NDB Functions

Functions

This page describes how to use the legacy bundled services and APIs. This API can only run in first-generation runtimes in the App Engine standard environment. If you are updating to the App Engine Python 3 runtime, refer to the migration guide to learn about your migration options for legacy bundled services.

ndb.add_flow_exception(exc)

Specify that an exception should not be logged, but is just part of normal program flow. (Normally, raising an exception writes a warning message to the application's logs.)

Arguments

exc

Exception class that should not be logged.

By default, the following exceptions are not logged:

• webob.exc.HTTPException (and its subclasses)
• ndb.Rollback

ndb.delete_multi(keys, **ctx_options)

Deletes entities identified by the passed sequence of keys.

Arguments

keys

Sequence of keys

**ctx_options

Context options

ndb.delete_multi_async(keys, **ctx_options)

Asynchronously deletes entities identified by the passed sequence of keys.

Arguments

keys

Sequence of keys

**ctx_options

Context options

Returns a list of Future objects. Each future's result will be None.

ndb.get_multi(keys, **ctx_options)

Fetches entities identified by the passed sequence of keys.

Arguments

keys

Sequence of keys

**ctx_options

Context options

Returns a list. Each list item is either a Model instance or None if the key wasn't found.

ndb.get_multi_async(keys, **ctx_options)

Asynchronously fetches entities identified by the passed sequence of keys.

Arguments

keys

Sequence of keys

**ctx_options

Context options

Returns a list of Future objects. Each future's result is a Model instance or None if the key wasn't found.

ndb.in_transaction()

Returns a Boolean indicating whether a transaction is currently active.

@ndb.non_transactional
@ndb.non_transactional(allow_existing=True)

Decorator to ensure that a function runs outside a transaction.

Arguments:

allow_existing

If True (the default) and if the decorated function is called by code in a transaction, the function runs independent of the transaction. If False and if the decorated function is called by code in a transaction, it raises an exception.

ndb.put_multi(entities, **ctx_options)

Stores a sequence of Model instances.

Arguments

entities

Sequence of Model instances

**ctx_options

Context options

Returns a list with the stored keys.

ndb.put_multi_async(entities, **ctx_options)

Asynchronously stores a sequence of Model instances.

Arguments

entities

Sequence of Model instances

**ctx_options

Context options

Returns a list of Future objects. Each future's result will be a stored key.

ndb.transaction(callback, **ctx_options)

Run a callback in a transaction.

Arguments

callback

Function or tasklet to be called

**ctx_options

Transaction options

Returns whatever callback returns. Raises whatever callback raises or a TransactionFailedError exception if the transaction fails.

To pass arguments to a callback function, use a lambda. For example,

def my_callback(key, inc):
  ...

transaction(lambda: my_callback(Key(...), 1))

ndb.transaction_async(callback, **ctx_options)

Asynchronously run a callback in a transaction.

Arguments

callback

Function or tasklet to be called

**ctx_options

Transaction options

Returns a Future. The future returns whatever callback returns, or raises whatever callback raises or a TransactionFailedError if the transaction fails.

To pass arguments to a callback function, use a lambda. For example,

def my_callback(key, inc):
  ...

transaction(lambda: my_callback(Key(...), 1))

@ndb.transactional
@ndb.transactional(**ctx_options)

Decorator to make a function automatically run in a transaction.

Arguments:

This decorator can have transaction options.

Context Options, Transaction Options

Context options allow you to run specific datastore operations with different configurations. For example, you might want to vary the read policy or the RPC deadline for individual requests. You can do this by passing context options to almost any operation. Some transaction-related functions take transaction options, which include additional options on top of a set of context options.

