NDB Query Class

A Query object represents an NDB query, a request for a filtered, sorted list of entities.

This page contains reference documentation. For a general discussion of NDB queries, see Queries.

Query Options

Many query methods take a standard set of additional options, either in the form of keyword arguments such as keys_only=True, or as QueryOptions object passed with options=QueryOptions(...).

Queries support a variety of configuration options. These are specified by keyword arguments to the Query methods:

Argument Type Default Description
keys_only bool False All operations return keys instead of entities.
projection tuple (or list) of properties (or strings) None Operations return entities with only the specified properties set. projection=[Article.title, Article.date] or projection=['title', 'date'] fetches entities with just those two fields set. (See Projection Queries.)
offset int 0 Number of query results to skip.
limit int No limit Maximum number of query results to return.
batch_size int 20 Batch size.

Affects efficiency of queries only; larger batch sizes use more memory but make fewer RPC calls.
prefetch_size int None Overrides batch size for first batch returned.
produce_cursors bool False Generate cursors from query (see Query Iterators. Query Cursors).
start_cursor Cursor None Starting point for search (see Query Cursors).
end_cursor Cursor None Ending point for search (see Query Cursors).
deadline int Depends on Context Overrides RPC deadline (which defaults to 5 seconds if not overridden when Context created)
read_policy ndb.EVENTUAL_CONSISTENCY The read policy. Set to ndb.EVENTUAL_CONSISTENCY to get perhaps-quicker results without waiting for the Datastore to apply pending changes to all returned records.

To run a query with a specific set of options, pass the keyword arguments to the applicable method:

qry = Employee.query().filter(...).order(...) # Create a query
for acct in qry.fetch(10, offset=20): # Skip the first 20
  print acct

Occasionally, you might want to keep a set of query options around and use them in various places. While you could just keep them in a dictionary and pass this dictionary to the methods using **kwds, you can also create a QueryOptions object and pass it using the options keyword argument. The following two examples are equivalent:

qo = ndb.QueryOptions(keys_only=True, offset=20)
results = qry.fetch(10, options=qo)
results = qry.fetch(10, keys_only=True, offset=20)

Constructor

Typically, an application creates a Query by calling Model.query(). But it's also possible to call ndb.Query().

Arguments

kind
Optional kind string. Normally, the name of a entity class.
ancestor
Optional ancestor Key.
filters
Optional Node representing a filter expression tree.
orders
Optional datastore_query.Order object.
app
Optional app id.
namespace
Optional namespace. If not specified, the default namespace at the time the query is executed will be used.
projection
Optional list or tuple of properties to project.
group_by
Optional list or tuple of properties to group by.
default_options
Optional QueryOptions object giving default query options to be used when the query is executed.

Instance Methods

filter(filter1, filter2, ...)
Returns a new Query with additional filter(s) applied. Takes filter arguments as described in Queries. qry.filter(filter1).filter(filter2) is equivalent to qry.filter(filter1filter2)
get(**q_options)
Returns the first query result, if any (otherwise None). This is similar to calling q.fetch(1) and returning the first item of the list of results.

Arguments

**q_options
All query options keyword arguments are supported.
order(order1, order2, ...)
Returns a new Query with additional sort order(s) applied. Takes one or more arguments which are properties or "negated" properties. For example, to sort users by age and "break ties" by name, you might use something like qry.order(-Account.birthday, Account.name)
bind(...values...)
This function is for use with GQL queries that use parameter bindings (:1, :2, etc.) or named bindings (:foo, :bar, etc.). It binds the passed values to the parameters.

To bind parameters, you might call something like qry.bind("USA", 49). To bind named parameters, you might call something like qry.bind(region = "USA", threshold = 49).

Returns a new Query object with the parameter values bound.

count(limit=None, **q_options)
Returns the number of query results, up to a limit. This returns the same result as len(q.fetch(limit)) but more efficiently.

Arguments

limit
How many results to count at most
**q_options
All query options keyword arguments and context options are supported.
count_async(limit, **q_options)
Asynchronously counts the number of query results, up to a limit; it returns a Future whose result is a number. This is the asynchronous version of count().
fetch(limit, **q_options)
Fetch a list of query results, up to a limit.

Arguments

limit
How many results to count at most
**q_options
All query options keyword arguments are supported.
fetch_async(limit, **q_options)
Asynchronously fetch a list of query results, up to a limit. Returns a Future whose result is a list of results. This is the asynchronous version of fetch().
fetch_page(page_size, **q_options)
Fetch a "page" of results. This is a specialized method for use by paging user interfaces.

Arguments

page_size
At most this many results will be returned.
**q_options
All query options keyword arguments are supported.
Returns a tuple (results, cursor, more):
  • results list of query results
  • cursor a query cursor pointing to the "next" batch of results. If there are no more results, this might be None.
  • more bool indicating whether there are (likely) more results after this batch. If False, there are no more results; if True, there are probably more results.

To fetch the next page, pass the cursor returned by one call to the next call using start_cursor=cursor. A common idiom is to pass the cursor to the client using cursor.urlsafe() and to reconstruct that cursor on a subsequent request using Cursor(urlsafe=string).

fetch_page_async(page_size, **q_options)
Asynchronously fetch a "page" of results. This is the asynchronous version of fetch_page().
get_async(**q_options)
Asynchronously returns the first query result, if any (otherwise None). This is the asynchronous version of get().
iter(**q_options)
Constructs and returns an iterator over the query.

Arguments

**q_options
All query options keyword arguments are supported.

Returns a QueryIterator object.

map(callback, pass_batch_into_callback=None, merge_future=None, **q_options)
Map a callback function or tasklet over the query results. That is, apply the function (or tasklet) to each entity in the query results.

Arguments

callback
A function or tasklet to be applied to each result.
pass_batch_into_callback
If True, calls the callback with batch information arguments as described below.
merge_future
Optional Future subclass; see below.
**q_options
All query options keyword arguments are supported.

Callback signature The callback is normally called with an entity as argument. However, if keys_only=True is given, it is called with a Key. If pass_batch_into_callback=True is given, the callback is called with three arguments: the current batch, the index within the batch, and the entity or Key at that index. The callback can return whatever it wants. If the callback is None, a trivial callback is assumed that just returns the entity or key passed in.

Optional merge_future The merge_future is an advanced argument that can be used to override how the callback results are combined into the overall map() return value. By default, a list of callback return values is produced. By substituting one of a small number of specialized alternatives you can arrange otherwise. See the source code for tasklets.MultiFuture for the default implementation and a description of the protocol the merge_future object must implement. Alternatives from the same module include QueueFuture, SerialQueueFuture and ReducingFuture.

Returns a list of the results of all the callbacks. (But see 'optional merge_future' above.) It returns when the query has run to completion and all callbacks have returned.

map_async(callback, pass_batch_into_callback=None, merge_future=None, **q_options)
Asynchronously map a callback function or tasklet over the query results. This is the asynchronous version of map().