Viewing file:  base.py (16.11 KB) -rw-r--r--
Select action/file-type:
 (+) |  (+) |  (+) | Code (+) | Session (+) |  (+) | SDB (+) |  (+) |  (+) |  (+) |  (+) |  (+) |
from contextlib import contextmanager
from .. import util
from ..util import sqla_compat
from . import batch
from . import schemaobj
from ..util.compat import exec_
import textwrap
import inspect
__all__ = ('Operations', 'BatchOperations')
try:
from sqlalchemy.sql.naming import conv
except:
conv = None
class Operations(util.ModuleClsProxy):
"""Define high level migration operations.
Each operation corresponds to some schema migration operation,
executed against a particular :class:`.MigrationContext`
which in turn represents connectivity to a database,
or a file output stream.
While :class:`.Operations` is normally configured as
part of the :meth:`.EnvironmentContext.run_migrations`
method called from an ``env.py`` script, a standalone
:class:`.Operations` instance can be
made for use cases external to regular Alembic
migrations by passing in a :class:`.MigrationContext`::
from alembic.migration import MigrationContext
from alembic.operations import Operations
conn = myengine.connect()
ctx = MigrationContext.configure(conn)
op = Operations(ctx)
op.alter_column("t", "c", nullable=True)
Note that as of 0.8, most of the methods on this class are produced
dynamically using the :meth:`.Operations.register_operation`
method.
"""
_to_impl = util.Dispatcher()
def __init__(self, migration_context, impl=None):
"""Construct a new :class:`.Operations`
:param migration_context: a :class:`.MigrationContext`
instance.
"""
self.migration_context = migration_context
if impl is None:
self.impl = migration_context.impl
else:
self.impl = impl
self.schema_obj = schemaobj.SchemaObjects(migration_context)
@classmethod
def register_operation(cls, name, sourcename=None):
"""Register a new operation for this class.
This method is normally used to add new operations
to the :class:`.Operations` class, and possibly the
:class:`.BatchOperations` class as well. All Alembic migration
operations are implemented via this system, however the system
is also available as a public API to facilitate adding custom
operations.
.. versionadded:: 0.8.0
.. seealso::
:ref:`operation_plugins`
"""
def register(op_cls):
if sourcename is None:
fn = getattr(op_cls, name)
source_name = fn.__name__
else:
fn = getattr(op_cls, sourcename)
source_name = fn.__name__
spec = inspect.getargspec(fn)
name_args = spec[0]
assert name_args[0:2] == ['cls', 'operations']
name_args[0:2] = ['self']
args = inspect.formatargspec(*spec)
num_defaults = len(spec[3]) if spec[3] else 0
if num_defaults:
defaulted_vals = name_args[0 - num_defaults:]
else:
defaulted_vals = ()
apply_kw = inspect.formatargspec(
name_args, spec[1], spec[2],
defaulted_vals,
formatvalue=lambda x: '=' + x)
func_text = textwrap.dedent("""\
def %(name)s%(args)s:
%(doc)r
return op_cls.%(source_name)s%(apply_kw)s
""" % {
'name': name,
'source_name': source_name,
'args': args,
'apply_kw': apply_kw,
'doc': fn.__doc__,
'meth': fn.__name__
})
globals_ = {'op_cls': op_cls}
lcl = {}
exec_(func_text, globals_, lcl)
setattr(cls, name, lcl[name])
fn.__func__.__doc__ = "This method is proxied on "\
"the :class:`.%s` class, via the :meth:`.%s.%s` method." % (
cls.__name__, cls.__name__, name
)
if hasattr(fn, '_legacy_translations'):
lcl[name]._legacy_translations = fn._legacy_translations
return op_cls
return register
@classmethod
def implementation_for(cls, op_cls):
"""Register an implementation for a given :class:`.MigrateOperation`.
This is part of the operation extensibility API.
.. seealso::
:ref:`operation_plugins` - example of use
"""
def decorate(fn):
cls._to_impl.dispatch_for(op_cls)(fn)
return fn
return decorate
@classmethod
@contextmanager
def context(cls, migration_context):
op = Operations(migration_context)
op._install_proxy()
yield op
op._remove_proxy()
@contextmanager
def batch_alter_table(
self, table_name, schema=None, recreate="auto", copy_from=None,
table_args=(), table_kwargs=util.immutabledict(),
reflect_args=(), reflect_kwargs=util.immutabledict(),
naming_convention=None):
"""Invoke a series of per-table migrations in batch.
