v0.5.49

What's Changed

• Create a new graph on POST to the Graph Store URL by @Qup42 in #2861
• Use IdTableView<0> in EvaluationContext and related interfaces by @joka921 in #2933
• Drop redundant member from TaskQueue by @RobinTF in #2947
• Don't show stripped columns in IndexScan descriptor by @ullingerc in #2941
• Fix GROUP BY optimizations for updates and materialized views by @RobinTF in #2913
• Properly resolve relative IRIs by @RobinTF in #2780
• Let OptionalJoin handle limit and offset by pushing to left child by @RobinTF in #2699
• Don't propagate limit and offset through explicit INTERNAL SORT BY by @RobinTF in #2974
• Drop accidental array wrap from SPARQL update response by @hannahbast in #2986
• Use unique temporary file names in CompressedRelationsTest by @Qup42 in #2950
• Add builtin function ql:toEpoch by @yarox-1 in #2841
• Remove deprecated parser/data/Iri.{h,cpp} by @marvin7122 in #2955
• Delegate management of index, cache, etc. from Server to Qlever class by @Damekt in #2965
• Fix use-after-stack-return in cheap compilation mode by @joka921 in #2993
• Replace const IdTable& by IdTableView<0> in several operations by @joka921 in #2999
• Add VariableCounter utility to count variable occurrences in a parsed query by @ullingerc in #2958
• Fix cmake compiler flags for ninja with clang by @ullingerc in #2987
• Add ZipMergeUniqueView for merging two sorted ranges by @Qup42 in #2975
• Log query start and end to new .metrics-log.jsonl file by @tanmay-9 in #2871
• Migrate join, path, and spatial-join call sites to Result::idTableView() by @joka921 in #3006
• Fix ZipMergeUniqueViewTest compilation on Clang-16 by @joka921 in #3014
• Migrate read-only call sites across the query engine to Result::idTableView() by @joka921 in #3018
• Introduce IndexAndViews and snapshot it per request by @RobinTF in #3019
• Add mode to HttpServer for lazy streaming of HTTP body by @joka921 in #2962
• Introduce BufferedWriteSerializer by @RobinTF in #2948
• Reduce code duplication in test/CMakeLists.txt by @RobinTF in #2946
• Bump more GitHub Actions to drop the Node 20 deprecation by @RobinTF in #2951
• Remove static keyword from free functions in headers by @RobinTF in #2994
• Bump libspatialjoin by @RobinTF in #2996
• Add construct-deduplication runtime parameter and corresponding DeduplicationMode by @marvin7122 in #2934
• Replace almost all remaining usages of Result::idTable by Result::idTableView<0> by @joka921 in #2932
• Add tryGet and insert to LRUCache by @marvin7122 in #2944
• Add IoUringManager and vocabulary batch-lookup result types by @marvin7122 in #3007
• Implement aligned serializers by @joka921 in #2791
• Don't throw on missing docs file in buildDocsDB by @RobinTF in #3002
• Faster permutation writer by @RobinTF in #3003
• Refactor BasicGraphPatternsInvariantTo using VariableCounter by @ullingerc in #3008
• Correctly set spatialjoin working directory and basename by @patrickbr in #3009
• Migrate test call sites from Result::idTable() to Result::idTableView() by @joka921 in #3030
• Fix CMake config when jemalloc is manually installed by @RobinTF in #3033
• Move added VALUES clause to the end of the SERVICE body by @RobinTF in #2912
• Add ql:simplifyGeometry function by @ullingerc in #3015
• Change ToEpoch logic to use milliseconds instead of nanoseconds by @yarox-1 in #3000
• Remove Result::idTable() by @joka921 in #3029
• Accept any InputRangeTypeErased for InputFileSpecification, not just std::vector by @joka921 in #2961
• Unify QueryPlan and PlannedQuery by @Damekt in #3011
• Replace ParallelBuffer with AsyncBlockSource realized via boost::asio by @joka921 in #3023
• Implement SortedSequence, which will replace std::set in LocatedTriples by @Qup42 in #3024
• Make cleanup of temporary files in IndexRebuilderTest more robust by @RobinTF in #3034
• Use a TaskQueue instead of std::async for writing the partial vocabularies by @RobinTF in #2949
• Fix use-after-free for LocalVocabIndex Ids in CONSTRUCT export by @hannahbast in #3025
• Move the implementation of the Join operation to JoinImpl by @MarkVeser in #3020
• Simplify IndexMetaData implementation and move away from mmap by @RobinTF in #2971
• Replace BatchedPipeline class with thread pools by @RobinTF in #2956
• Fix caching and cloning of non-deterministic operations like RAND() or BNODE() by @RobinTF in #2798
• Find statement boundaries by scanning blocks from the back by @RobinTF in #2959
• Fix wrong results and assertion errors for certain expressions in GROUP BY by @RobinTF in #2966
• Move query planning from Server class into Qlever class by @Damekt in #3038
• Save and restore the global logging stream in tests by @Qup42 in #3054

