• Added support for quantized dense vector options (#1948)
• Added support for composable index templates (#1943)

• Autogenerate parts of the library using documentation and types from the Elasticsearch specification
  • query classes (#1890)
  • aggregation classes (#1918)
  • response classes (#1929 #1932)
• Support pipe syntax to declare optional document fields (#1937)
• Ignore document attributes typed with ClassVar (#1936)
• Support Python 3.13 (#1938)

• Fixed the use of dictionaries as values in Terms query (#1920)

• Fixed regression introduced in Terms query class (#1907)
• Removed unnecessary filter argument in AggBase.__getitem__ (#1903)
• Fixed deserialization of datetime.date fields (#1914)

• Added support for any iterables to the Terms query (#1887)
• Added back support for tuples and other iterables to Search.source() method (#1895)
• Added recursive option to AttrDict.to_dict() (#1892)
• Removed unused analyzer from search as you type example (#1883)

• Added support for the semantic_text field and semantic query type (#1881)
• Removed extra __orig_class__ attribute in aggregation responses (#1877)

8.15.0 (2024-08-09)

• Added the option to use Python types to declare document fields (#1845)
• Added type annotations (#1533)
• Added support for bulk document operations with Document.bulk() (#1864)
• Added the ConstantKeyword field to the top-level package (#1843)
• Added async_connections to the top-level package (#1865)
• Added index creation to the aggregations example (#1862)

• Added text_expansion query clause (#1837)
• Added Response.search_after() and Search.search_after() methods for efficient iteration (#1829)
• Added point in time support and the iterate() method in the Search class (#1833)
• Added support for slicing multiple times in Search class (#1771) Added support for regular expressions in Completion.suggest() (#1836)
• Fixed suggest() method of the Completion class to format requests correctly. (#1836)
• Fixed Document.update() to accept fields set to None or empty (#1820)
• Started work on type hints (Thanks Caio Fontes for leading this effort!)
  • Added Type hints to function.py (#1827)
  • Added Type hints to query.py (#1821)

• Added support for knn as a query option (#1770)
• Made the dims attribute of the dense vector type optional (#1776)
• Added missing inner_hits option to search.knn() method (#1777)
• Added support for detecting document updates in InnerDoc attributes (#1535)
• Changed _expand__to_dot setting to resolve at runtime (#1633)
• Added explicit error message when unsupported minimum_should_match values are used (#1774)
• Added the EmptySearch class (#1780)
• Added several missing aggregations:
  • AdjacencyMatrix (#1553)
  • CategorizeText (#1588)
  • GeohexGrid (#1590)
  • IPPrefix (#1592)
  • RandomSampler (#1594)
  • GeoLine (#1628)
  • MatrixStats (#1630)
  • TopMetrics (#1706)
• Added params option to the FacetedSearch object (#1500)
• Added support for passing a dictionary in the script option for a document update (#1560)
• Added keys() and items() methods to AttrDict class (#1784)
• Added a to_list() method to the AttrList class (#1584)
• Fixed various documentation issues and typos (#1769, #1615, #1585, #1318, #1223)
• Added a vector search example (#1778)

• Added asyncio support (#1714)
• Dropped support for Python 3.7 (#1717)
• Stopped mixing body and parameters in UpdateByQuery (#1702)

• Added Search.knn() method (#1691)
• Added Search.rank() method (undocumented as it still is in technical preview) (#1692)
• Fixed importing collapse from dictionary (#1689)

• Added support for Python 3.12 (#1680)
• Added Search.collapse() (#1649, contributed by @qcoumes)

• Added Elasticsearch 8.x support (#1664)
• Dropped support for Python 2.7 and 3.5 (#1606, contributed by @hugovk)
• Added support for Python 3.10 and 3.11 (#1608, contributed by @hugovk)
• Added the MultiTerms aggregation (#1543, contributed by @Telomeraz)
• Added the CombinedFields query (#1557, contributed by @Telomeraz)

• Added Elasticsearch 8.x support (#1664)
• Dropped support for Python 2.7 and 3.5 (#1606, contributed by @hugovk)
• Added support for Python 3.10 and 3.11 (#1608, contributed by @hugovk)
• Added the MultiTerms aggregation (#1543, contributed by @Telomeraz)
• Added the CombinedFields query (#1557, contributed by @Telomeraz)

• Fixed DeprecationWarnings that would be emitted from deprecated usages of the body parameter in the Python Elasticsearch client.

