Releases: blevesearch/bleve
Releases · blevesearch/bleve
v2.4.2
Bug Fixes
- Set
MaxSegmentFileSize
for force merge whenSingleSegmentMergePlanOptions
goes into use (#2050) - Protect code against any bolt failures (#2043)
- Fix size estimation for various field types (#2052)
- Address an out-of-bounds panic that could occur with zapx/v16 in the event of a single non-vector field (#2058)
Improvements
- Support new search params for vector search queries - [
ivf_probe_pct
,ivf_max_codes_pct
] (#2049)
Milestone
v2.4.1
Bug Fixes
- Addressed a bug where
ClientContextID
was missing from SearchRequest when library used withvectors
(#2014) - Fix to a memory leak in vector query path (#2023)
- Fix to issue #2027, error in parsing a certain datetime syntax (#2030)
- Fix that mitigates a race between persister's stale segment removal and index copy operations (#2032)
- Nested field mapping determination was broken (#2031)
Improvements
- Vector search requires the faiss dynamic library to be built from blevesearch/faiss@d9db66a
- Support for new data type -
vector_base64
which allows for interpreting vector float32s encoded as base64 following littleEndian byte ordering (#2012) - Vector dimensionality support up to 4096 from 2048 (#2015)
- New vector index optimization -
memory_efficient
alongside pre-existingrecall
andlatency
(#2040) - Better resource usage and efficiency in vector indexing (blevesearch/zapx#241, blevesearch/zapx#234)
- Delayed releasing of deserialized vector indexes to support better query throughput (#2016)
Milestone
v2.4.0
Bug Fixes
- Fixes to spanish accents normalization (#1957)
- Critical fix around facets merging which could've caused inconsistent results (#1946)
Improvements
- Support for vectors indexing and search
- Please read: vectors.md for more details
- With this we will now default to a new zap format: v16, which co-locates text and vector indexes within scorch segments. v16 will be able to interpret v15 zap segments.
- Requires the faiss dynamic library to be built from blevesearch/faiss@7b119f4
- An empty
query
attribute within aSearchRequest
will now be interpreted as amatch_none
(#2000)
Milestone
v2.3.10
Bug Fixes
- Fix format of date times displayed in search result, to not drop precision (#1868)
Improvements
- Upgraded to
golang.org/x/[email protected]
to overcome a vulnerability (#1845) - Support for polish language analyzer -
pl
(#1825) - Additional attribute in search response to determine partial/complete hits for match queries (#1848)
- Performance optimization to speed up documents processing per mapping definition (#1865)
- Support fuzziness factor for matchphrase, multiphrase & phrase queries (#1847)
- Geo shape polygon decode optimization w/ buffer re-use (#1864)
- Override-able Marshal/Unmarshal methods (#1880)
- Enhanced data time format support:
sanitizedgo
that validates layout against supported golang constants (#1877)- Ability to apply custom datetime parsers to query criteria and facets (#1860, #1878)
- Ability to parse date times in these formats -
unix_micro
,unix_sec
,unix_milli
,unix_nano
(#1870) - Additional layouts support with
percentstyle
andisostyle
(#1866)
Milestone
v2.3.9
Bug Fixes
- Fixes along recording IO bytes read for different query types (#1829)
- Also changes the attribute name in the
SearchResult
construct frombytesRead
tocost
- Also changes the attribute name in the
- Fix incorrect facet counts when facetted queries are coupled with sorting over the same field (#1836)
Improvements
- Upgrade to RoaringBitmap/roaring@v1.2.3 (#1822)
- Upgrade to etcd-io/bblot@v1.3.7 (#1824)
- Consolidated some redundant work in a hot index path with #1832 which showed index time speed up by about 30% in certain situations especially while working with large documents.
- Upgrade to vellum@v1.0.10 which contains a minor optimization (#1838)
Milestone
v2.3.8
Bug Fixes
- Addresses panics that could occur within zap (with [email protected])
- Fixes analyzer inheritance for default mapping with #1807
- Addresses #1816 by fixing CLI's completion generation with #1817
Enhancements
- Introduces a new not-so-aggressive plural stemmer (
stemmer_en_plural
) for english with #1808
Milestone
v2.3.7
Bug Fixes
- Replaced deprecated io/ioutil functions with #1781
- Fixed situation where bytesRead will not appear in search response if query gets everything it needs from mmaped content with #1788
- Fixed a panic due to misalignment within chunkedContentCoder on 32-bit platforms with blevesearch/zapx#148 and #1791 (only applicable to zap@v15)
- A bot upgraded the golang.org/x/text version we use from 0.3.7 to 0.3.8 with #1792 🥇
- Include bytesWritten into bbolt (metadata for scorch) with #1774
Enhancements
- Support for improved rollback behavior by allowing sampling of snapshots based on a time interval with #1776 . Default behavior remains the same
- Changed the signature of scorch's AsyncErrorCallback with #1800 to provide more context. This is a breaking change.
Milestone
v2.3.6
Bug Fixes
- More efficient alignment for chunkedContentCoder (blevesearch/zapx#147)
- Addressed issue in indexing unigrams before punctuations in
Japanese (cjk)
(#1724) - Addressed facets string output limitation for date ranges and numeric ranges reported by #1763 with #1764
- Fixed inaccurate IO stats accounting while loading segment metadata (#1769)
Enhancements
- New
bytesRead
attribute in search response (#1752, blevesearch/zapx#142)- Applicable to zapx v15 only
- New tests showcasing performance benchmarks for popular queries (#1755)
Milestone
v2.3.5
Bug Fixes
- Fixed throughput regression reported with #1731
- Fix intersection logic for geo JSON searches, more unit tests
- Fix to support IP Range query
Other
- Breaking change in defining custom analyzers
- #1729 (this is needed to allow custom analyzer definition without needing to always define CharFilters, Tokenizer, TokenFilters)
- Update spf13/cobra version v1.4.0 (to drop a number of indirect dependencies)