cassandra.query - Prepared Statements, Batch Statements, Tracing, and Row Factories

cassandra.query.tuple_factory(colnames, rows)[source]

Returns each row as a tuple


>>> from cassandra.query import tuple_factory
>>> session = cluster.connect('mykeyspace')
>>> session.row_factory = tuple_factory
>>> rows = session.execute("SELECT name, age FROM users LIMIT 1")
>>> print rows[0]
('Bob', 42)

Changed in version 2.0.0: moved from cassandra.decoder to cassandra.query

cassandra.query.named_tuple_factory(colnames, rows)[source]

Returns each row as a namedtuple. This is the default row factory.


>>> from cassandra.query import named_tuple_factory
>>> session = cluster.connect('mykeyspace')
>>> session.row_factory = named_tuple_factory
>>> rows = session.execute("SELECT name, age FROM users LIMIT 1")
>>> user = rows[0]

>>> # you can access field by their name:
>>> print "name: %s, age: %d" % (, user.age)
name: Bob, age: 42

>>> # or you can access fields by their position (like a tuple)
>>> name, age = user
>>> print "name: %s, age: %d" % (name, age)
name: Bob, age: 42
>>> name = user[0]
>>> age = user[1]
>>> print "name: %s, age: %d" % (name, age)
name: Bob, age: 42

Changed in version 2.0.0: moved from cassandra.decoder to cassandra.query

cassandra.query.dict_factory(colnames, rows)[source]

Returns each row as a dict.


>>> from cassandra.query import dict_factory
>>> session = cluster.connect('mykeyspace')
>>> session.row_factory = dict_factory
>>> rows = session.execute("SELECT name, age FROM users LIMIT 1")
>>> print rows[0]
{u'age': 42, u'name': u'Bob'}

Changed in version 2.0.0: moved from cassandra.decoder to cassandra.query

cassandra.query.ordered_dict_factory(colnames, rows)[source]

Like dict_factory(), but returns each row as an OrderedDict, so the order of the columns is preserved.

Changed in version 2.0.0: moved from cassandra.decoder to cassandra.query

class cassandra.query.SimpleStatement(query_string, retry_policy=None, consistency_level=None, routing_key=None, serial_consistency_level=None, fetch_size=<object object>, keyspace=None, custom_payload=None, is_idempotent=False)[source]

A simple, un-prepared query.

query_string should be a literal CQL statement with the exception of parameter placeholders that will be filled through the parameters argument of Session.execute().

See Statement attributes for a description of the other parameters.

class cassandra.query.PreparedStatement[source]

A statement that has been prepared against at least one Cassandra node. Instances of this class should not be created directly, but through Session.prepare().

A PreparedStatement should be prepared only once. Re-preparing a statement may affect performance (as the operation requires a network roundtrip).

A note about * in prepared statements: Do not use * in prepared statements if you might change the schema of the table being queried. The driver and server each maintain a map between metadata for a schema and statements that were prepared against that schema. When a user changes a schema, e.g. by adding or removing a column, the server invalidates its mappings involving that schema. However, there is currently no way to propagate that invalidation to drivers. Thus, after a schema change, the driver will incorrectly interpret the results of SELECT * queries prepared before the schema change. This is currently being addressed in CASSANDRA-10786.


Creates and returns a BoundStatement instance using values.

See BoundStatement.bind() for rules on input values.

class cassandra.query.BoundStatement(prepared_statement, retry_policy=None, consistency_level=None, routing_key=None, serial_consistency_level=None, fetch_size=<object object>, keyspace=None, custom_payload=None)[source]

A prepared statement that has been bound to a particular set of values. These may be created directly or through PreparedStatement.bind().

prepared_statement should be an instance of PreparedStatement.

See Statement attributes for a description of the other parameters.

prepared_statement = None

The PreparedStatement instance that this was created from.

values = None

The sequence of values that were bound to the prepared statement.


