SQLAlchemy 1.3 has deprecated engine.contextual_connect() which
will be removed in 1.4. Additionally, private methods like
Engine._run_visitor will also be removed.
Change-Id: I319785d7dd83ffe2c6e651a2494b073becc84684
The "force" parameter in SQLAlchemy IdentifierPreparer.quote()
has been a no-op since 0.9 in
031ef08078,
which was six years ago. In SQLAlchemy 1.3 this parameter
will be removed entirely. Bump requirements to 0.9 series
here and remove usage of the "force" flag.
Change-Id: I4492df2e7d2075fefbf13d6782de11f7d402f6b8
Use "PRAGMA legacy_alter_table = ON" with sqlite >= 3.26 when
using "ALTER TABLE RENAME TO migration_tmp" to maintain legacy
behavior.
As of sqlite version 3.26, when a table is renamed using
"ALTER TABLE RENAME TO", REFERENCES clauses that refer to the
table will be updated. To maintain legacy (3.24 and earlier)
behavior, "PRAGMA legacy_alter_table" can be set to true and
"PRAGMA foreign_keys" can be set to false. [1]
[1] https://www.sqlite.org/src/info/ae9638e9c0ad0c36
Thanks to "László Böszörményi (GCS)" <gcs@debian.org> for
providing the code for this patch, which has since been
slightly modified.
Change-Id: I539988ab2ad6df6c8f423ecec15364ad8fcc7267
Closes-Bug: 1807262
Starting in SQLAlchemy 1.1, the rules for when "autoincrement=True"
may be set on a column are more strict. The migrate tests are
testing the alteration of a column from Integer to String
and then regenerating; this means we need to set autoincrement
to False as well. A related issue in SQLAlchemy 1.1 is
also being fixed (see https://bitbucket.org/zzzeek/sqlalchemy/issues/3835/),
however this fix is not needed in order for the tests to pass here.
Change-Id: Ibd3a75fff13312411df87e17b6e5764865d69728
Lift the requirements to support SQLAlchemy 1.0. Two tests
were calling upon revised APIs and required adjustment.
Change-Id: Ic91a91bb3c915027b522eace302f2ed074233294
The ibmdb2 code calls _index_identifier() when it handles index name. This
method only exists from sqlalchemy 0.6.5 to 0.7.*. Nova code change
https://review.openstack.org/#/c/153123/ attempts to drop a db constraint and
it fails to sync nova db with sqlalchemy 0.9.8 running against db2. Need to let
ibmdb2 code identify sqlalchemy version and call the correct method to handle
index name.
Closes-Bug: 1428477
Change-Id: Ie6333f9cea0209c1ea290356873a1a1bcf409bed
This implements the ability to drop foreign keys
with sqlite. It's basically the same implementation
used for dropping unique constraints so the common
code is refactored.
The existing FKey test that was skipping sqlite is
no longer skipped to show this works.
Change-Id: Idaaf4229e34af4c21c3bcead4b4e22491d24238e
Closes-Bug: #1423955
Commit 8d6ce64cd08c0598963a92844495782997cd59f3 started using the native
quote attribute built into sqlalchemy 0.9 but missed the changes to the
ibmdb2 changeset, so alter table statements fail for DB2 on sqlalchemy
>= 0.9 (tested against 0.9.8).
This fixes the same issue for the ibmdb2 changeset.
Change-Id: Ia3fa6c3090b5eab29ed7746f4795d502990b8a2f
The changeset module was adding a warnings filter on import. This
affects all applications that wind up importing it. A library
shouldn't modify the warnings filters unless asked.
Closes-Bug: #1407736
Change-Id: I893f8be48efd3d3642e977ab587c9e6dc867258b
The ibmdb2 unique constraint code was accessing the private _all_cols
member var in iterating over columns which breaks in sqlalchemy 0.9 so
fix up the code to not use internals of sqlalchemy.
UniqueConstraint in sqlalchemy extends ColumnCollectionConstraint
which implements __iter__ to generate a tuple of the columns in
the constraint, so we just iterate over the constraint as the fix.
This is based on a patch from Rahul Priyadarshi in ibm-db-sa issue
158:
https://code.google.com/p/ibm-db/issues/detail?id=158
Co-Authored-By: Rahul Priyadarshi <rahul.priyadarshi@in.ibm.com>
Change-Id: I0f06f6314c382e83573d762abe5981db0a02a83a
oslo-incubator commit 3f503faac for making sqlite work with dropping
unique constraints in database migrations. This was made in
oslo-incubator since at the time sqlalchemy-migrate was not in
stackforge. Now that we can update sqlalchemy-migrate, move the patch
over from oslo.
