Django gives you a few ways to control how database transactions are managed.
\nDjango’s default behavior is to run in autocommit mode. Each query is\nimmediately committed to the database, unless a transaction is active.\nSee below for details.
\nDjango uses transactions or savepoints automatically to guarantee the\nintegrity of ORM operations that require multiple queries, especially\ndelete() and update() queries.
\nDjango’s TestCase class also wraps each test in a\ntransaction for performance reasons.
A common way to handle transactions on the web is to wrap each request in a\ntransaction. Set ATOMIC_REQUESTS to\nTrue in the configuration of each database for which you want to enable\nthis behavior.
It works like this. Before calling a view function, Django starts a\ntransaction. If the response is produced without problems, Django commits the\ntransaction. If the view produces an exception, Django rolls back the\ntransaction.
\nYou may perform subtransactions using savepoints in your view code, typically\nwith the atomic() context manager. However, at the end of the view,\neither all or none of the changes will be committed.
In practice, this feature simply wraps every view function in the atomic()\ndecorator described below.
Note that only the execution of your view is enclosed in the transactions.\nMiddleware runs outside of the transaction, and so does the rendering of\ntemplate responses.
\nWhen ATOMIC_REQUESTS is enabled, it’s\nstill possible to prevent views from running in a transaction.
This decorator will negate the effect of ATOMIC_REQUESTS for a given view:
from django.db import transaction\n\n@transaction.non_atomic_requests\ndef my_view(request):\n do_stuff()\n\n@transaction.non_atomic_requests(using='other')\ndef my_other_view(request):\n do_stuff_on_the_other_database()\nIt only works if it’s applied to the view itself.
\nDjango provides a single API to control database transactions.
\nAtomicity is the defining property of database transactions. atomic\nallows us to create a block of code within which the atomicity on the\ndatabase is guaranteed. If the block of code is successfully completed, the\nchanges are committed to the database. If there is an exception, the\nchanges are rolled back.
atomic blocks can be nested. In this case, when an inner block\ncompletes successfully, its effects can still be rolled back if an\nexception is raised in the outer block at a later point.
atomic is usable both as a decorator:
from django.db import transaction\n\n@transaction.atomic\ndef viewfunc(request):\n # This code executes inside a transaction.\n do_stuff()\nand as a context manager:
\nfrom django.db import transaction\n\ndef viewfunc(request):\n # This code executes in autocommit mode (Django's default).\n do_stuff()\n\n with transaction.atomic():\n # This code executes inside a transaction.\n do_more_stuff()\nWrapping atomic in a try/except block allows for natural handling of\nintegrity errors:
from django.db import IntegrityError, transaction\n\n@transaction.atomic\ndef viewfunc(request):\n create_parent()\n\n try:\n with transaction.atomic():\n generate_relationships()\n except IntegrityError:\n handle_exception()\n\n add_children()\nIn this example, even if generate_relationships() causes a database\nerror by breaking an integrity constraint, you can execute queries in\nadd_children(), and the changes from create_parent() are still\nthere. Note that any operations attempted in generate_relationships()\nwill already have been rolled back safely when handle_exception() is\ncalled, so the exception handler can also operate on the database if\nnecessary.
In order to guarantee atomicity, atomic disables some APIs. Attempting\nto commit, roll back, or change the autocommit state of the database\nconnection within an atomic block will raise an exception.
atomic takes a using argument which should be the name of a\ndatabase. If this argument isn’t provided, Django uses the "default"\ndatabase.
Under the hood, Django’s transaction management code:
\nopens a transaction when entering the outermost atomic block;
creates a savepoint when entering an inner atomic block;
releases or rolls back to the savepoint when exiting an inner block;
commits or rolls back the transaction when exiting the outermost block.
You can disable the creation of savepoints for inner blocks by setting the\nsavepoint argument to False. If an exception occurs, Django will\nperform the rollback when exiting the first parent block with a savepoint\nif there is one, and the outermost block otherwise. Atomicity is still\nguaranteed by the outer transaction. This option should only be used if\nthe overhead of savepoints is noticeable. It has the drawback of breaking\nthe error handling described above.