v0.5.48

You find the official QLever packages for Debian, Ubuntu, Ubuntu-derived distributions, and macOS at https://docs.qlever.dev/quickstart/

What's Changed

• During index build, free memory from parsing stage before vocabulary merging stage by @joka921 in #2843
• Complete the refactoring of the CONSTRUCT evaluation code by @marvin7122 in #2837
• Add started-at field to result from cmd=dump-active-queries by @tanmay-9 in #2836
• Reduce allocations in index builder and simplify vocabulary merger code by @RobinTF in #2839
• Check that rebuild-index never writes outside of original directory and never overwrites by @RobinTF in #2860
• Fix another use-after-free, this time in MessageSender by @hannahbast in #2869
• Fix reported buffer size in RDF parser error message by @MikiDi in #2382
• Minor cleanup of #2823 by @marvin7122 in #2840
• Make IndexTest.updateInputFileSpecificationsAndLog pass at LOGLEVEL=WARN by @Qup42 in #2848
• Properly catch exceptions during rebuild-index by @RobinTF in #2852
• Fix assertion error when building an empty text index by @RobinTF in #2799
• Don't precompute sort keys for the partial vocabularies by @RobinTF in #2856
• Treat FetchContent dependencies as SYSTEM and drop the GCC 12/13 -Wno-error workarounds by @joka921 in #2873
• Reduce in-memory buffer for idTriples during partial-vocab building by @RobinTF in #2857
• AD_LOG_DEBUG the actually-used num-triples-per-batch by @RobinTF in #2858
• Avoid redundant Sort operation in query plan for EXISTS by @RobinTF in #2874
• Support path{n}, path{n,m}, path{n,}, path{,n} in SPARQL property paths by @SvenPVoigt in #2240
• Update clang-format from 16 to 18 by @RobinTF in #2880
• Avoid a GCC 12 -Werror=array-bounds false positive in ExplicitIdTableOperationTest by @RobinTF in #2888
• Fix bug in unifying local vocab IDs when building a partial vocabulary by @RobinTF in #2891
• Fix YEAR(?date) prefilter when value is non-numeric or non-integral by @RobinTF in #2882
• Store Index and MaterializedViewsManager via shared_ptr by @RobinTF in #2892
• Fix assertion error in IndexScan when joining materialized views by @ullingerc in #2898
• Fix bug in optimized GROUP BY with COUNT by @RobinTF in #2901
• For cmd=dump-active-queries, return {} instead of null when no active queries by @RobinTF in #2903
• Turn vocabulary-order check during index build into an assertion by @RobinTF in #2894
• Fix handling of prefix regexes with character class escapes by @hannahbast in #2926
• Use 7 chars instead of 6 for git hash by @RobinTF in #2914
• Add addition for dates and durations by @yarox-1 in #2814
• Make logging thread-safe and the log level runtime-configurable by @joka921 in #2834
• Propagate limit and offset through Sort operation by @RobinTF in #2797
• Fix error for HTTP response with empty body by @Qup42 in #2870
• Make tests runnable in parallel again by @RobinTF in #2876
• Run rebuild-index independently of the query memory limit by @RobinTF in #2893
• Make lazy GROUP BY as fast as non-lazy GROUP BY by @RobinTF in #2921
• Fix copy-paste error in CMakeLists.txt by @S7uX in #2945
• Make getBlocksForJoin linear-time by @joka921 in #2928
• Bump Docker GitHub Actions to drop the Node 20 deprecation by @RobinTF in #2927
• Read on-disk vocabulary via pread instead of mmap by @RobinTF in #2931

v0.5.47

You find the official QLever packages for Debian, Ubuntu, Ubuntu-derived distributions, and macOS at https://docs.qlever.dev/quickstart/

