Changelog for strus-0.15.7.3-1.2.x86_64.rpm :

* Thu Mar 15 2018 Patrick Frey 0.15.7-0.1- added make target cppcheck for static code analysis- adapted changes in error handling defined in strus base: simplified error codes- improve some error messages- added skipPrefix (skip to first element with a prefix) in valueIteratorInterface- fix some smaller build issues
* Mon Feb 12 2018 Patrick Frey 0.15.6-0.1- added some functions for introspection needed by the webservice- adapted error handling to changes in base: mandatory error codes everywhere
* Fri Dec 29 2017 Patrick Frey 0.15.5-0.1- fixed several build issues: Build with Clang, Fix bugs in travis build
* Mon Dec 04 2017 Patrick Frey 0.15.4-0.1- changed paramater renaming option name in summarizer \'metadata\' and \'attribute\'- added -Wshadow for C++ build
* Fri Oct 13 2017 Patrick Frey 0.15.3-0.1- RPath fixes and some CMake file reorganization- fixed missing check of meta data element names, causing a logic error in the commit
* Wed Sep 27 2017 Patrick Frey 0.15.2-0.1- bring up travis build, some minor fixes
* Mon Sep 04 2017 Patrick Frey 0.15.1-0.1- release to cleanup mess with missing CHANGELOG in 0.15.0
* Sun Sep 03 2017 Patrick Frey 0.15.0-0.1- remove hardcoded setting of C++98 in CMake files- added method close() to vector storage client- fixed SEGFAULT under some circumstances after StorageClient::close hss been called- implemented posting join operator \'sequence_imm\'- some method renamings- added ACL reader interface and method to create it in storage client interface- fixed bug in user access restriction handling (storing ACLs)- fixed bug in docnoAllMatchItr: endless loop under some circumstances- added weighting function \'scalar\' that calculates a scalar function on constants, variables and document metadata defined as string- many CMake build fixes
* Wed Feb 15 2017 Patrick Frey 0.14.0-0.1- added possibility to precalculate weights in weighting function \'constant\'- update for terms in forward index and search index- explicit specification of df for features (for example to sopecify df for forward index features)
* Mon Feb 06 2017 Patrick Frey 0.13.0-0.1- improve standard summarization and weighting functions- renamings of summarizer and weighting functions- new summarizer \'accunear\' that accumulates forward index features weighted by near matches of query features- improved vector interfaces- fixed some bugs in weighting and summarization- many build fixes
* Sat Oct 01 2016 Patrick Frey 0.12.0-0.1- increment version parallel to analyzer (some interface files reorganization)
* Fri Sep 30 2016 Patrick Frey 0.11.1-0.1- moved vector space model implementation into own project strusvector
* Wed Sep 28 2016 Patrick Frey 0.11.0-0.1- integrated the armadillo library for fast linear algebra operations- define vector space model interface and implement standard vector space model for unsupervised learning of features from vectors
* Tue Jul 05 2016 Patrick Frey 0.10.0-0.1- added support for cardinality in posting iterator intersect- many build fixes
* Thu Jun 09 2016 Patrick Frey 0.9.2-0.1- adapted license change of CompactNodeTrie
* Wed May 18 2016 Patrick Frey 0.9.1-0.1- fixed wrong counting of documents when deleting documents
* Thu Apr 28 2016 Patrick Frey 0.9.0-0.1- changes in the object builder interfaces
* Sun Apr 24 2016 Patrick Frey 0.8.3-0.1- put some restricting constants of weighting functions and summarizers into header files
* Fri Apr 15 2016 Patrick Frey 0.8.2-0.1- moved numeric variant from strus to strus base
* Thu Apr 14 2016 Patrick Frey 0.8.1-0.1- package fix: remove deployment instructions for files removed from this project
* Wed Apr 13 2016 Patrick Frey 0.8.0-0.1- moved libstrus_utils (no named libstrus_base) and libstrus_error to new project strusbase (strus depends now on strusbase)
* Fri Apr 01 2016 Patrick Frey 0.7.4-0.1- removed deprecated formula weighting function- added weighting function \'Smart\' that allows to specify weighting with a scalar function expression on ff,df,N (collection size) and document metadata
* Thu Mar 31 2016 Patrick Frey 0.7.3-0.1- fixed missing dependency to libstrus_scalarproc
* Thu Mar 31 2016 Patrick Frey 0.7.2-0.1- fixed missing dependency to libstrus_scalarproc
* Wed Mar 30 2016 Patrick Frey 0.7.1-0.1- fixed bug in scalar functions in 0.7.0
* Wed Mar 30 2016 Patrick Frey 0.7.0-0.1- introducing scalar function objects used for combining ranking functions or to pass functions to weighting functions or summarizers
* Wed Mar 23 2016 Patrick Frey 0.6.1-0.1- improved BM25pff weighting function: Two more parameters for a proximity ff cap for feature occurrencies not fulfilling minimal proximity requirements
* Mon Mar 21 2016 Patrick Frey 0.6.0-0.1- changed license from GPLv3 to MPLv2 (Mozilla Public License, Version 2)
* Wed Mar 16 2016 Patrick Frey 0.5.3-0.1- some more normalization factors configurable in BM25pff- some small fixes in BM25 and Formula weighting method
* Mon Mar 07 2016 Patrick Frey 0.5.2-0.1- more fixes for bringing up packages again
* Mon Mar 07 2016 Patrick Frey 0.5.1-0.1- small fixes for bringing up packages again