You may use atomic when autocommit is turned off. It will only use\nsavepoints, even for the outermost block.
In the SQL standards, each SQL query starts a transaction, unless one is\nalready active. Such transactions must then be explicitly committed or rolled\nback.
\nThis isn’t always convenient for application developers. To alleviate this\nproblem, most databases provide an autocommit mode. When autocommit is turned\non and no transaction is active, each SQL query gets wrapped in its own\ntransaction. In other words, not only does each such query start a\ntransaction, but the transaction also gets automatically committed or rolled\nback, depending on whether the query succeeded.
\nPEP 249, the Python Database API Specification v2.0, requires autocommit to\nbe initially turned off. Django overrides this default and turns autocommit\non.
\nTo avoid this, you can deactivate the transaction management, but it isn’t recommended.
\nYou can totally disable Django’s transaction management for a given database\nby setting AUTOCOMMIT to False in its\nconfiguration. If you do this, Django won’t enable autocommit, and won’t\nperform any commits. You’ll get the regular behavior of the underlying\ndatabase library.
This requires you to commit explicitly every transaction, even those started\nby Django or by third-party libraries. Thus, this is best used in situations\nwhere you want to run your own transaction-controlling middleware or do\nsomething really strange.
\nSometimes you need to perform an action related to the current database\ntransaction, but only if the transaction successfully commits. Examples might\ninclude a Celery task, an email notification, or a cache invalidation.
\nDjango provides the on_commit() function to register callback functions\nthat should be executed after a transaction is successfully committed:
Pass any function (that takes no arguments) to on_commit():
from django.db import transaction\n\ndef do_something():\n pass # send a mail, invalidate a cache, fire off a Celery task, etc.\n\ntransaction.on_commit(do_something)\nYou can also wrap your function in a lambda:
\ntransaction.on_commit(lambda: some_celery_task.delay('arg1'))\nThe function you pass in will be called immediately after a hypothetical\ndatabase write made where on_commit() is called would be successfully\ncommitted.
If you call on_commit() while there isn’t an active transaction, the\ncallback will be executed immediately.
If that hypothetical database write is instead rolled back (typically when an\nunhandled exception is raised in an atomic() block), your function will\nbe discarded and never called.
Savepoints (i.e. nested atomic() blocks) are handled correctly. That is,\nan on_commit() callable registered after a savepoint (in a nested\natomic() block) will be called after the outer transaction is committed,\nbut not if a rollback to that savepoint or any previous savepoint occurred\nduring the transaction:
with transaction.atomic(): # Outer atomic, start a new transaction\n transaction.on_commit(foo)\n\n with transaction.atomic(): # Inner atomic block, create a savepoint\n transaction.on_commit(bar)\n\n# foo() and then bar() will be called when leaving the outermost block\nOn the other hand, when a savepoint is rolled back (due to an exception being\nraised), the inner callable will not be called:
\nwith transaction.atomic(): # Outer atomic, start a new transaction\n transaction.on_commit(foo)\n\n try:\n with transaction.atomic(): # Inner atomic block, create a savepoint\n transaction.on_commit(bar)\n raise SomeError() # Raising an exception - abort the savepoint\n except SomeError:\n pass\n\n# foo() will be called, but not bar()\nOn-commit functions for a given transaction are executed in the order they were\nregistered.
\nIf one on-commit function within a given transaction raises an uncaught\nexception, no later registered functions in that same transaction will run.\nThis is, of course, the same behavior as if you’d executed the functions\nsequentially yourself without on_commit().
Your callbacks are executed after a successful commit, so a failure in a\ncallback will not cause the transaction to roll back. They are executed\nconditionally upon the success of the transaction, but they are not part of\nthe transaction. For the intended use cases (mail notifications, Celery tasks,\netc.), this should be fine. If it’s not (if your follow-up action is so\ncritical that its failure should mean the failure of the transaction itself),\nthen you don’t want to use the on_commit() hook. Instead, you may want\ntwo-phase commit such as the psycopg Two-Phase Commit protocol support\nand the optional Two-Phase Commit Extensions in the Python DB-API\nspecification.
