poliqarp (1.3.13) unstable; urgency=low

  * bpng:
    + In NKJP4TEI parsing, <text> elements without xml:id are no longer
      skipped. Thanks to Michał Lenart for the bug report.
  * Common code:
    + Polish translation of “unexpected character” error message was fixed.
      Thanks to Janusz S. Bień for the bug report.
  * Python interface:
    + set_query_flags() method was fixed.
  * Build system:
    + Support for OpenJDK 7 was added.
    + Error handling was improved.
    + A script to regenerate all autoconf files (devel/autogen) was
      implemented.
    + Enabling/disabling optional components in the configure script (a
      regression of Poliqarp 1.3.11) was fixed.
  * Debian package:
    + Detection of JREs in multi-arch directories was fixed.
    + Work-around for OpenJDK bug, which causes displaying only empty gray
      windows when used with non-re-parenting window managers was implemented.
    + Support for OpenJDK 7 was added.
    + debian/rules was rewritten from scratch, without using dh.
  * Documentation:
    + Protocol description in the poliqarpd manual page was updated:
      - “SET query-flags” command is now documented.
      - “CLOSE” → “CLOSE-CORPUS” rename is now documented.
    + “fullstoppedness” was added to the list of attributes in the NKJP
       tagset.

 -- Jakub Wilk <jwilk@jwilk.net>  Wed, 01 Aug 2012 00:50:23 +0200

poliqarp (1.3.12) unstable; urgency=low

  * bpng:
    + A crash when two documents are using a different set of subdocument
      types was fixed (a regression of Poliqarp 1.3.10). Thanks to Krzysztof
      Szafran for the bug report.
    + The corpus builder will now abort early (rather than produce a broken
      corpus) when a limit of 2.1G segments or 2.1G metadata items is reached.
  * Common code:
    + Error handling was improved (issue #143). Thanks to Carsten Schnober for
      the bug report.
    + Thread-safety was improved (issue #144).
  * Documentation:
    + It is now documented that very large corpora (more than 2.1G segments or
      2.1G metadata items) are not supported.
    + Performance issues related to the libxml2 library are now documented.

 -- Jakub Wilk <jwilk@jwilk.net>  Fri, 19 Aug 2011 18:57:46 +0200

poliqarp (1.3.11) unstable; urgency=low

  * bpng:
    + The -j/--parallel option was added, that allows parsing XML files in
      multiple threads of execution.
    + New XPath function is now available: strip-prefix.
  * bp:
    + Unknown options are not silently ignored anymore.
  * bpindexer:
    + Unknown options are not silently ignored anymore.
    + Error handling was improved.
  * Python interface:
    + Minor bug fixes.
  * Common code:
    + An off-by-one error in the metadata handling code, which caused
      crashes for certain corpora, was fixed.

 -- Jakub Wilk <jwilk@jwilk.net>  Mon, 18 Apr 2011 20:12:44 +0200

poliqarp (1.3.10) unstable; urgency=low

  * Query language:
    + Parsing of \' and \" escape sequences was fixed (a regression of
      Poliqarp 1.3.7).
    + Parsing of ' inside double-quoted strings was fixed.
  * poliqarpd:
    + Fix race condition between the query thread and the main thread.

 -- Jakub Wilk <jwilk@jwilk.net>  Fri, 30 Jul 2010 16:48:41 +0200

poliqarp (1.3.9) unstable; urgency=low

  * bpng:
    + Corpus builder has been rewritten from scratch and now supports new a
      TEI-based source format. The old corpus builder (bp) is still available.
  * bpindex:
    + Error handling was improved.
  * poliqarpc:
    + A bug in handling the -L option (a regression of Poliqarp 1.3.4) was
      fixed.
  * poliqarpd:
    + A race condition in locale handling was fixed.
  * Query language:
    + \xhh, \uhhhh and \Uhhhhhhhh escape sequences in strings are now
      accepted.
  * Python interface:
    + Many minor bug fixes.

 -- Jakub Wilk <jwilk@jwilk.net>  Thu, 20 May 2010 18:54:35 +0200

poliqarp (1.3.8) unstable; urgency=low

  * poliqarpd:
    + retrieve-ids session option was added.
    + Logging is a bit more verbose now.
    + Network performance bug was fixed.
    + SET-LOCALE command was added.
  * Python interface:
    + Many minor bug fixes.
  * Documentation:
    + Recommended query rewrite rules were fixed.
  * Build system:
    + Compatibility with glibc 2.10 was fixed.