What's Changed

• Remove the globalSingletonIndex_ static pointer from IndexImpl by @RobinTF in #2818
• Compute diff between two snapshots and remap IDs to the new index by @RobinTF in #2822
• Set permutationWriterNumThreads to 5 for index build by @hannahbast in #2833
• Simplify mergeVocabulary phase and remove memory spike by @RobinTF in #2827
• Make WordWriterBase::finishWasCalled_ atomic by @RobinTF in #2828
• Add GraphNameManager for managing unique graph IRIs for Graph Store Protocol by @Qup42 in #2744
• Remove two files that were completely unused by @RobinTF in #2830
• Refactor triple instantiation for CONSTRUCT queries into dedicated files by @marvin7122 in #2824
• Make the RDF parsers work for arbitrary input streams by @joka921 in #2838
• Add ql:has-word triples if option --add-has-word-triples by @hannahbast in #2579
• Add EnumWithStrings base class and refactor VocabularyType to use it by @joka921 in #2835

v0.5.46

You find the official QLever packages for Debian, Ubuntu, Ubuntu-derived distributions, and macOS at https://docs.qlever.dev/quickstart/

What's Changed

• Preparations for rewriting queries to use materialized views by @ullingerc in #2692
• Fix wrong size estimate for full scan of a materialized view by @ullingerc in #2707
• Allow creating materialized views as part of index build by @ullingerc in #2693
• Don't deduplicate rows implicitly when scanning materialized views by @ullingerc in #2708
• Re-enable the clang 21 build by @RobinTF in #2734
• Rewrite simple chain join to scan of materialized view if possible by @ullingerc in #2649
• Add template preprocessing phase for CONSTRUCT queries by @marvin7122 in #2710
• Add workflow for native build on macOS using AppleClang by @tanmay-9 in #2630
• Simpler lifetime management for the bug fix from #2684 by @RobinTF in #2690
• Encode literals from STRDT as ValueId by @ullingerc in #2713
• Fix sparql-results+json for solutions where all variables are unbound by @RobinTF in #2715
• Fix handling of UNDEF values in PrefixRegexExpression by @RobinTF in #2714
• Ensure that tests can be compiled and run on all LOGLEVELs by @Qup42 in #2719
• Fix order for Cartesian product join when UNDEF in left input by @RobinTF in #2726
• Add runtime parameter service-allowed-iri-prefixes to specify whitelist for SERVICE calls by @ullingerc in #2728
• Fix nested LIMIT and OFFSET and redundant StripColumns operation by @RobinTF in #2741
• Fix regression involving useGraphPostProcessing_ by @ullingerc in #2745
• Make destructor of CallbackOnEndView exception-safe by @joka921 in #2533
• Change log level of "No timeout check" messages from WARN to DEBUG by @hannahbast in #1635
• Normalize away the xsd:string datatype by @RobinTF in #2732
• Prove UndefStatus::AlwaysDefined for more expressions by @RobinTF in #2733
• Fix division by zero for IN with RHS with more than 1000 elements by @RobinTF in #2747
• Update most GitHub workflows to Ubuntu 24.04 by @joka921 in #2743
• String produced by GROUP_CONCAT should never have a language tag by @RobinTF in #2751
• Let Permutation know to which MaterializedView it belongs, it any by @ullingerc in #2753
• Add ql:envelopeLowerLeft and ql:envelopeUpperRight helper functions by @ullingerc in #2754
• For each BIND expression in a materialized view query, remember the column by @ullingerc in #2756
• Distinguish between "cancelled" and "failed" status for lazy query results by @RobinTF in #2628
• Add subtraction for dates by @yarox-1 in #2645
• Remove outdated documentation and refer to https://docs.qlever.dev by @ullingerc in #2631
• Allow to completely bypass QLever's cache by @joka921 in #2758
• Faster compilation with type-erased StringExpression by @joka921 in #2735
• Fix the compilation with CHEAPER_COMPILATION=ON by @joka921 in #2767
• Implement Graph Store Protocol operation HEAD, and fix status codes for PUT and DELETE by @Qup42 in #2532
• Fix the sonarcloud uploader by @joka921 in #2774
• Support redirects in SERVICE requests by @hannahbast in #2435
• Compute BIND column via materialized view if possible by @ullingerc in #2742
• Store UndefStatus for each column of a materialized view by @ullingerc in #2772
• Fix spurious test failures in IndexTest (hopefully) by @joka921 in #2775
• Add non-owning variants of LiteralOrIri, Iri, and Literal by @joka921 in #2736
• Improve the performance of rebuild-index and of the graph metadata update by @RobinTF in #2738
• Add batch evaluation of variables for CONSTRUCT queries by @marvin7122 in #2723
• Add operation vacuum-delta-triples that removes redundant update triples by @Qup42 in #2737
• Rewrite star join to scan on materialized view if possible by @ullingerc in #2776
• Let SpatialJoin use bounding boxes from an underlying materialized view by @ullingerc in #2769
• Add internal function ql:prefix-match by @RobinTF in #2755
• Extend IndexScan descriptor to include all scanned columns by @ullingerc in #2790
• Fix rewrite of chain join to materialized view for edge case (subject fixed and not sorted by subject) by @ullingerc in #2789
• Rename --parse-parallel option of qlever-index to --parallel-parsing by @nbdani in #2796
• Reduce cyclic dependency between index and engine libraries by @marvin7122 in #2765
• Index nested loop join for MINUS and EXISTS when left side is larger by @RobinTF in #2312
• Clone subtrees before applying LIMIT and OFFSET by @RobinTF in #2782
• Refactor graph search in TransitivePath operation by @schaetzr in #2730
• Finish the migration to C++17 by @joka921 in #2786
• Add tryGet method to LRUCache and LRUCacheWithStatistics by @marvin7122 in #2763
• Update GitHub Actions to avoid Node.js 20 deprecation issues by @RobinTF in #2811
• Add magic service ql:external-values by @joka921 in #2778
• Implement remaining cases of date and duration subtraction by @yarox-1 in #2784
• Make IdTable cheaper to move, and IdTableView cheaper to create by @joka921 in #2759
• Remove redundant globalSingletonComparator_ by @RobinTF in #2801
• Change interface of toValueId to take const IndexImpl& by @RobinTF in #2808
• Make disable-caching runtime parameter accessible via HTTP by @hannahbast in #2809
• Fix four independent "use-after-free" bugs discovered with ASAN by @joka921 in #2812
• Fix three build failures in the CPP17/GCC 8 and AppleClang workflows by @RobinTF in #2815
• Distinguish between IdOrLiteralOrIri and IdOrLocalVocabEntry by @RobinTF in #2813
• Fix two more lifetime issues (one ASAN and one SEGV) by @hannahbast in #2816
• Drop redundant detail:: qualifiers in ParallelMultiwayMerge.h by @marvin7122 in #2823