Callbacks are not run until autocommit is restored on the connection following\nthe commit (because otherwise any queries done in a callback would open an\nimplicit transaction, preventing the connection from going back into autocommit\nmode).
\nWhen in autocommit mode and outside of an atomic() block, the function\nwill run immediately, not on commit.
On-commit functions only work with autocommit mode\nand the atomic() (or ATOMIC_REQUESTS) transaction API. Calling on_commit() when\nautocommit is disabled and you are not within an atomic block will result in an\nerror.
Django’s TestCase class wraps each test in a transaction\nand rolls back that transaction after each test, in order to provide test\nisolation. This means that no transaction is ever actually committed, thus your\non_commit() callbacks will never be run. If you need to test the results\nof an on_commit() callback, use a\nTransactionTestCase instead.
A rollback hook is harder to implement robustly than a commit hook, since a\nvariety of things can cause an implicit rollback.
\nFor instance, if your database connection is dropped because your process was\nkilled without a chance to shut down gracefully, your rollback hook will never\nrun.
\nThe solution is simple: instead of doing something during the atomic block\n(transaction) and then undoing it if the transaction fails, use\non_commit() to delay doing it in the first place until after the\ntransaction succeeds. It’s a lot easier to undo something you never did in the\nfirst place!
Django provides a straightforward API in the django.db.transaction\nmodule to manage the autocommit state of each database connection.
These functions take a using argument which should be the name of a\ndatabase. If it isn’t provided, Django uses the "default" database.
Autocommit is initially turned on. If you turn it off, it’s your\nresponsibility to restore it.
\nOnce you turn autocommit off, you get the default behavior of your database\nadapter, and Django won’t help you. Although that behavior is specified in\nPEP 249, implementations of adapters aren’t always consistent with one\nanother. Review the documentation of the adapter you’re using carefully.
\nYou must ensure that no transaction is active, usually by issuing a\ncommit() or a rollback(), before turning autocommit back on.
Django will refuse to turn autocommit off when an atomic() block is\nactive, because that would break atomicity.
A transaction is an atomic set of database queries. Even if your program\ncrashes, the database guarantees that either all the changes will be applied,\nor none of them.
\nDjango doesn’t provide an API to start a transaction. The expected way to\nstart a transaction is to disable autocommit with set_autocommit().
Once you’re in a transaction, you can choose either to apply the changes\nyou’ve performed until this point with commit(), or to cancel them with\nrollback(). These functions are defined in django.db.transaction.
These functions take a using argument which should be the name of a\ndatabase. If it isn’t provided, Django uses the "default" database.
Django will refuse to commit or to rollback when an atomic() block is\nactive, because that would break atomicity.
A savepoint is a marker within a transaction that enables you to roll back\npart of a transaction, rather than the full transaction. Savepoints are\navailable with the SQLite, PostgreSQL, Oracle, and MySQL (when using the InnoDB\nstorage engine) backends. Other backends provide the savepoint functions, but\nthey’re empty operations – they don’t actually do anything.
\nSavepoints aren’t especially useful if you are using autocommit, the default\nbehavior of Django. However, once you open a transaction with atomic(),\nyou build up a series of database operations awaiting a commit or rollback. If\nyou issue a rollback, the entire transaction is rolled back. Savepoints\nprovide the ability to perform a fine-grained rollback, rather than the full\nrollback that would be performed by transaction.rollback().
When the atomic() decorator is nested, it creates a savepoint to allow\npartial commit or rollback. You’re strongly encouraged to use atomic()\nrather than the functions described below, but they’re still part of the\npublic API, and there’s no plan to deprecate them.
Each of these functions takes a using argument which should be the name of\na database for which the behavior applies. If no using argument is\nprovided then the "default" database is used.
Savepoints are controlled by three functions in django.db.transaction:
Creates a new savepoint. This marks a point in the transaction that is\nknown to be in a «good» state. Returns the savepoint ID (sid).