Added

• Added the ConstantKeyword, RankFeatures field types (#1456, #1465)
• Added the ScriptScore query type (#1464)
• Added UpdateByQueryResponse.success() method (#1463)
• Added return_doc_meta parameter to Document.save() and Document.update() for accessing the complete API response (#1466)
• Added support for calendar_interval and fixed_interval to DateHistogramFacet (#1467)
• Added Document.exists() method (#1447, contributed by @dem4ply)
• Added support for the year interval to DateHistogramFacet (#1502, contributed by @nrsimha)

Fixed

• Fixed issue where to_dict() should be called recursively on Search.extras and **kwargs (#1458)
• Fixed inverse of an empty Bool query should be MatchNone (#1459)
• Fixed issue between retry_on_conflict and optimistic concurrency control within Document.update() (#1461, contributed by @armando1793)

Added

• Added Intervals, MatchBoolPrefix, Shape, and Wrapper queries (#1392, #1418)
• Added Boxplot, RareTerms, VariableWidthHistogram, MedianAbsoluteDeviation, TTest, CumulativeCardinality, Inference, MovingPercentiles, and Normalize aggregations (#1416, #1418)
• Added __all__ and removed all star imports from elasticsearch_dsl namespace to avoid leaking unintended names (#1390)

Fixed

• Fixed an issue where Object and Nested could mutate the inner doc_class mapping (#1255, contributed by @l1nd3r0th)
• Fixed a typo in query SpanContaining, previously was SpanContainining (#1418)

• Fixed issue when slicing a Search that would result in a negative size instead of a size of 0. (#1360, contributed by @bk-equityzen)

• Added support for geotile_grid aggregation (#1350, contributed by @owrcasstevens)

• Added the DenseVector and SparseVector data types (#1278)

• Added the SearchAsYouType field (#1295, contributed by @dpasqualin)

• Fixed name of DoubleRange (#1272, contributed by @braunsonm)

• Optimistic concurrent control for Document.delete

• Removing deprecated DocType

• Proper count caching for ES 7.x

• Support for multiplexer token filter

• Don't substitute for __ in FacetedSearch

• Compatibility with Elasticsearch 7.x

• Document.save() now returns "created" or "updated"

• Dropped support for Python 2.6, 3.2, and 3.3

• When using fields the values are no longer merged into the body of the document and have to be accessed via .meta.fields only

• Index.document now correctly sets the Document's _index only when using default index (#1091)

• Document inheritance allows overriding Object and Nested field metadata like dynamic

• adding auto_date_histogram aggregation

• Do not change data in place when (de)serializing

• Analyzer.simulate now supports built-in analyzers

• proper (de)serialization of the Range wrapper

• Added search_analyzer to Completion field

• Fixed logic around defining a different doc_type name.

• Added retry_on_conflict parameter to Document.update.

• fields defined on an index are now used to (de)serialize the data even when not defined on a Document

• Allow Index.analyzer to construct the analyzer

• Detect conflict in analyzer definitions when calling Index.analyzer

• Detect conflicting mappings when creating an index

• Add simulate method to analyzer object to test the analyzer using the _analyze API.

• Add script and script_id options to Document.update

• Facet can now use other metric than doc_count

• Range objects to help with storing and working with _range fields

• Improved behavior of Index.save where it does a better job when index already exists

• Composite aggregations now correctly support multiple sources aggs

• UpdateByQuery implemented by @emarcey

• allow users to redefine doc_type in Index (#929)

• include DocType in elasticsearch_dsl module directly (#930)

Backwards incompatible change - DocType refactoring.

In 6.2.0 we refactored the DocType class and renamed it to Document. The primary motivation for this was the support for types being dropped from elasticsearch itself in 7.x - we needed to somehow link the Index and Document classes. To do this we split the options that were previously defined in the class Meta between it and newly introduced class Index. The split is that all options that were tied to mappings (like setting dynamic = MetaField('strict')) remain in class Meta and all options for index definition (like settings, name, or aliases) got moved to the new class Index.

You can see some examples of the new functionality in the examples directory. Documentation has been updated to reflect the new API.

DocType is now just an alias for Document which will be removed in 7.x. It does, however, work in the new way which is not fully backwards compatible.