Binds a sequence of values for the prepared statement parameters and returns this instance. Note that values must be:

  • a sequence, even if you are only binding one value, or
  • a dict that relates 1-to-1 between dict keys and columns

Changed in version 2.6.0: UNSET_VALUE was introduced. These can be bound as positional parameters in a sequence, or by name in a dict. Additionally, when using protocol v4+:

  • short sequences will be extended to match bind parameters with UNSET_VALUE
  • names may be omitted from a dict with UNSET_VALUE implied.

Changed in version 3.0.0: method will not throw if extra keys are present in bound dict (PYTHON-178)

class cassandra.query.Statement[source]

An abstract class representing a single query. There are three subclasses: SimpleStatement, BoundStatement, and BatchStatement. These can be passed to Session.execute().

retry_policy = None

An instance of a cassandra.policies.RetryPolicy or one of its subclasses. This controls when a query will be retried and how it will be retried.

consistency_level = None

The ConsistencyLevel to be used for this operation. Defaults to None, which means that the default consistency level for the Session this is executed in will be used.

fetch_size = <object object>

How many rows will be fetched at a time. This overrides the default of Session.default_fetch_size

This only takes effect when protocol version 2 or higher is used. See Cluster.protocol_version for details.

New in version 2.0.0.

keyspace = None

The string name of the keyspace this query acts on. This is used when TokenAwarePolicy is configured for Cluster.load_balancing_policy

It is set implicitly on BoundStatement, and BatchStatement, but must be set explicitly on SimpleStatement.

New in version 2.1.3.

custom_payload = None

Custom Payloads to be passed to the server.

These are only allowed when using protocol version 4 or higher.

New in version 2.6.0.

is_idempotent = False

Flag indicating whether this statement is safe to run multiple times in speculative execution.


The partition_key portion of the primary key, which can be used to determine which nodes are replicas for the query.

If the partition key is a composite, a list or tuple must be passed in. Each key component should be in its packed (binary) format, so all components should be strings.


The serial consistency level is only used by conditional updates (INSERT, UPDATE and DELETE with an IF condition). For those, the serial_consistency_level defines the consistency level of the serial phase (or “paxos” phase) while the normal consistency_level defines the consistency for the “learn” phase, i.e. what type of reads will be guaranteed to see the update right away. For example, if a conditional write has a consistency_level of QUORUM (and is successful), then a QUORUM read is guaranteed to see that write. But if the regular consistency_level of that write is ANY, then only a read with a consistency_level of SERIAL is guaranteed to see it (even a read with consistency ALL is not guaranteed to be enough).

The serial consistency can only be one of SERIAL or LOCAL_SERIAL. While SERIAL guarantees full linearizability (with other SERIAL updates), LOCAL_SERIAL only guarantees it in the local data center.

The serial consistency level is ignored for any query that is not a conditional update. Serial reads should use the regular consistency_level.

Serial consistency levels may only be used against Cassandra 2.0+ and the protocol_version must be set to 2 or higher.

See Lightweight Transactions (Compare-and-set) for a discussion on how to work with results returned from conditional statements.

New in version 2.0.0.


Specifies an unset value when binding a prepared statement.

Unset values are ignored, allowing prepared statements to be used without specify

See for further details on semantics.

New in version 2.6.0.

Only valid when using native protocol v4+

class cassandra.query.BatchStatement(batch_type=BatchType.LOGGED, retry_policy=None, consistency_level=None)[source]

A protocol-level batch of operations which are applied atomically by default.

New in version 2.0.0.

batch_type specifies The BatchType for the batch operation. Defaults to BatchType.LOGGED.

retry_policy should be a RetryPolicy instance for controlling retries on the operation.

consistency_level should be a ConsistencyLevel value to be used for all operations in the batch.

custom_payload is a Custom Payloads passed to the server. Note: as Statement objects are added to the batch, this map is updated with any values found in their custom payloads. These are only allowed when using protocol version 4 or higher.