Releases savepoint sid. The changes performed since the savepoint was\ncreated become part of the transaction.
Rolls back the transaction to savepoint sid.
These functions do nothing if savepoints aren’t supported or if the database\nis in autocommit mode.
\nIn addition, there’s a utility function:
\nResets the counter used to generate unique savepoint IDs.
\nThe following example demonstrates the use of savepoints:
\nfrom django.db import transaction\n\n# open a transaction\n@transaction.atomic\ndef viewfunc(request):\n\n a.save()\n # transaction now contains a.save()\n\n sid = transaction.savepoint()\n\n b.save()\n # transaction now contains a.save() and b.save()\n\n if want_to_keep_b:\n transaction.savepoint_commit(sid)\n # open transaction still contains a.save() and b.save()\n else:\n transaction.savepoint_rollback(sid)\n # open transaction now contains only a.save()\nSavepoints may be used to recover from a database error by performing a partial\nrollback. If you’re doing this inside an atomic() block, the entire block\nwill still be rolled back, because it doesn’t know you’ve handled the situation\nat a lower level! To prevent this, you can control the rollback behavior with\nthe following functions.
Setting the rollback flag to True forces a rollback when exiting the\ninnermost atomic block. This may be useful to trigger a rollback without\nraising an exception.
Setting it to False prevents such a rollback. Before doing that, make sure\nyou’ve rolled back the transaction to a known-good savepoint within the current\natomic block! Otherwise you’re breaking atomicity and data corruption may\noccur.
While SQLite supports savepoints, a flaw in the design of the sqlite3\nmodule makes them hardly usable.
When autocommit is enabled, savepoints don’t make sense. When it’s disabled,\nsqlite3 commits implicitly before savepoint statements. (In fact, it\ncommits before any statement other than SELECT, INSERT, UPDATE,\nDELETE and REPLACE.) This bug has two consequences:
If you’re using MySQL, your tables may or may not support transactions; it\ndepends on your MySQL version and the table types you’re using. (By\n«table types,» we mean something like «InnoDB» or «MyISAM».) MySQL transaction\npeculiarities are outside the scope of this article, but the MySQL site has\ninformation on MySQL transactions.
\nIf your MySQL setup does not support transactions, then Django will always\nfunction in autocommit mode: statements will be executed and committed as soon\nas they’re called. If your MySQL setup does support transactions, Django\nwill handle transactions as explained in this document.
\nInside a transaction, when a call to a PostgreSQL cursor raises an exception\n(typically IntegrityError), all subsequent SQL in the same transaction\nwill fail with the error «current transaction is aborted, queries ignored\nuntil end of transaction block». While simple use of save() is unlikely\nto raise an exception in PostgreSQL, there are more advanced usage patterns\nwhich might, such as saving objects with unique fields, saving using the\nforce_insert/force_update flag, or invoking custom SQL.
There are several ways to recover from this sort of error.
\nThe first option is to roll back the entire transaction. For example:
\na.save() # Succeeds, but may be undone by transaction rollback\ntry:\n b.save() # Could throw exception\nexcept IntegrityError:\n transaction.rollback()\nc.save() # Succeeds, but a.save() may have been undone\nCalling transaction.rollback() rolls back the entire transaction. Any\nuncommitted database operations will be lost. In this example, the changes\nmade by a.save() would be lost, even though that operation raised no error\nitself.
You can use savepoints to control\nthe extent of a rollback. Before performing a database operation that could\nfail, you can set or update the savepoint; that way, if the operation fails,\nyou can roll back the single offending operation, rather than the entire\ntransaction. For example:
\na.save() # Succeeds, and never undone by savepoint rollback\nsid = transaction.savepoint()\ntry:\n b.save() # Could throw exception\n transaction.savepoint_commit(sid)\nexcept IntegrityError:\n transaction.savepoint_rollback(sid)\nc.save() # Succeeds, and a.save() is never undone\nIn this example, a.save() will not be undone in the case where\nb.save() raises an exception.