This change also adds the support for the case that a unique constraint
is dropped because the column it's on is dropped.
Note that there are already unit tests that cover dropping a unique
constraint directly and implicitly via dropping a column that is in
the unique constraint.
Related-Bug: #1307266
Change-Id: I5ee8082a83aebf66f6e1dacb093ed79e13f73f5e
Brief summary of the modifications:
* Use six for compatibility with both Python 2 and 3;
* Replace UserDict.DictMixin with collections.MutableMapping;
* Fix relative imports;
* Use test-requirements.txt for requirements that are common to both Python 2
and 3, and test-requirements-py{2,3}.txt for version-specific requirements;
* Miscellaneous fixes.
* Use a specific test_db_py3.cfg file for Python 3, that only runs tests on
sqlite.
Thanks to Victor Stinner who co-wrote this patch.
Change-Id: Ia6dc536c39d274924c21fd5bb619e8e5721e04c4
Co-Authored-By: Victor Stinner <victor.stinner@enovance.com>
AbstractType not longer exists in the class hierarchy for types.
TypeEngine was its direct descendant, so use that instead.
Change-Id: Idbfaee4b0d3acbc4795913ddf2ab4e1c9b6d065c
In SQLA 0.9 there is now a native .quote attribute on many objects.
Conditionally use this instead of the old method if the attribute
exists, to remove deprecation messages (and prepare for when the
other way will be fully removed).
Change-Id: I3c5fada13e044c1c4102acc0455226ce1524f2e2
Since ibm_db_sa is not part of sqlalchemy, we need to handle the
conditional import of the module in visitor.py so we don't get an
ImportError if ibm_db_sa is not available.
Closes-Bug: #1287229
Change-Id: Ida070b629ce3b9be727ae49973bb6a71543c1dcf
This patch fixes get_constraint_name in the
ANSIConstraintCommon class. It's part of the
fixes needed for SQLA 0.9.x compat.
Change-Id: I1f1648af48f459bd18f99bb42fa9a272186fb37d
Remove all trailing spaces and tabs in every file in the project.
People have editors configured to do this, which causes them to
accidentally make little whitespace changes in unrelated commits,
which makes those commits harder to review. Better to fix them all
at once.
Change-Id: I17d89f55f41d8599e0ab1a31f646cd161289703e
This patch adds the initial support for DB2 10.5 to migrate. It
includes:
1. The dialect implementation for DB2.
2. The dialect registration with the visitor.
3. Code to parse the engine name in version.py.
4. A new dependency on ibm_db_sa in test-requirements.txt.
5. A connection string in test_db.cfg for ibm_db_sa.
Part of blueprint add-db2-support
Co-authored-by: Sheng Bo Hou <sbhou@cn.ibm.com>
Co-authored-by: Thuy Christenson <thuy@us.ibm.com>
Co-authored-by: Rahul Priyadarshi <rahul.priyadarshi@in.ibm.com>
Change-Id: I745ec615487b1b06c5d1a09ea316f376d66ee4c0
Version 0.8 of SQLAlchemy added support of indexes
on expressions in addition to plain table columns,
which changed the way indexes are created.
This broke support of dropping columns of composite
indexes for SQLite: due to limitations of ALTER in
SQLite every time a column is dropped, we recreate
the whole table without the given column; if a
column is a part of a composite index, we change the
index definition to omit that column and then indexes
are recreated too.
SQLAlchemy versions starting from 0.8 no more pay
attention to 'columns' attribute of Index instances
when generating DDL for indexes, so when one of columns
of a composite index is dropped, we try to create a
new index on the column that doesn't exist anymore,
which of course fails.
Closes-Bug: #1241038
Change-Id: I777b8ce36e36f49bfb0889908811a063cf1a527b
In commit 0.7.2-16-gc670d1d the _index_identifier() implementation
was copied from sqlalchemy, as that function was renamed
in sqlalchemy 0.8. Instead handle call the renamed function
when appropriate, to decouple ourselves from the sqlalchemy
implementation.
Change-Id: I97b22c20d96758fc5b6bd55318218edb26c5b5d0