Example usage:

insert_user = session.prepare("INSERT INTO users (name, age) VALUES (?, ?)")
batch = BatchStatement(consistency_level=ConsistencyLevel.QUORUM)

for (name, age) in users_to_insert:
    batch.add(insert_user, (name, age))


You can also mix different types of operations within a batch:

batch = BatchStatement()
batch.add(SimpleStatement("INSERT INTO users (name, age) VALUES (%s, %s)"), (name, age))
batch.add(SimpleStatement("DELETE FROM pending_users WHERE name=%s"), (name,))

New in version 2.0.0.

Changed in version 2.1.0: Added serial_consistency_level as a parameter

Changed in version 2.6.0: Added custom_payload as a parameter

serial_consistency_level = None

The same as Statement.serial_consistency_level, but is only supported when using protocol version 3 or higher.

batch_type = None

The BatchType for the batch operation. Defaults to BatchType.LOGGED.


This is a convenience method to clear a batch statement for reuse.

Note: it should not be used concurrently with uncompleted execution futures executing the same BatchStatement.

add(statement, parameters=None)[source]

Adds a Statement and optional sequence of parameters to be used with the statement to the batch.

Like with other statements, parameters must be a sequence, even if there is only one item.

add_all(statements, parameters)[source]

Adds a sequence of Statement objects and a matching sequence of parameters to the batch. Statement and parameter sequences must be of equal length or one will be truncated. None can be used in the parameters position where are needed.

class cassandra.query.BatchType[source]

A BatchType is used with BatchStatement instances to control the atomicity of the batch operation.

New in version 2.0.0.


Atomic batch operation.


Non-atomic batch operation.


Batches of counter operations.

class cassandra.query.ValueSequence[source]

A wrapper class that is used to specify that a sequence of values should be treated as a CQL list of values instead of a single column collection when used as part of the parameters argument for Session.execute().

This is typically needed when supplying a list of keys to select. For example:

>>> my_user_ids = ('alice', 'bob', 'charles')
>>> query = "SELECT * FROM users WHERE user_id IN %s"
>>> session.execute(query, parameters=[ValueSequence(my_user_ids)])
class cassandra.query.QueryTrace[source]

A trace of the duration and events that occurred when executing an operation.

request_type = None

A string that very generally describes the traced operation.

duration = None

A datetime.timedelta measure of the duration of the query.

client = None

The IP address of the client that issued this request

This is only available when using Cassandra 2.2+

coordinator = None

The IP address of the host that acted as coordinator for this request.

parameters = None

A dict of parameters for the traced operation, such as the specific query string.

started_at = None

A UTC datetime.datetime object describing when the operation was started.

events = None

A chronologically sorted list of TraceEvent instances representing the steps the traced operation went through. This corresponds to the rows in for this tracing session.

trace_id = None

uuid.UUID unique identifier for this tracing session. Matches the session_id column in system_traces.sessions and

populate(max_wait=2.0, wait_for_complete=True, query_cl=None)[source]

Retrieves the actual tracing details from Cassandra and populates the attributes of this instance. Because tracing details are stored asynchronously by Cassandra, this may need to retry the session detail fetch. If the trace is still not available after max_wait seconds, TraceUnavailable will be raised; if max_wait is None, this will retry forever.

wait_for_complete=False bypasses the wait for duration to be populated. This can be used to query events from partial sessions.

query_cl specifies a consistency level to use for polling the trace tables, if it should be different than the session default.

class cassandra.query.TraceEvent[source]

Representation of a single event within a query trace.

description = None

A brief description of the event.

datetime = None

A UTC datetime.datetime marking when the event occurred.

source = None

The IP address of the node this event occurred on.

source_elapsed = None

A datetime.timedelta measuring the amount of time until this event occurred starting from when source first received the query.

thread_name = None

The name of the thread that this event occurred on.

exception cassandra.query.TraceUnavailable[source]

Raised when complete trace details cannot be fetched from Cassandra.