* Sun Mar 06 2016 Patrick Frey 0.5.0-0.1- add upperbound seek with upper key restriction to database interface (not used yet in storage)- change return value of document term iterator skipDoc from boolean to document number- more helpers for posting iterator implementations (just a start) and reorganize code in posting join operator iterators (eliminate copy paste code)- changed logarithm base in weighting from e to 10, fixed summarizer metadata- fixed metadata summarizer: could not extract first metadata element- rewrote posting set join of the \'contains\', \'sequence\' and \'within\' operators- fixed bug in \'randoc\' test: cases for within were not correctly generated in case of position conflicts in strict case- implemented browse posting iterator with metadata restriction for browsing documents matching a boolean condition without query evaluation involved- changed query interface: result is now a structure containing also some statistical information like the number of matching documents and not only a vector of ranks- use unordered_map instead of compact node trie for keymaps. They are about 30% slower, but the problem with memory limits and an overflow map, that eats up any performance gain, does not appear (much better)- do not read keymaps during insertion (performance gain in bulk insert of factors)- added new weighting function BM25pff (BM25 with proximity based ff)
* Tue Feb 09 2016 Patrick Frey 0.4.3-0.1- changed return value of weighting function from float to double
* Sun Jan 17 2016 Patrick Frey 0.4.2-0.1- fix build errors
* Sun Jan 17 2016 Patrick Frey 0.4.1-0.1- change default forward index block size to 128- fix typos in src license reference
* Sat Jan 02 2016 Patrick Frey 0.4.0-0.1- changed posting iterator interface: new method skipDocCandidate that does what skipDoc was doing before and method skipDoc that always delivers a real match and not only a candidate, removed obsolete method subExpressions
* Fri Jan 01 2016 Patrick Frey 0.3.5-0.1- fixed bug in Reference::reset()- changed type of document frequency from GlobalCounter to Index- fixed new segfault in case of no statistics populated
* Thu Dec 24 2015 Patrick Frey 0.3.4-0.1- fixed bug in transaction: statistics number of documents (collection size) was not set correctly
* Thu Dec 24 2015 Patrick Frey 0.3.3-0.1- fixed bug in statistics viewer: wrong decoding of global number of documents (collection size)
* Wed Dec 23 2015 Patrick Frey 0.3.2-0.1- fixing bug in metadata cache layout: was completely broken
* Mon Dec 21 2015 Patrick Frey 0.3.1-0.1- fixing versioning problems of the Open Suse build cluster
* Fri Dec 18 2015 Patrick Frey 0.3.0-0.1- better introspection for functions- removed docno allocator interface and fdatasync on every new document insert when not using the docno allocator
* Tue Dec 01 2015 Patrick Frey 0.2.6-0.1- added interfaces for introspection (user names, docids, term types, term values)
* Tue Dec 01 2015 Patrick Frey 0.2.5-0.1- define method PeerMessageQueue start for register/deregister of peers- change way how peer message processor is passed to storage: Now an argument of createClient and an argument of StorageClient::createPeerMessageQueue- reimplemented peer message processing interface completely (because we are still prealpha only a patch version number increment)- fixed: throw error instead of segfaulting when peer message methods called without peer message processor defined- fixed bug in storage transaction: cleanup of forward index not complete when only documents deleted in a transaction
* Thu Nov 26 2015 Patrick Frey 0.2.4-0.1- incremented version to 0.2.4 to resolve OBS trouble with dependencies
* Tue Nov 24 2015 Patrick Frey 0.2.3-0.1- ABI change. StorageClient::documentNumber( const std::string&) returns 0, if document does not exist
* Tue Nov 24 2015 Patrick Frey 0.2.2-0.1- more info in corrupt index error message.- better summarization for first sentences in summarizer match phrase (abstracting)- fix bug in summarizer: if end of sentence was not existing in a document, then the end of sentence of the previous document was taken- packaging fixes: fixed a type in 12.04 DSC file, some date fixes in redhat spec file, cmake patches the version now, also adapted the changelogs- fixed error in summarizer match variables and accumulate variables: fetch from forward index only if element exists
* Sun Nov 15 2015 Patrick Frey 0.2.1-0.1- added posting join operator \"inrange\" (like \"within\", but not strict, different positions)- added new summarizer \"accuvariable\" to accumulate weights of variable references- remove weighting method \'BM23_dpfc\' from core (moved to module in wikipedia demo project)
