October 20, 2016
- Added constant specultive exectution which preemptively runs mutiple of the same query in an effort to smooth out latency spikes.
- Added idempotent queries to support speculative execution and allow automatic
- SSL can be now be used without initializing the underlying implmentation
(e.g. OpenSSL). Use
- Fixed an issue where a connection is incorrectly marked READY after it’’s already been marked DEFUNCT (CPP-401)
- Fixed pointer aliasing issues in
- Fixed an issue where random partitioner hash values were being incorrectly generated (CPP-403)
August 22, 2016
- Contact points are now randomized by default (CPP-193)
- Token-aware routing can be used without enabling schema metadata (CPP-387)
- Multiple IP addresses can be resolved from a single domain (CPP-364)
- Fixed issue that would cause quadradic ring processing with invalid replilcation factors (CPP-298)
- Fixed issue where schema change handler could hang if an error is returned by Cassandra (CPP-381)
- Fixed crash caused by connecting seperate sessions in multiple threads (CPP-385)
- Fixed issue where the control connection could timeout as a result of schema or token map processing (CPP-388)
- Greatly improved the performance of building the token map for token aware routing (CPP-389)
- Fixed issue where heartbeats were restarted on unresponseive connections and prevented the connection from being terminated (CPP-392)
June 24, 2016
- Added the per-request timeouts to statement and batch
- Added the ability to remove custom payload items
- Fixed issue where cass_date_time_to_epoch() unable to handle dates before Cassandra date epoch (value: 2147483648)
June 9, 2016
- Fixed issue where
cass_future_get_result()and similiar methods would dereference a NULL pointer if a future is set to an error.
June 1, 2016
- Added the ability to add a custom SASL authenticator
- Added reverse DNS lookup for IP addresses which can be enabled using
- Added a DNS resolve timeout that can be set using
- Added the ability for SSL peer identity to be verified by hostname.
- Added the ability to bind/set/append custom types with a class name.
CassStatementfunction to clear/resize parameters
- Fixed crash caused by tables with the same name in adjacent keyspaces.
- Fixed an invalid memory read in processing strategy options metadata.
- Fixed issue where
nullvalues didn’’t have proper datatypes.
- Fixed issue where an “UP” event could cause the session connection process to hang.
- Reduced reconnection log chatter at the
CASS_LOG_ERRORlevel. A failed reconnection attempt is only logged one time as an error.
March 14, 2016
- Added support for materialized view schema metadata.
- Added support for secondary index schema metadata.
- Added cluster key order to table schema metadata via
- Added frozen flag to
- Added support getting the connected cluster’’s Cassandra version via
- Added blacklist, whitelist DC, and blacklist DC load balancing policies. Thanks to Walid Aitamer (waitamer).
- Fixed the control connection to only become “ready” during the initial connection process.
- Fixed assertion in data type comparison for collection/tuple types.
- Fixed reference count issue in
- Fixed retry consistency levels for batches. Batches now correctly use the downgraded consistency levels supplied from retry policies.
- Fixed function name typo:
cass_data_type_sub_type_count(). The old spelling is still present, but deprecated.
cass::VersionNumberto handle tick-tock release version where only “major.minor” are provided.
- Fixed function
cass_session_connect_keyspace()to correctly return an error when the keyspace is invalid.
- Fixed invalid logic in schema metadata swap method which could cause unecessary copy-on-writes.
- Fixed bottleneck in
cass::Sessionby using copy-on-write instead of a highly contended lock to track the current keyspace.
- Changed “major/minor” to “major_version/minor_version” in
cass::VersionNumberto avoid naming conflicts on some platforms.
December 14, 2015
- Added basic support for Cassandra 3.0. The driver supports Cassandra 3.0 schema metadata for keyspaces, tables, columns, user types, functions, and aggregates.
- Changed the default consistency from
November 19, 2015
- Fixed issue where cass_data_type_sub_data_type() was always referencing a NULL pointer
- Fixed cass::Atomic (intrinsic) for 32-bit MSVC
October 29, 2015
- Refactored the schema metadata API (
CassSchema) to use concrete types (
CassTableMeta, etc.) instead of the single
- Added support for UDF/UDA schema metadata
- Added support for whitelist filtering i.e. whitelist load balancing policy
Address::compare()issue that cause invalid comparisons for IPv4
- Fixed issue that caused
StreamManagerto generate invalid stream IDs in Windows release builds.
- Fixed issue where
cass_cluster_set_protocol_version()would return an error for protocol versions higher than v3
- Changed the default consistency from
- When using prepared statements with protocol v4 (and higher) the driver will use partition key metadata returned in the prepared response instead of using schema metadata.
- Improved the performance of UUID to string and UUID from string conversions
September 21, 2015
- Added support for Cassandra 2.2 data types
cass_uint32_tsupport for the Cassandra 2.2
- Added functions to convert from Unix Epoch time (in seconds) to and from the
- Added support for sending and recieving custom payloads to and from Cassandra
- Added support for server-side warnings
- Added new error response functions to handle the ‘Read_failure’, ‘Write_failure’, ‘Function_failure’, and ‘Already_exists’ errors
- Fixed SSL ring buffer memory leak
- Fixed potential memory issue caused by using named parameters
- Fixed hanging issue when calling
- Fixed hanging issue caused by
- Fixed invalid state assertion in
- Fixed UDTs with
- A value of
nullis no longer implictly used for unbound statement parameters. An error is returned (
CASS_ERROR_LIB_PARAMETER_UNSET) for Cassandra 2.1 and earlier or the UNSET value is used for Cassandra 2.2 and later
August 11, 2015
- Added support for retry policies
- Added support for client-side timestamps
- Exposed raw paging state token via
cass_result_paging_state_token()and it can be added to a statement using
- Added support for enabling/disabling schema metadata using
- Connections that timeout will not disconnect an entire conneciton pool
TCP_NODELAYis enabled by default
- Connections now support using 32,767 stream IDs when using the CQL Native Protocol v3
cass_uuid_max_from_time()now correctly considers the time parameter as epoch time
July 7, 2015
- Added support for tuples and UDTs
- Added support for nested collections
- Added support for named parameters when using simple queries (non-prepared)
- Added support for protocol version 3 (and schema changes)
- Fixed SSL error handling
- Added libuv read buffer caching in cass::Connection
May 15, 2015
- Removed logging thread.
cass_log_set_queue_size()are now deprecated. A logging callback registered with
cass_log_set_callback()will now be called on mulitple threads concurrently.