Here are a few examples using context options. To set the RPC deadline to 1 second when reading an entity, you can use:

key.get(deadline=1)

To set the memcache timeout to 30 seconds when writing an entity, you can use:

ent.put(ndb_memcache_timeout=30)

To delete an item that has been cached and force its reload, you can use:

key.delete(use_datastore=False)

The special keyword arguments options and config (which have identical meanings for historical reasons) allow one to specify several options as a Configuration object. This can be an ndb.ContextOptions object or (for the transactional functions and decorator) an ndb.TransactionOptions object. For example, key.get(options=ndb.ContextOptions(use_cache=True)) is equivalent to key.get(use_cache=True). The options set in such an options object can be overridden by keyword parameters.

The following context options are available:

Option	Type	Description
deadline	float	Datastore call deadline, specified as a number of seconds. (By default, the call is only interrupted by the request handler deadline.)
read_policy	ndb.EVENTUAL_CONSISTENCY	Set this to ndb.EVENTUAL_CONSISTENCY if, instead of waiting for the Datastore to finish applying changes to all returned results, you wish to get possibly-not-current results faster.
force_writes	bool	Specifies whether a write request should succeed even if the app is read-only. (This only applies to user controlled read-only periods.)
use_cache	bool	Specifies whether to store entities in in-process cache; overrides in-process cache policy for this operation.
use_memcache	bool	Specifies whether to store entities in memcache; overrides memcache policy for this operation.
use_datastore	bool	Specifies whether to store entities in Datastore; overrides Datastore policy for this operation.
memcache_timeout	int	Maximum lifetime for entities in memcache; overrides memcache timeout policy for this operation.
max_memcache_items	int	Maximum batch size for the auto-batching feature of the Context memcache methods. For example, with the default size of max_memcache_items (100), up to 100 memcache set operations will be combined into a single set_multi operation.
For some transaction-related functions, the following transaction options are available (along with the inherited context options listed above):
Option	Type	Description
xg	bool	Allow cross-group (XG) transactions. False by default.
propagation	int	NDB provides limited support for transactions within transactions, which are known as "nested transactions". The propagation parameter controls what happens if your code tries to start a nested transaction. The propagation policy for @ndb.transactional defaults to ALLOWED. The propagation policy for ndb.transaction() defaults to NESTED. The NESTED policy is not supported by NDB so your code will throw a BadRequestError exception. NDB sets an unsupported default value, in this case, so that programmers are explicitly aware of the limitations of nested transactions. The propagation parameter can be one of the following values: ndb.TransactionOptions.NESTED The NESTED propagation policy would commit all changes in the outer and inner transactions together when the outer policy commits. However, if an exception is thrown in the inner transaction all changes there would get thrown out but allow the outer transaction to optionally recover and continue. The NESTED policy is not supported. If you use this policy, your code will throw a BadRequestError exception. ndb.TransactionOptions.MANDATORY Always propagate an existing transaction; throw an exception if there is no existing transaction. If a function that uses this policy throws an exception, it's probably not safe to catch the exception and commit the outer transaction; the function may have left the outer transaction in a bad state. ndb.TransactionOptions.ALLOWED If there is an existing transaction, propagate it. If a function that uses this policy throws an exception, it's probably not safe to catch the exception and commit the outer transaction; the function may have left the outer transaction in a bad state. ndb.TransactionOptions.INDEPENDENT Always use a new transaction, "pausing" any existing transactions. A function that uses this policy should not return any entities read in the new transaction, as the entities are not transactionally consistent with the caller's transaction.
retries	int	How many times to retry automatically in case of transaction failures. Zero means to try once but not re-try.

In some cases, options are ignored because of caching. For example, if you specify an RPC deadline for a read operation that is satisfied from the in-context cache, the deadline is ignored. On the other hand, unrecognized options cause TypeError to be raised.

Operations with different options are grouped together when auto-batching applies. For example, if you use put_async() to write some entities with deadline = 5 and some without specifying a deadline, and all are eligible for auto-batching, the auto-batcher will make two separate RPC calls—one for the group of entities with deadline = 5 and one for the other group—even though the default RPC deadline is also 5! This applies even if the option specified is irrelevant to the RPC operation (for example, ndb_should_cache).