Batch mode allows a series of operations specific to a table
to be syntactically grouped together, and allows for alternate
modes of table migration, in particular the "recreate" style of
migration required by SQLite.
"recreate" style is as follows:
1. A new table is created with the new specification, based on the
migration directives within the batch, using a temporary name.
2. the data copied from the existing table to the new table.
3. the existing table is dropped.
4. the new table is renamed to the existing table name.
The directive by default will only use "recreate" style on the
SQLite backend, and only if directives are present which require
this form, e.g. anything other than ``add_column()``. The batch
operation on other backends will proceed using standard ALTER TABLE
operations.
The method is used as a context manager, which returns an instance
of :class:`.BatchOperations`; this object is the same as
:class:`.Operations` except that table names and schema names
are omitted. E.g.::
with op.batch_alter_table("some_table") as batch_op:
batch_op.add_column(Column('foo', Integer))
batch_op.drop_column('bar')
The operations within the context manager are invoked at once
when the context is ended. When run against SQLite, if the
migrations include operations not supported by SQLite's ALTER TABLE,
the entire table will be copied to a new one with the new
specification, moving all data across as well.
The copy operation by default uses reflection to retrieve the current
structure of the table, and therefore :meth:`.batch_alter_table`
in this mode requires that the migration is run in "online" mode.
The ``copy_from`` parameter may be passed which refers to an existing
:class:`.Table` object, which will bypass this reflection step.
.. note:: The table copy operation will currently not copy
CHECK constraints, and may not copy UNIQUE constraints that are
unnamed, as is possible on SQLite. See the section
:ref:`sqlite_batch_constraints` for workarounds.
:param table_name: name of table
:param schema: optional schema name.
:param recreate: under what circumstances the table should be
recreated. At its default of ``"auto"``, the SQLite dialect will
recreate the table if any operations other than ``add_column()``,
``create_index()``, or ``drop_index()`` are
present. Other options include ``"always"`` and ``"never"``.
:param copy_from: optional :class:`~sqlalchemy.schema.Table` object
that will act as the structure of the table being copied. If omitted,
table reflection is used to retrieve the structure of the table.
.. versionadded:: 0.7.6 Fully implemented the
:paramref:`~.Operations.batch_alter_table.copy_from`
parameter.
.. seealso::
:ref:`batch_offline_mode`
:paramref:`~.Operations.batch_alter_table.reflect_args`
:paramref:`~.Operations.batch_alter_table.reflect_kwargs`
:param reflect_args: a sequence of additional positional arguments that
will be applied to the table structure being reflected / copied;
this may be used to pass column and constraint overrides to the
table that will be reflected, in lieu of passing the whole
:class:`~sqlalchemy.schema.Table` using
:paramref:`~.Operations.batch_alter_table.copy_from`.
.. versionadded:: 0.7.1
:param reflect_kwargs: a dictionary of additional keyword arguments
that will be applied to the table structure being copied; this may be
used to pass additional table and reflection options to the table that
will be reflected, in lieu of passing the whole
:class:`~sqlalchemy.schema.Table` using
:paramref:`~.Operations.batch_alter_table.copy_from`.
.. versionadded:: 0.7.1
:param table_args: a sequence of additional positional arguments that
will be applied to the new :class:`~sqlalchemy.schema.Table` when
created, in addition to those copied from the source table.
This may be used to provide additional constraints such as CHECK
constraints that may not be reflected.
:param table_kwargs: a dictionary of additional keyword arguments
that will be applied to the new :class:`~sqlalchemy.schema.Table`
when created, in addition to those copied from the source table.
This may be used to provide for additional table options that may
not be reflected.
.. versionadded:: 0.7.0
:param naming_convention: a naming convention dictionary of the form
described at :ref:`autogen_naming_conventions` which will be applied
to the :class:`~sqlalchemy.schema.MetaData` during the reflection
process. This is typically required if one wants to drop SQLite
constraints, as these constraints will not have names when
reflected on this backend. Requires SQLAlchemy **0.9.4** or greater.
.. seealso::
:ref:`dropping_sqlite_foreign_keys`
.. versionadded:: 0.7.1
.. note:: batch mode requires SQLAlchemy 0.8 or above.
