместно с реализациейSchemaType
, обычноEnum
илиBoolean
, обеспечивая распространение событий для каждой таблицы от типа реализации к внешнему типу. Эти события используются для обеспечения того, чтобы ограничения или типы PostgreSQL (например, ENUM) были правильно созданы (и, возможно, удалены) вместе с родительской таблицей.References: #2919
Поведение конструкции
union()
и связанных с ней конструкций, таких какQuery.union()
, теперь обрабатывает случай, когда встроенные операторы SELECT необходимо заключать в круглые скобки из-за того, что они включают LIMIT, OFFSET и/или ORDER BY. Эти запросы не работают на SQLite, и на этом бэкенде, как и раньше, будут выполняться с ошибками, но теперь они должны работать на всех других бэкендах.См.также
A UNION or similar of SELECTs with LIMIT/OFFSET/ORDER BY now parenthesizes the embedded selects
References: #2528
schema¶
Функции генерации по умолчанию, передаваемые объектам
Column
, теперь запускаются через «update_wrapper» или эквивалентную функцию, если передана вызываемая не-функция, так что инструменты интроспекции сохраняют имя и docstring обернутой функции. Pull request courtesy hsum.
postgresql¶
Добавлена поддержка PostgreSQL’s INSERT..ON CONFLICT с использованием нового специфического для PostgreSQL объекта
Insert
. Pull-запрос и значительные усилия здесь приложены Робином Томасом.References: #3529
DDL для DROP INDEX будет выдавать «CONCURRENTLY», если флаг
postgresql_concurrently
установлен послеIndex
и если используемая база данных определена как PostgreSQL версии 9.2 или выше. Для CREATE INDEX также добавлено определение версии базы данных, которое будет опускать пункт, если версия PG меньше 8.2. Pull request любезно предоставлен Iuri de Silvio.Добавлен новый параметр
PGInspector.get_view_names.include
, позволяющий указать, какие виды представлений должны быть возвращены. В настоящее время включены «простые» и «материализованные» представления. Pull request любезно предоставлен Себастьяном Банком.References: #3588
Добавлено
postgresql_tablespace
в качестве аргумента кIndex
для возможности указания TABLESPACE для индекса в PostgreSQL. Дополняет одноименный параметр вTable
. Pull request любезно предоставлен Бенджамином Бертраном.References: #3720
Добавлен новый параметр
GenerativeSelect.with_for_update.key_share
, который будет выводитьFOR NO KEY UPDATE
версиюFOR UPDATE
иFOR KEY SHARE
вместоFOR SHARE
на бэкенде PostgreSQL. Pull request любезно предоставлен Сергеем Скопиным.Добавлен новый параметр
GenerativeSelect.with_for_update.skip_locked
, который будет отображать фразуSKIP LOCKED
для блокировкиFOR UPDATE
илиFOR SHARE
на бэкендах PostgreSQL и Oracle. Pull request любезно предоставлен Джеком Чжоу.Добавлен новый диалект PyGreSQL для диалекта PostgreSQL. Спасибо Кристофу Цвершке и Kaolin Imago Fire за их усилия.
Добавлена новая константа
JSON.NULL
, указывающая, что для значения JSON NULL должно использоваться значение NULL независимо от других настроек.См.также
References: #3514
Модуль
sqlalchemy.dialects.postgres
, давно устаревший, удаляется; он выдавал предупреждение в течение многих лет, и проекты должны обращаться кsqlalchemy.dialects.postgresql
. Однако URL-адреса движков в формеpostgres://
будут продолжать работать.В PostgreSQL-версию метода
Inspector.get_view_definition()
добавлена поддержка отражения источника материализованных представлений.References: #3587
Использование объекта
ARRAY
, который ссылается на подтипEnum
илиENUM
, теперь будет выдавать ожидаемые «CREATE TYPE» и «DROP TYPE». DDL, когда тип используется внутри «CREATE TABLE» или «DROP TABLE».References: #2729
Флаг «hashable» для специальных типов данных, таких как
ARRAY
,JSON
иHSTORE
, теперь установлен на False, что позволяет этим типам быть извлекаемыми в ORM-запросах, включающих сущности внутри строки.References: #3499
Тип PostgreSQL
ARRAY
теперь поддерживает многомерный индексированный доступ, например, выражения типаsomecol[5][6]
без необходимости явного приведения или принуждения типов, при условии, что параметрARRAY.dimensions
установлен на желаемое количество измерений.References: #3487
Тип возврата для
JSON
иJSONB
при использовании индексированного доступа был исправлен так, чтобы работать как сам PostgreSQL, и возвращал выражение, которое само имеет типJSON
илиJSONB
. Ранее аксессор возвращалNullType
, что не позволяло использовать последующие JSON-подобные операторы.References: #3503
Типы данных
JSON
,JSONB
иHSTORE
теперь позволяют полностью контролировать тип возврата из операции индексированного текстового доступа, либоcolumn[someindex].astext
для типа JSON, либоcolumn[someindex]
для типа HSTORE, через параметрыJSON.astext_type
иHSTORE.text_type
.References: #3503
Модификатор
Comparator.astext
больше не вызываетColumnElement.cast()
неявно, так как типы JSON/JSONB в PG также допускают перекрестное обращение друг к другу. Код, использующийColumnElement.cast()
при индексированном доступе к JSON, напримерcol[someindex].cast(Integer)
, должен быть изменен для явного вызоваComparator.astext
.References: #3503
mysql¶
Добавлена поддержка «autocommit» в драйверах MySQL, через настройку уровня изоляции AUTOCOMMIT. Pull request любезно предоставлен Романом Подолякой.
