- compatibility with Elasticsearch 6.0.0
streaming_bulkhelper now supports retries with incremental backoff
scanhelper properly checks for successful shards instead of just checking
- compatible release with elasticsearch
- fix handling of UTF-8 surrogates
bulkhelpers now extract
pipelineparameter from the action dictionary.
Compatibility with elasticsearch 5.3
The client now automatically sends
Content-Type http header set to
application/json. If you are explicitly passing in other encoding than
json you need to set the header manually.
- Fixed sniffing
Fixed performance regression in
Version compatible with elasticsearch 5.0
- when using SSL certificate validation is now on by default. Install
certifior supply root certificate bundle.
elasticsearch.tracelogger now also logs failed requests, signature of internal logging method
log_request_failhas changed, all custom connection classes need to be updated
headersarg to connections to support custom http headers
- passing in a keyword parameter with
Noneas value will cause that param to be ignored
pingnow ignores all
TransportErrorexceptions and just returns
- increase default size for
scanhelper to 1000
Transport.perform_requestto just return the body, not status as well.
client_keyargument to configure client certificates
- debug logging now includes response body even for failed requests
Due to change in json encoding the client will no longer mask issues with
encoding - if you work with non-ascii data in python 2 you must use the
unicode type or have proper encoding set in your environment.
- adding additional options for ssh -
ssl_assert_fingerprintto the default connection class
- fix sniffing
- move multiprocessing import inside parallel bulk for Google App Engine
- Elasticsearch 2.0 compatibility release
- removed thrift and memcached connections, if you wish to continue using those, extract the classes and use them separately.
- added a new, parallel version of the bulk helper using thread pools
- In helpers, removed
bulk_indexas an alias for
- elasticsearch 2.0 compatibility
- thrift now deprecated, to be removed in future version
- make sure urllib3 always uses keep-alive
helpers.reindexnow supports reindexing parent/child documents
- Add support for
query_cacheparameter when searching
- helpers have been made more secure by changing defaults to raise an exception on errors
- removed deprecated options
replicationand the deprecated benchmark api.
AddonClientclass to allow for extending the client from outside
- Using insecure SSL configuration (
verify_cert=False) raises a warning
reindexhelper to accept any kwargs for underlying
- when doing an initial sniff (via
sniff_on_start) ignore special sniff timeout
- option to treat
TransportErroras normal failure in
- fixed an issue with sniffing when only a single host was passed in
- Timeout now doesn’t trigger a retry by default (can be overriden by setting
- Introduced new parameter
(503, 504, )) controls which http status code should lead to a retry.
- Implemented url parsing according to RFC-1738
- Added support for proper SSL certificate handling
- Required parameters are now checked for non-empty values
- ConnectionPool now checks if any connections were defined
- DummyConnectionPool introduced when no load balancing is needed (only one connection defined)
- Fixed a race condition in ConnectionPool
Compatibility with newest (1.3) Elasticsearch APIs.
- Filter out master-only nodes when sniffing
- Improved docs and error messages
Bugfix release fixing escaping issues with
Compatibility with newest Elasticsearch APIs.
- Test helpers -
get_test_clientfor use in your tests
- Python 3.2 compatibility
simplejsonif installed instead of stdlib json library
- Introducing a global
request_timeoutparameter for per-call timeout
- Bug fixes
Elasticsearch 1.0 compatibility. See 0.4.X releases (and 0.4 branch) for code compatible with 0.90 elasticsearch.
- major breaking change - compatible with 1.0 elasticsearch releases only!
- Add an option to change the timeout used for sniff requests (
- empty responses from the server are now returned as empty strings instead of None
nameas another optional parameter due to issue #4539 in es repo. Note that the order of params have changed so if you are not using keyword arguments this is a breaking change.
helpers.bulk(alias put in place for backwards compatibility, to be removed in future versions)
helpers.streaming_bulkto consume an iterator and yield results per operation
helpers.streaming_bulkare no longer limited to just index operations.
- unicode body (for
incices.analyzefor example) is now handled correctly
Connectionclasses to return headers as well. This is a backwards incompatible change for people who have developed their own connection class.
- changed deserialization mechanics. Users who provided their own serializer that didn’t extend
JSONSerializerneed to specify a
- minor bug fixes
- Fixes to
helpers.bulk_index, better error handling
- More benevolent
hostsargument parsing for
requestsno longer required (nor recommended) for install
ignoreparam accepted by all APIs
- Fixes to