• Percolator field now expects Query objects as values

• you can no longer access meta fields on a Document instance by specifying ._id or similar. Instead all access needs to happen via the .meta attribute.

• Implemented NestedFacet for FacetedSearch. This brought a need to slightly change the semantics of Facet.get_values which now expects the whole data dict for the aggregation, not just the buckets. This is a backwards incompatible change for custom aggregations that redefine that method.

• Document.update now supports refresh kwarg

• DslBase._clone now produces a shallow copy, this means that modifying an existing query can have effects on existing Search objects.

• Empty Search no longer defaults to match_all query and instead leaves the query key empty. This is backwards incompatible when using suggest.

• Removed String field.

• Fixed issue with Object/Nested deserialization

Fixing wheel package for Python 2.7 (#803)

Backwards incompatible release compatible with elasticsearch 6.0, changes include:

• use doc as default DocType name, this change includes: * DocType._doc_type.matches method is now used to determine which DocType should be used for a hit instead of just checking _type

• Nested and Object field refactoring using newly introduced InnerDoc class. To define a Nested/Object field just define the InnerDoc subclass and then use it when defining the field:

  class Comment(InnerDoc):
      body = Text()
      created_at = Date()
  
  class Blog(DocType):
      comments = Nested(Comment)

• methods on connections singleton are now exposed on the connections module directly.

• field values are now only deserialized when coming from elasticsearch (via from_es method) and not when assigning values in python (either by direct assignment or in __init__).

• fix ip_range aggregation and rename the class to IPRange.

  Iprange is kept for bw compatibility

• fix bug in loading an aggregation with meta data from dict

• add support for normalizer parameter of Keyword fields

• IndexTemplate can now be specified using the same API as Index

• Boolean field now accepts "false" as False

• fix constant score query definition

• DateHistogramFacet now works with datetime objects

• respect __ in field names when creating queries from dict

• make sure all response structers are pickleable (for caching)

• adding exclude to Search

• fix metric aggregation deserialization

• expose all index-level APIs on Index class

• adding delete to Search which calls delete_by_query API

• Renamed Result and ResultMeta to Hit and HitMeta respectively

• Response now stores Search which it gets as first arg to __init__

• aggregation results are now wrapped in classes and properly deserialized

• Date fields now allow for numerical timestamps in the java format (in millis)

• Added API documentation

• replaced generated classes with manually created

Version compatible with elasticsearch 5.0.

Breaking changes:

• String field type has been deprecated in favor of Text and Keyword

• fields method has been removed in favor of source filtering

• accessing missing string fields no longer returned '' but returns

  None instead.

• fix issues with bool's | and & operators and minimum_should_match

• inner_hits are now also wrapped in Response

• + operator is deprecated, .query() now uses & to combine queries

• added mget method to DocType

• fixed validation for "empty" values like '' and []

Compatibility with Elasticsearch 2.x:

• Filters have been removed and additional queries have been added. Instead of F objects you can now use Q.

• Search.filter is now just a shortcut to add queries in filter context

• support for pipeline aggregations added

Backwards incompatible changes:

• list of analysis objects and classes was removed, any string used as tokenizer, char or token filter or analyzer will be treated as a builtin

• internal method Field.to_python has been renamed to deserialize and an optional serialization mechanic for fields has been added.

• Custom response class is now set by response_class method instead of a kwarg to Search.execute

Other changes:

• FacetedSearch now supports pagination via slicing

• Search can now be iterated over to get back hits

• Search now caches responses from Elasticsearch

• DateHistogramFacet now defaults to returning empty intervals

• Search no longer accepts positional parameters

• Experimental MultiSearch API

• added option to talk to _suggest endpoint (execute_suggest)

• FacetedSearch now uses its own Facet class instead of built in

  aggregations

• 0.0.5 and 0.0.6 was released with broken .tar.gz on pypi, just a build fix

• Metadata fields (such as id, parent, index, version etc) must be stored (and

  retrieved) using the meta attribute (#58) on both Result and DocType objects or using their underscored variants (_id, _parent etc)

• query on Search can now be directly assigned

• suggest method added to Search

• Search.doc_type now accepts DocType subclasses directly

• Properties.property method renamed to field for consistency

• Date field now raises ValidationException on incorrect data

Added persistence layer (Mapping and DocType), various fixes and improvements.

Fix for python 2