 -- Jakub Wilk <jwilk@jwilk.net>  Sat, 28 Nov 2009 23:17:26 +0100

poliqarp (1.3.7) unstable; urgency=low

  * Python interface to poliqarpd was included.
  * poliqarpc:
    + Option for producing random samples was added.
    + Option for producing all results was added.
    + Memory usage was significantly reduced.
  * poliqarpd:
    + Support for producing random samples was added.
    + Logging on big-endian machines was fixed.
    + A memory leak was fixed.
    + For consistency reasons, some commands were renamed. Old names are still
      accepted, but their use is deprecated.
    + GET-NUM-RESULTS command was added.
  * Query language:
    + Unknown string escapes are now handled in a less confusing manner.
  * Common code:
    + Context is no longer allowed to cross document boundaries.
    + It is now possible to compile parts of Poliqarp without a threading
      library.
    + Support for producing random samples was added.
    + A bug which prevented from return strictly adjacent result was fixed.
    + A bug (introduced in Poliqarp 1.3.5) in the string handling library,
      which led to excessive memory usage, was fixed.
    + A memory leak was fixed.
  * Documentation:
    + Description of the statistical extension was included.

 -- Jakub Wilk <jwilk@jwilk.net>  Sat, 17 Oct 2009 12:46:43 +0200

poliqarp (1.3.6) unstable; urgency=low

  * bp:
    + A memory leak (introduced in Poliqarp 1.3.4) was fixed (issue #120).
    + The --log-file option was removed.
  * bpindexer:
    + Reading/writing CDF files is now done more carefully.
    + Error handling was improved.
  * poliqarpd:
    + Maximum number of sessions was increased to 256.
    + Maximum number of connections was increased to 128.
  * Query language:
    + All whitespace characters are now recognized as such.
    + Explicit query terminators were removed.
    + Dummy parser for statistical queries was added (issue #100).
  * Common code:
    + Reading/writing CDF files is now done more carefully.
    + Error handling was significantly improved.
    + Enums are not used in binary structures anymore.
    + Code to assure proper UTF-8 environment is more robust now.

 -- Jakub Wilk <jwilk@jwilk.net>  Mon, 01 Jun 2009 21:25:31 +0200

poliqarp (1.3.5) unstable; urgency=low

  * bpupgrade:
    + A major bug of bpupgrade was fixed (a regression of Poliqarp 1.3.4,
      issue #112).
  * Common code:
    + Error handling was improved.
    + Re-enable compilation on big-endian machines (a regression of Poliqarp
      1.3.1, issue #111).

 -- Jakub Wilk <jwilk@jwilk.net>  Tue, 17 Feb 2009 23:25:28 +0100

poliqarp (1.3.4) unstable; urgency=low

  * bp:
    + Access to uninitialized memory was fixed.
    + Current LC_COLLATE setting are now shown.
    + Building corpora with syntax annotations was fixed (a regression of
      Poliqarp 1.3.1, issue #108).
    + New routines to collate strings were provided.
  * poliqarpc:
    + It is now possible to use query rewriting (issue #94)
    + It is now possible to change the match buffer size.
    + It is now possible to run more than one query at once (issue #17).
    + Limitations on context size was lifted.
  * poliqarpd:
    + Ability to detach from console/terminal was added (issue #47).
    + Windows platform only: an interface to notify an other process that
      the server is ready to accept connections was added (issue #47).
    + Error handling was improved (issue #78).

  * Common code:
    + Access to uninitialized memory was fixed (issue #88).
    + Several memory leaks were fixed (issue #89, #90).
    + Error handling was significantly improved (issue #75, #91).
    + String hashing function was replaced with Paul Hsieh's one.
    + Floating-point arithmetic is no longer used (issue #85).
    + The fastest available memory unit is used for bitsets.
    + Code to assure proper UTF-8 environment was added. Check if regular
      expression engine is aware of UTF-8 was added. (issue #38, #39)
  * GUI:
    + Race condition in the Poliqarp launcher were fixed (issue #47).
  * Build system:
    + C99 mode was enabled.
    + Explicitly target JVM 1.5, not a later version (issue #81).
    + Generator of the Makefile.dep was rewritten.

 -- Jakub Wilk <jwilk@jwilk.net>  Wed, 11 Feb 2009 14:03:24 +0100

poliqarp (1.3.3) unstable; urgency=low

  * Parser was cleaned up (issue #27). See the NEWS file for details.
  * bp:
    + Pathname splitting was made more reliable (issue #37).
    + Having a path separator in a corpus basename is now explicitly
      disallowed.
  * poliqarpd:
    + Several memory leaks were fixed (issue #40).
  * Common code:
    + Regular expressions engine was re-imported from Tcl 8.5.3.
  * Query language:
    + SORT, BY, DIST and ONCE are no long reserved words (issue #68).
    + It is now possible to query for a space (issue #60). See the NEWS file
      for details.
  * GUI:
    + Change current working directory to $HOME at start (issue #66).
    + Default open/save filter are now set (issue #65).
    + JGoodies Looks in no longer embedded in the source package (issue #48).
    + JGoodies Forms are now used instead of non-free TableLayout (issue #71).
  * Common code:
    + Access to uninitialized memory was fixed.
    + Error handling was improved.
    + Section names in configuration files are now case-insensitive.
    + A threshold controlling whether to use indices was increased.
  * Build system:
    + Java detection was simplified.
    + Documentation is now included in the Windows package.
    + Handling for --enable-… options was fixed.

