django/db/migrations/migration.py
import re
from django.db.migrations.utils import get_migration_name_timestamp
from django.db.transaction import atomic
from .exceptions import IrreversibleError
class Migration:
"""
The base class for all migrations.
Migration files will import this from django.db.migrations.Migration
and subclass it as a class called Migration. It will have one or more
of the following attributes:
- operations: A list of Operation instances, probably from
django.db.migrations.operations
- dependencies: A list of tuples of (app_path, migration_name)
- run_before: A list of tuples of (app_path, migration_name)
- replaces: A list of migration_names
Note that all migrations come out of migrations and into the Loader or
Graph as instances, having been initialized with their app label and name.
"""
# Operations to apply during this migration, in order.
operations = []
# Other migrations that should be run before this migration.
# Should be a list of (app, migration_name).
dependencies = []
# Other migrations that should be run after this one (i.e. have
# this migration added to their dependencies). Useful to make third-party
# apps' migrations run after your AUTH_USER replacement, for example.
run_before = []
# Migration names in this app that this migration replaces. If this is
# non-empty, this migration will only be applied if all these migrations
# are not applied.
replaces = []
# Is this an initial migration? Initial migrations are skipped on
# --fake-initial if the table or fields already exist. If None, check if
# the migration has any dependencies to determine if there are dependencies
# to tell if db introspection needs to be done. If True, always perform
# introspection. If False, never perform introspection.
initial = None
# Whether to wrap the whole migration in a transaction. Only has an effect
# on database backends which support transactional DDL.
atomic = True
def __init__(self, name, app_label):
self.name = name
self.app_label = app_label
# Copy dependencies & other attrs as we might mutate them at runtime
self.operations = list(self.__class__.operations)
self.dependencies = list(self.__class__.dependencies)
self.run_before = list(self.__class__.run_before)
self.replaces = list(self.__class__.replaces)
def __eq__(self, other):
return (
isinstance(other, Migration)
and self.name == other.name
and self.app_label == other.app_label
)
def __repr__(self):
return "<Migration %s.%s>" % (self.app_label, self.name)
def __str__(self):
return "%s.%s" % (self.app_label, self.name)
def __hash__(self):
return hash("%s.%s" % (self.app_label, self.name))
def mutate_state(self, project_state, preserve=True):
"""
Take a ProjectState and return a new one with the migration's
operations applied to it. Preserve the original object state by
default and return a mutated state from a copy.
"""
new_state = project_state
if preserve:
new_state = project_state.clone()
for operation in self.operations:
operation.state_forwards(self.app_label, new_state)
return new_state
def apply(self, project_state, schema_editor, collect_sql=False):
"""
Take a project_state representing all migrations prior to this one
and a schema_editor for a live database and apply the migration
in a forwards order.
Return the resulting project state for efficient reuse by following
Migrations.
"""
for operation in self.operations:
# If this operation cannot be represented as SQL, place a comment
# there instead
if collect_sql:
schema_editor.collected_sql.append("--")
schema_editor.collected_sql.append("-- %s" % operation.describe())
schema_editor.collected_sql.append("--")
if not operation.reduces_to_sql:
schema_editor.collected_sql.append(
"-- THIS OPERATION CANNOT BE WRITTEN AS SQL"
)
continue
collected_sql_before = len(schema_editor.collected_sql)
# Save the state before the operation has run
old_state = project_state.clone()
operation.state_forwards(self.app_label, project_state)
# Run the operation
atomic_operation = operation.atomic or (
self.atomic and operation.atomic is not False
)
if not schema_editor.atomic_migration and atomic_operation:
# Force a transaction on a non-transactional-DDL backend or an
# atomic operation inside a non-atomic migration.
with atomic(schema_editor.connection.alias):
operation.database_forwards(
self.app_label, schema_editor, old_state, project_state
)
else:
# Normal behaviour
operation.database_forwards(
self.app_label, schema_editor, old_state, project_state
)
if collect_sql and collected_sql_before == len(schema_editor.collected_sql):
schema_editor.collected_sql.append("-- (no-op)")
return project_state
def unapply(self, project_state, schema_editor, collect_sql=False):
"""
Take a project_state representing all migrations prior to this one
and a schema_editor for a live database and apply the migration
in a reverse order.
The backwards migration process consists of two phases:
1. The intermediate states from right before the first until right
after the last operation inside this migration are preserved.
2. The operations are applied in reverse order using the states
recorded in step 1.
"""
# Construct all the intermediate states we need for a reverse migration
to_run = []
new_state = project_state
# Phase 1
for operation in self.operations:
# If it's irreversible, error out
if not operation.reversible:
raise IrreversibleError(
"Operation %s in %s is not reversible" % (operation, self)
)
# Preserve new state from previous run to not tamper the same state
# over all operations
new_state = new_state.clone()
old_state = new_state.clone()
operation.state_forwards(self.app_label, new_state)
to_run.insert(0, (operation, old_state, new_state))
# Phase 2
for operation, to_state, from_state in to_run:
if collect_sql:
schema_editor.collected_sql.append("--")
schema_editor.collected_sql.append("-- %s" % operation.describe())
schema_editor.collected_sql.append("--")
if not operation.reduces_to_sql:
schema_editor.collected_sql.append(
"-- THIS OPERATION CANNOT BE WRITTEN AS SQL"
)
continue
collected_sql_before = len(schema_editor.collected_sql)
atomic_operation = operation.atomic or (
self.atomic and operation.atomic is not False
)
if not schema_editor.atomic_migration and atomic_operation:
# Force a transaction on a non-transactional-DDL backend or an
# atomic operation inside a non-atomic migration.
with atomic(schema_editor.connection.alias):
operation.database_backwards(
self.app_label, schema_editor, from_state, to_state
)
else:
# Normal behaviour
operation.database_backwards(
self.app_label, schema_editor, from_state, to_state
)
if collect_sql and collected_sql_before == len(schema_editor.collected_sql):
schema_editor.collected_sql.append("-- (no-op)")
return project_state
def suggest_name(self):
"""
Suggest a name for the operations this migration might represent. Names
are not guaranteed to be unique, but put some effort into the fallback
name to avoid VCS conflicts if possible.
"""
if self.initial:
return "initial"
raw_fragments = [op.migration_name_fragment for op in self.operations]
fragments = [re.sub(r"\W+", "_", name) for name in raw_fragments if name]
if not fragments or len(fragments) != len(self.operations):
return "auto_%s" % get_migration_name_timestamp()
name = fragments[0]
for fragment in fragments[1:]:
new_name = f"{name}_{fragment}"
if len(new_name) > 52:
name = f"{name}_and_more"
break
name = new_name
return name
class SwappableTuple(tuple):
"""
Subclass of tuple so Django can tell this was originally a swappable
dependency when it reads the migration file.
"""
def __new__(cls, value, setting):
self = tuple.__new__(cls, value)
self.setting = setting
return self
def swappable_dependency(value):
"""Turn a setting value into a dependency."""
return SwappableTuple((value.split(".", 1)[0], "__first__"), value)