v0.5.45

There are now official QLever packages for Debian, Ubuntu, Ubuntu-derived distributions, and macOS; see https://docs.qlever.dev/quickstart/

What's Changed

• Allow writing materialized views from queries with less than four columns by @ullingerc in #2639
• Avoid hashing twice in BlankNodeManager::allocateBlock by @RobinTF in #2648
• When writing a materialized view, remember the query string by @ullingerc in #2653
• Show name of materialized view in IndexScan descriptor by @ullingerc in #2655
• Ensure that tests also compile and run for LOGLEVEL < WARN by @Qup42 in #2671
• Unload materialized view when overwritten by @ullingerc in #2657
• Use the new binary names qlever-index and qlever-server in all Dockerfiles by @ullingerc in #2677
• Load materialized views when starting the server by @ullingerc in #2659
• Native WebAssembly aka Wasm build with Emscripten by @ZamirCS-369 in #2577
• Fix blank node label consistency in parallel parsing by @joka921 in #2676
• Preparations for binary index rebuild by @RobinTF in #2660
• Fix destruction order of transformation_ and view_ in CachingTransformInputRange by @tanmay-9 in #2682
• Refactor code for the evaluation of CONSTRUCT queries by @marvin7122 in #2652
• More efficient size estimate for IndexScan, in particular for a full index scan by @RobinTF in #2680
• Enable fast "binary" rebuilding of the index, including updates by @RobinTF in #2408
• Lazy and prefiltered OPTIONAL by @joka921 in #2695
• Prove UndefStatus::AlwaysDefined for a variety of expressions by @joka921 in #2698
• Fix division by zero bug for FILTER(... IN ()) by @RobinTF in #2702
• Insert ql:langtag triple when inserting triple with an object with a language tag by @RobinTF in #2674

v0.5.44

There are now official QLever packages for Debian, Ubuntu, Ubuntu-derived distributions, and macOS; see https://docs.qlever.dev/quickstart/

Version for the Autocompletion Paper

Please use this version only for reproducing the results of our Autocompletion Paper.
In this case, please consider the Caveats from the README.md
For all other uses, a stable version from the master branch is probably better suited.