Release 0.15.2

Features

• Added ByteString data type to Cottontail DB (thanks to @lucaro)
• Added entity delete-row command to CLI (thanks to @silvanheller)
• Added ability to execute LIKE queries via the query find CLI command

Bugfixes

• Fixed an issue in CSV exporter, which throw an out-of-bounds error (thanks to @sauterl)
• Fixed issue with SLF4j dependency (thanks to @rasmuswilli)

Changes

• Switched from OpenJDK 17 (deprecated) to Eclipse-Temurin based for Docker container
• Various quality of life changes to data import and export functionality in CLI (thanks to @sauterl)

Full Changelog: v0.15.1...0.15.2

Release 0.15.1

Bugfixes

• Fixed an issue with caching of Tx objects for multi-statement transactions.
• Fixed an issue that led to an erroneous ProductQuantizer when loading it from aPQConfig.
• Fixed various issues surrounding the TRUNCATE command.
• Fixed issue with handling of non-unique values in presence of a UniqueHashIndex.
• Fixed issue when exporting entities containing NaN values to JSON.
• Fixed issue in transaction handling when COMMITS are in conflict with other other transactions

Changes

• Optimised test-cases related to indexes and endpoints.
• Changed Cosine Similarity to Cosine Distance.

Thanks to @silvanheller and @lucaro for reporting issues and contributing various test cases.

Full Changelog: v0.15.0...0.15.1

Release 0.15.0

New features

• Cottontail DB now features a new storage engine, based on JetBrains Xodus
• Cottontail DB should handle intra-query parallelism more efficiently
• Reworked high-dimensional index structures. Most importantly, VAF and PQ indexes should now work reliably for various distance functions

Bugfix

• Fixed various bugs

Release 0.14.2

Format.PROTO is now the default value for all AbstractCottontailComma…

Release 0.14.1

Bugfix

• Fixed a critical bug that would lead to erroneous results when executing a Boolean query that exhibits intra-query parallelism. Thanks to @lucaro for helping to find the issue.

Full Changelog: v0.14.0...0.14.1

Release 0.14.0

New features

• Introducing C(arrot)LI, a separate CLI for Cottontail DB.
• Cottontail DB comes now in three modules core, dbms and cli
• Sort operators now produce consistent results between invocations even if compared columns result in a tie. This is achieved by ultimately using the tupleId of a record as a tie breaker.

Bugfix

• Fixed a major issue introduced in 0.13.0 connected to the new function invocation feature and intra-query parallelism, which could lead to inconsistent results

Release 0.13.3

New features

• The Docker container now comes for the linux/amd64 and the linux/arm64 platforms (thanks to @Spiess)

Bugfix

• Fixed an issue when parsing the LESS-than (<) comparison operator (thanks to @vogti).

Release 0.13.2

Optimization

• LuceneIndex now simply parses the query string when executing a MATCH operation, which gives access to full Lucene query syntax.
• Optimised handling of CottontailGrpcServer and closing of DefaultCatalogue when shutting down that server.

Bugfix

• Fixed issue with query planner that could lead to selection of non-executable query plans.

Release 0.13.1

Security

• Updated to Log4j2 2.17.0 which mitigates CVE-2021-45105, CVE-2021-45046 and CVE-2021-44228 (thanks to @sauterl)

Bugfix

• Fixed issue that caused locks to not be released properly if preparation of a query failed.
• Fixed issue with parallelisation of queries

Release 0.13.0

Query Language

• Query Language now support the evaluation of functions as part of a PROJECTION or a PREDICATE clause
• Removed kNN Predicate as a primary language construct. NNS can now be expressed as a combination of FUNCTION execution, ORDER BY and LIMIT. This allows for greater flexibility in terms of queries that can be executed.
• Removed MATCH operator.
• Transaction and query information is now provided in a dedicated Metadata object in every query request.

General

• Improved implementation of SimpleClient, which now provides access to most query language constructs
• TupleIterator now implements Closeable which aborts an ongoing query (no close is required if iterator is drained)
• Added notion of a FunctionRegistry that serves as a lookup facility for functions that can be executed as part of a query.
• Migrated all distance to proper, executable function implementations and adjusted query planner to support execution of NNS queries based on this new facility.

CLI

• There is now a kill command that can be used to abort ongoing transactions.

Bugfixes

• Various improvements related to transaction management and locking.