References: #3332
Добавлен
JSON
для MySQL 5.7. Тип JSON обеспечивает сохранение значений JSON в MySQL, а также базовую поддержку операторов «getitem» и «getpath», используя функциюJSON_EXTRACT
для обращения к отдельным путям в структуре JSON.См.также
References: #3547
Диалект MySQL больше не генерирует дополнительную директиву «KEY» при генерации CREATE TABLE DDL для таблицы, использующей InnoDB с составным первичным ключом с AUTO_INCREMENT на столбце, который не является первым столбцом; чтобы преодолеть ограничение InnoDB здесь, ограничение PRIMARY KEY теперь генерируется со столбцом AUTO_INCREMENT, расположенным первым в списке столбцов.
См.также
No more generation of an implicit KEY for composite primary key w/ AUTO_INCREMENT
The .autoincrement directive is no longer implicitly enabled for a composite primary key column
References: #3216
sqlite¶
Диалект SQLite теперь отражает фразы ON UPDATE и ON DELETE в ограничениях внешнего ключа. Pull request любезно предоставлен Михалом Петрухой.
Диалект SQLite теперь отражает имена ограничений первичного ключа. Pull request любезно предоставлен Дианой Кларк.
References: #3629
Добавлена поддержка диалекта SQLite для работы метода
Inspector.get_schema_names()
с SQLite; запрос на исправление любезно предоставлен Брайаном Ван Клавереном. Также исправлена поддержка создания индексов со схемами, а также отражение ограничений внешнего ключа в таблицах, связанных со схемами.См.также
Обходной путь для право-вложенных объединений на SQLite, где они переписываются как подзапросы, чтобы обойти отсутствие поддержки этого синтаксиса в SQLite, отменяется при обнаружении SQLite версии 3.7.16 или выше.
References: #3634
Обходной путь для неожиданной доставки SQLite имен столбцов в виде
tablename.columnname
для некоторых типов запросов теперь отключен при обнаружении SQLite версии 3.10.0 или выше.References: #3633
mssql¶
Флаг
mssql_clustered
, доступный наUniqueConstraint
,PrimaryKeyConstraint
,Index
, теперь по умолчанию равенNone
, и может быть установлен в False, что приведет к тому, что ключевое слово NONCLUSTERED не будет использоваться для первичного ключа, что позволит использовать другой индекс в качестве «кластерного». Pull request courtesy Saulius Žemaitaitis.Добавлена поддержка базового уровня изоляции для диалектов SQL Server через параметры
create_engine.isolation_level
иConnection.execution_options.isolation_level
.References: #3534
Флаг
legacy_schema_aliasing
, введенный в версии 1.0.5 как часть #3424 для отключения попыток диалекта MSSQL создать псевдонимы для таблиц, определенных схемой, теперь имеет значение по умолчанию False; старое поведение теперь отключено, если не включено явным образом.References: #3434
Корректировка диалекта mxODBC для использования символа
BinaryNull
, когда это уместно, в сочетании с типом данныхVARBINARY
. Pull request любезно предоставлен Шейлой Аллен.Исправлена проблема, когда диалект SQL Server отражал строковый или другой тип столбца переменной длины с неограниченной длиной, присваивая атрибуту длины строки маркер
"max"
. Хотя явное использование маркера"max"
поддерживается диалектом SQL Server, оно не является частью обычного контракта базовых строковых типов, и вместо этого длина должна быть оставлена как None. Теперь диалект присваивает длине значение None при отражении типа, чтобы тип вел себя нормально в других контекстах.References: #3504
misc¶
Добавлены вспомогательные классы
MutableSet
иMutableList
к расширению Отслеживание мутаций. Pull request любезно предоставлен Jeong YunWon.References: #3297
docstring, указанный для гибридного свойства или метода, теперь учитывается на уровне класса, что позволяет работать с такими инструментами, как Sphinx autodoc. Механика здесь обязательно включает некоторую обертку выражений для гибридных свойств, что может привести к их различному отображению при использовании интроспекции.
References: #3653
Неподдерживаемый диалект Sybase теперь выдает ошибку
NotImplementedError
при попытке составить запрос, включающий «offset»; Sybase не имеет прямой функции «offset».References: #2278