 -- Jakub Wilk <jwilk@jwilk.net>  Thu, 11 Dec 2008 21:39:11 +0100

poliqarp (1.3.2) unstable; urgency=low

  * poliqarpd:
    + It is now possible to pre-define corpora in the configuration file. (See
      corpus configuration option.)
    + It is now possible to restrict maximum match length. (See
      max-match-length configuration option.)
    + ‘log-file’ is now the preferred spelling of the ‘logfile’ configuration
      option.
    + Number of segments that are needed to be inspected are logged.
    + Hostname resolving was made more robust.
    + A few concurrency issues were fixed (issue #21, #24).
    + Error handling was improved (issue #33).
  * poliqarpc:
    + The -c, -m and -t options works now as documented (issue #19).
  * bpindexer:
    + Error handling was improved.
  * bpupgrade:
    + Verbosity was increased.
    + Windows platform only: a simple graphical interface was provided.
  * Query language:
    + Experimental support for variables in the query language was added
      (issue #28). See the NEWS file for details.
  * Common code:
    + Major bug preventing partial matches to overlap with matches, was fixed
      (issue #12).
    + An awkward step in the automata “determinization” was removed
      (issue #14).
    + Possible buffer overflow in the lexer was fixed (issue #30).
    + The lexer was made re-rentrant to avoid concurrency issues (issue #31).
    + A crash with syntax queries was fixed (regression of Poliqarp 1.3.1,
      issue #35).
    + A crash with syntax queries and a corpora without syntax annotations was
      fixed (issue #36).
    + As an compile-time option, it is now possible to memory-map whole copora
      images into memory. (Enabled by default on 64-bit machines.)
    + Indices are now used more eagerly.
    + Results with empty matches are not allowed anymore (issue #11).
    + Cache for string routines was turned off to avoid concurrency issues
      (issue #23).
    + Error handling was improved in several places.
    + Windows platform only:
      - String formatting was fixed in several places (issue #34).
      - File I/O was make 64-bit-clean.
      - pthreads library was upgraded to 2.8.0.
  * Build system:
    + Debian packaging was provided.
    + Manual pages were provided.
    + Files necessary to build a Windows installed were provided.
    + Generator of the Makefile.dep was included.
    + Generator of tags file was included.

 -- Jakub Wilk <jwilk@jwilk.net>  Mon, 24 Nov 2008 18:41:32 +0100

poliqarp (1.3.1) unstable; urgency=low

  * New binary format (isseu #4, #10). See the NEWS file for details.
  * bp:
    + Locale settings are respected.
    + Error handling was fixed.
  * bpindexer:
    + New name (was: indexer).
    + A regression of Poliqarp 1.2, which lead to infinite loops (and
      possibly broken indices), was fixed.
    + Gigabyte as a memory unit is allowed.
  * bpupgrade:
    + New tool to handle binary format upgrades.
  * poliqarpc:
    + Locale settings are respected.
  * poliqarpd:
    + By default, poliqarpd listens on 127.0.0.1 only. (See ‘hostname’
      configuration option)
    + A regression of Poliqarp 1.2, which disabled asynchronous results
      notifications, was fixed.
    + RFC-3339 timestamps are used in log files.
    + A crash if a peer disconnected was fixed.
    + A race condition in job cleanup code was fixed.
    + It's now possible to extract error message of an error during corpus
      opening.
    + SUSPEND-SESSION command, which is needed to avoid race condition in
      Poliqarp web clients, was added.
  * Common code:
    + Some error messages were made more meaningful.
  * Query language:
    + Precedence of & and | operators was fixed (issue #5).
    + Identifiers can contain non-leading ?, + and * characters.
  * Build system:
    + The build system was completely rewritten. Parallel builds are now possible.
      Binaries are put into bin/ subdirectory.
    + JAVA_HOME variable (rather than JDK) is inspected in order to determine
      find a Java compiler.

 -- Jakub Wilk <jwilk@jwilk.net>  Thu, 06 Nov 2008 20:03:16 +0100

poliqarp (1.2) unstable; urgency=low

  * New upstream version, packaged for Debian for the first time.
  * Major changes since 1.0:
    + New feature: statistical queries in the GUI;
    + New feature: support for syntactic queries;
    + New utility: poliqarpc, a simple command-line front-end for Poliqarp;
    + Improvement: Poliqarp now can be used on big-endian machines;
    + Improvement: bp is now more cross-platform;
    + Improvement: GUI now explains errors in queries in more detail;
    + Many bug fixes.

 -- Daniel Janus <nathell@korpus.pl>  Wed, 30 Jan 2008 09:42:52 +0100