.. seealso::
:ref:`batch_migrations`
"""
impl = batch.BatchOperationsImpl(
self, table_name, schema, recreate,
copy_from, table_args, table_kwargs, reflect_args,
reflect_kwargs, naming_convention)
batch_op = BatchOperations(self.migration_context, impl=impl)
yield batch_op
impl.flush()
def get_context(self):
"""Return the :class:`.MigrationContext` object that's
currently in use.
"""
return self.migration_context
def invoke(self, operation):
"""Given a :class:`.MigrateOperation`, invoke it in terms of
this :class:`.Operations` instance.
.. versionadded:: 0.8.0
"""
fn = self._to_impl.dispatch(
operation, self.migration_context.impl.__dialect__)
return fn(self, operation)
def f(self, name):
"""Indicate a string name that has already had a naming convention
applied to it.
This feature combines with the SQLAlchemy ``naming_convention`` feature
to disambiguate constraint names that have already had naming
conventions applied to them, versus those that have not. This is
necessary in the case that the ``"%(constraint_name)s"`` token
is used within a naming convention, so that it can be identified
that this particular name should remain fixed.
If the :meth:`.Operations.f` is used on a constraint, the naming
convention will not take effect::
op.add_column('t', 'x', Boolean(name=op.f('ck_bool_t_x')))
Above, the CHECK constraint generated will have the name
``ck_bool_t_x`` regardless of whether or not a naming convention is
in use.
Alternatively, if a naming convention is in use, and 'f' is not used,
names will be converted along conventions. If the ``target_metadata``
contains the naming convention
``{"ck": "ck_bool_%(table_name)s_%(constraint_name)s"}``, then the
output of the following:
op.add_column('t', 'x', Boolean(name='x'))
will be::
CONSTRAINT ck_bool_t_x CHECK (x in (1, 0)))
The function is rendered in the output of autogenerate when
a particular constraint name is already converted, for SQLAlchemy
version **0.9.4 and greater only**. Even though ``naming_convention``
was introduced in 0.9.2, the string disambiguation service is new
as of 0.9.4.
.. versionadded:: 0.6.4
"""
if conv:
return conv(name)
else:
raise NotImplementedError(
"op.f() feature requires SQLAlchemy 0.9.4 or greater.")
def inline_literal(self, value, type_=None):
"""Produce an 'inline literal' expression, suitable for
using in an INSERT, UPDATE, or DELETE statement.
When using Alembic in "offline" mode, CRUD operations
aren't compatible with SQLAlchemy's default behavior surrounding
literal values,
which is that they are converted into bound values and passed
separately into the ``execute()`` method of the DBAPI cursor.
An offline SQL
script needs to have these rendered inline. While it should
always be noted that inline literal values are an **enormous**
security hole in an application that handles untrusted input,
a schema migration is not run in this context, so
literals are safe to render inline, with the caveat that
advanced types like dates may not be supported directly
by SQLAlchemy.
See :meth:`.execute` for an example usage of
:meth:`.inline_literal`.
The environment can also be configured to attempt to render
"literal" values inline automatically, for those simple types
that are supported by the dialect; see
:paramref:`.EnvironmentContext.configure.literal_binds` for this
more recently added feature.
:param value: The value to render. Strings, integers, and simple
numerics should be supported. Other types like boolean,
dates, etc. may or may not be supported yet by various
backends.
:param ``type_``: optional - a :class:`sqlalchemy.types.TypeEngine`
subclass stating the type of this value. In SQLAlchemy
expressions, this is usually derived automatically
from the Python type of the value itself, as well as
based on the context in which the value is used.
.. seealso::
:paramref:`.EnvironmentContext.configure.literal_binds`
"""
return sqla_compat._literal_bindparam(None, value, type_=type_)
def get_bind(self):
"""Return the current 'bind'.
Under normal circumstances, this is the
:class:`~sqlalchemy.engine.Connection` currently being used
to emit SQL to the database.
In a SQL script context, this value is ``None``. [TODO: verify this]
"""
return self.migration_context.impl.bind
class BatchOperations(Operations):
"""Modifies the interface :class:`.Operations` for batch mode.
This basically omits the ``table_name`` and ``schema`` parameters
from associated methods, as these are a given when running under batch
mode.
.. seealso::
:meth:`.Operations.batch_alter_table`
Note that as of 0.8, most of the methods on this class are produced
dynamically using the :meth:`.Operations.register_operation`
method.
"""
def _noop(self, operation):
raise NotImplementedError(
"The %s method does not apply to a batch table alter operation."
% operation)
|