Version 2.0.1-beta, 22 apr 2011

New general features

New SphinxQL features

New command-line switches

  • added --print-queries switch to indexer that dumps SQL queries it runs
  • added --sighup-each switch to indexer that rotates indexes one by one
  • added --strip-path switch to searchd that skips file paths embedded in the index(-es)
  • added --dumpconfig switch to indextool that dumps an index header in sphinx.conf format

Major changes and optimizations

  • changed default preopen_indexes value to 1
  • optimized English stemmer (results in 1.3x faster snippets and indexing with morphology=stem_en)
  • optimized snippets, 1.6x general speedup
  • optimized const-list parsing in SphinxQL
  • optimized full-document highlighting CPU/RAM use
  • optimized binlog replay (improved performance on K-list update)

Bug fixes

  • fixed #767, joined fields vs ODBC sources
  • fixed #757, wordforms shared by indexes with different settings
  • fixed #733, loading of indexes in formats prior to v.14
  • fixed #763, occasional snippets failures
  • fixed #648, occasionally missed rotations on multiple SIGHUPs
  • fixed #750, an RT segment merge leading to false positives and/or crashes in some cases
  • fixed #755, zones in snippets output
  • fixed #754, stopwords counting at snippet passage generation
  • fixed #723, fork/prefork index rotation in children processes
  • fixed #696, freeze on zero threshold in quorum operator
  • fixed #732, query escaping in SphinxSE
  • fixed #739, occasional crashes in MT mode on result set send
  • fixed #746, crash with a named list in SphinxQL option
  • fixed #674, AVG vs group order
  • fixed #734, occasional crashes attempting to report NULL errors
  • fixed #829, tail hits within field position modifier
  • fixed #712, missing query_mode, force_all_words snippet option defaults in Java API
  • fixed #721, added dupe removal on RT batch INSERT/REPLACE
  • fixed #720, potential extraneous highlighting after a blended keyword
  • fixed #702, exceptions vs star search
  • fixed #666, ext2 query grouping vs exceptions
  • fixed #688, WITHIN GROUP ORDER BY related crash
  • fixed #660, multi-queue batches vs dist_threads
  • fixed #678, crash on dict=keywords vs xmlpipe vs min_prefix_len
  • fixed #596, ECHILD vs scripted configs
  • fixed #653, dependency in expression, sorting, grouping
  • fixed #661, concurrent distributed searches vs workers=threads
  • fixed #646, crash on status query via UNIX socket
  • fixed #589, libexpat.dll missing from some Win32 build types
  • fixed #574, quorum match order
  • fixed multiple documentation issues (#372, #483, #495, #601, #623, #632, #654)
  • fixed that ondisk_dict did not affect RT indexes
  • fixed that string attributes check in indextool --check was erroneously sensitive to string data order
  • fixed a rare crash when using BEFORE operator
  • fixed an issue with multiforms vs BuildKeywords()
  • fixed an edge case in OR operator (emitted wrong hits order sometimes)
  • fixed aliasing in docinfo accessors that lead to very rare crashes and/or missing results
  • fixed a syntax error on a short token at the end of a query
  • fixed id64 filtering and performance degradation with range filters
  • fixed missing rankers in libsphinxclient
  • fixed missing SPH04 ranker in SphinxSE
  • fixed column names in sql_attr_multi sample (works with example.sql now)
  • fixed an issue with distributed local+remote setup vs aggregate functions
  • fixed case sensitive columns names in RT indexes
  • fixed a crash vs strings from multiple indexes in result set
  • fixed blended keywords vs snippets
  • fixed secure_connection vs MySQL protocol vs MySQL.NET connector
  • fixed that Python API did not works with Python 2.3
  • fixed overshort_step vs snippets
  • fixed keyword staistics vs dist_threads searching
  • fixed multiforms vs query parsing (vs quorum)
  • fixed missed quorum words vs RT segments
  • fixed blended keywords occasionally skipping extra character when querying (eg “abc[]”)
  • fixed Python API to handle int32 values
  • fixed prefix and infix indexing of joined fields
  • fixed MVA ranged query
  • fixed missing blended state reset on document boundary
  • fixed a crash on missing index while replaying binlog
  • fixed an error message on filter values overrun
  • fixed passage duplication in snippets in weight_order mode
  • fixed select clauses over 1K vs remote agents
  • fixed overshort accounting vs soft-whitespace tokens
  • fixed rotation vs workers=threads
  • fixed schema issues vs distributed indexes
  • fixed blended-escaped sequence parsing issue
  • fixed MySQL IN clause (values order etc)
  • fixed that post_index did not execute when 0 documents were succesfully indexed
  • fixed field position limit vs many hits
  • fixed that joined fields missed an end marker at field end
  • fixed that xxx_step settings were missing from .sph index header
  • fixed libsphinxclient missing request cleanup in sphinx_query() (eg after network errors)
  • fixed that index_weights were ignored when grouping
  • fixed multi wordforms vs blend_chars
  • fixed broken MVA output in SphinxQL
  • fixed a few RT leaks
  • fixed an issue with RT string storage going missing
  • fixed an issue with repeated queries vs dist_threads
  • fixed an issue with string attributes vs buffer overrun in SphinxQL
  • fixed unexpected character data warnings within ignored xmlpipe tags
  • fixed a crash in snippets with NEAR syntax query
  • fixed passage duplication in snippets
  • fixed libsphinxclient SIGPIPE handling
  • fixed libsphinxclient vs VS2003 compiler bug
News
Downloads
Services
Community
Resources
About