- Fixed an issue where the driver might crash/hang when
system.localcontains no entries
- Fixed an issue where the driver might hang under high-load due to missing
memory fence before
April 23, 2015
- Added latency-aware load balancing. To enable use:
- Added performance metrics. To get metrics use:
CassInetstring conversion functions:
- Removed internal dependency on Boost. Optional: Boost Atomic can still be used as an implementation for internal atomic operations and memory fences.
- Removed wrapper types
- Added string length functions (functions with the suffix “_n”) for all
functions that previously took a null-terminated or
- Added missing
- Fixed issue where time-based UUIDs (
CassUuidv4) had the incorrect time on 32-bit systems
Mar 2, 2015
- Added testing documentation
- Added docmentation for the
- Added a contributing document
- Added version defines to
- Fixed an issue where connections were attempted against decommissioned nodes
- Fixed an issue where an invalid key index was used for a single entry token-aware routing key
- Improved build error output when an incompatabile version of Boost is used to build the driver
Feb 2, 2015
- Libuv 1.x supported
cass_cluster_set_num_threads_io()now returns an error code
- Fixed an issue where a node in a remote DC could reconnect even if it was supposed to be ignored.
- Reduced the defaults for IO worker threads (1), core connections (1), and max connections (2).
- Fixed crash caused by a null token-aware key value.
Dec 22, 2014
- Added global logging. Logging is no longer configured per session.
- DC-aware load balancing policy is now the default and uses the DC of the first connected contact point. The DC-aware policy now has settings to control the number of remote hosts used in each DC (after exhausting the hosts in the local DC) and whether to ignore local consistency levels.
- Added socket options for TCP nodelay and keepalive:
CassSessionobjects must now be explictly allocated/freed using
cass_cluster_connect()has been removed and replaced by
- SIGPIPE is ignored on driver threads
- Connection timeout now covers the entire connection handshake instead of just the socket connection.
- UUID API has been updated to use explict state via
CassUuidGeninstead of using internal global state. See
cass_uuid_gen_new()and corresponding functions.
Nov 20, 2014
- Added SSL support
- Added token-aware load balancing. It’s enable by default. It can
be disable using
- Added functions to get schema metadata,
cass_session_get_schema()can be use to get a snapshot of the schema metadata and
cass_meta_*()functions can be used to retrieve detailed information about keyspaces, tables, and columns.
- Significantly improved request throughput by batching socket writes.
Batch sizes can be configured using the function
cass_cluster_set_max_requests_per_flush(). The default is 128 because the maximum write size is bounded by the number of pipelined requests per connection (for version 1 and 2 of the native protocol).
- Added backpressure mechanism that can be configured for by the
number of outstanding bytes and the number of requests on a connection. These can
be configured using
- Fixed a crash caused by a request timing out followed by a call to the conneciton read callback then the write callback.
- Fixed an issue that prevented the driver from recovering from a full cluster outage.
- Fixed an issue that allowed
uv_queue_work()to be called from an application thread.
- Added logic to prevent redundant node refreshes when estabilishing a new connection from multiple IO workers.
- Fixed a crash caused by Cassandra 2.1.0 not returning metadata on execute requests.
- Fixed a memory leak that could leak request handlers if the control connection ran out of stream ids.
- Fixed a crash caused by calling an invalid timeout callback when waiting for connection to become available in a pool.
- Build will use an external version of Boost if present.
Sep 11, 2014
- Added node discovery and now handle topology/status change events
- Queries that modify schema wait for cluster schema agreement
- Added DC aware load balancing policy. The default policy is set to round robin.
- Added new methods to configure the load balancing policy,
- It is now possible for a session future to return an error when connecting to a cluster. This can occur when the driver is unable to connect to any contact points or the provided authentication credentials are invalid.
Aug 13, 2014
- Added callbacks to futures,
- Added plaintext authentication support,
- Added map iterator,
cass_iterator_from_map(), key and value pairs can be retreived using
- Values can be bound by name (statements created from prepared statments only)
- Values can be retreived by name from result sets
- Reduced overhead of prepared statements
- Small performance improvement for all queries via reduced allocations
Jul 17, 2014
- Removed C++1x dependency
- Support for Cassandra 1.2 (native protocol version 1)
- Support for automatic paging
- Added method to set serial consistency on statement
- Added batch type to batches
- Exposed options
cass_result_column_name()to get column name from a result set
StreamManagerso it no longer causes stack corruption
- Fixed issue where connections would log “’Timed out during startup’ error on startup for …” when a Session was closed while a connection was during its startup process
- Fixed Row and collections iterator so they return the first and last item properly
- Improved logic for creating new connections so it’s based request throughput
- Removed the
setoptmethod added set methods for all options
- Removed the
cass_collection_new()no longer take a consistency parameter
cass_collection_new()now requires a colleciton type parameter
is_mapparameter removed from
Jun 16, 2014
- Support for Cassandra 2.0 (native protocol version 2)
- Support for prepared and batch statements
- Support for collections