* Thu Nov 12 2015 Patrick Frey 0.2.0-0.1- make ranked document list deterministic (docno is 2nd sort criterion)- removed \'pushDuplicate\' method in queryInterface- fixed bug in parsing helper method: string was not extracted correctly in extractStringFromConfigString- make aligned malloc more portable- make debug output of weighting formula weighting scheme more readable- renamed fixedStructAllocator to fixedSizeStructAllocator- weighting also weight of title in BM23_dpfc- better error messages for mixing feature parameters with numeric or string values in summarizers and weighting functions- fixed missing default initialization of max nof ranks in query- fixed bug in accumulator: Specifying an evaluation set with Query::addDocumentEvaluationSet could cause an endless loop- fixed bug in metadata record (unsigned integers of 8 and 16 bit were not assigned correctly (mapped to an integer)- fixed bug in forward index iterator: return illegal chunks- remove hardcoded doclen meta data element (has to be explicitely defined now with an aggregator)- fixed description of summarizer match phrase- fixed missing initialization of error handler in forward iterator (causing a segfault)- fixed bug in new string maps used in storage: calling C++ delete of memory block allocated with C malloc- better error messages for formula interpreter in formula weighting method- improve visibility of error position in messages for formula weighting scheme- better error message in formula weighting scheme- added all boost libraries when linking in randoc test not only boost_thread (because of missing boost_atomic)- added a missing boost_system in testLoggingMalloc- added a missing strus_private_utils library dependency for formula test- strus_database_leveldb depends on leveldb only not on snappy too- updated OBS package build support- fixed seek of alternative matches in struct join operators- fixed some small bugs in reversed struct withing- make backwards sequence work- added join operator chain and struct_chain, that do not require a strict ordering of the member elements- added weight as parameter to summarizer- make summarizer \'matchvariables\' work- added methods for retrieving description of functions- implement printing of floats and doubles with precision in strus_snprintf- reorganise \'contains\' iterator join- added new expression parameter \'cardinality\'- do not checkin cmake generated files- (origin/memfrag, memfrag) fix segfault (error in last checkins)- (origin) remove string maps also from forward index- replace STL string map with StringMap- replace map of strings in transaction and peer storage transaction by new StringMap, for better memory management, prevent heap fragmentation- fixing missing dependency to private utils in logging malloc library test- added new library that logs malloc call statistics (to be used with LD_PRELOAD)- added error buffer method to change the maximum number of threads- set default max nof threads default in error buffer- delete all objects passed with ownership to functions that have no return value. caller cannot always determine if function was successful- make createStorage return bool (success/failure)- make error buffer run in multithreaded environment- exception free interfaces (also change guidelines)- fixed missing linking to private_utils static library in weighing- revert visibility of linked symbols after merge with master- fix a bug in ranker: segfault when retrieving more than 128 results. Do not build summarizers if result is empty- added weighting function defined by formula specified as string- added test for formula parser- added formula interpreter for weighting functions with formula defined from string- copy variable assignments too with pushDuplicate- fixed bad memory management in query (came in with last checkin)- implicitely define order of elements in subset query- add possibility to evaluate query on a subset of documents. This allows us to make a multi pass query evaluation like evaluating a query on the N best results of a BM25 query- added an unreliable malloc library that fails every N-th malloc (to use with LD_PRELOAD) for testing - fixed typos and updated documentation- adapted dependency description in online documentation
* Tue Aug 18 2015 Patrick Frey v0.1.6-0.1- increased default maximum key size in keymap cache- updated wegpage with language logos- updated landing page and documentation- throw is there is no selection feature defined in the query- test also random order insert in peer message processor- redesigned storage peer message processing completely. You do not pass an interface to a messaging, but just an interface for packing and unpacking messages. The peer message processing is passive: You fetch messages that have to be sent to peers and you push messages from peers- added flags to peer message processor- added start/rollback to peer message builder- added library for packing unpacking of peer storage messages- put common CMAKE stuff of all strus relaed projects into includes- some small storage peer interface changes and documentation- some changes in peer storage interface
* Sun Aug 02 2015 Patrick Frey v0.1.5-0.1- fixed bug in expression evaluation in query.cpp (causing UMR)- updated utilities documentation
* Fri Jul 24 2015 Patrick Frey v0.1.4-0.1- added dependency to libboost_atomic needed by ARM for boost 1.54 (see https://svn.boost.org/trac/boost/ticket/8810)
* Tue Jul 14 2015 Patrick Frey v0.1.3-0.1- skip entries starting with \'.\' in read files/directories (fileio.cpp)- fixed segfault (due to latest changes)- incremented patch version number- extended fileio with function to read directory only (assumed in programs that directories do not have a file extension - HACK)- some documentation adapted
* Thu Jul 09 2015 Patrick Frey v0.1.2-0.1- added exists check to database interface- reverted create option (latest changes) because it is bullshit- added boolean option \'create\' to database client configuration: true=create database implicitely if it does not exist yet
* Fri Jul 03 2015 Patrick Frey v0.1.1-0.1- Initial revision