xref: /dports/deskutils/pinot/pinot-1.20/NEWS

2021/10/16 version_1_2_0
General :
 - dropped xdgmime in favour of gio
 - require giomm >= 2.6
 - run rst2txt on RST files, if available
IndexSearch :
 - removed obsolete search plugins, dead code
 - added a plugin for Arxiv
 - better results with the dir query filter
 - fixed size range queries
 - synced with how Xapian Omega indexes file extensions
 - added sort by file size (descending)
 - pinot-index shows document IDs to facilitate troubleshooting
 with xapian-delve
 - require Xapian >= 1.4.10
 - reindexing is recommended.
Daemon :
 - fixed a bug that would cause the loss of documents for one directory
 when crawling ends for the next one
 - overhauled the D-Bus implementation, extended the interface
 - dropped dependency on dbus-glib
 - reduced flushing
 - a IndexFlushEpoch property indicates the last time index changed
 were flushed to disk
 - rely on org.freedesktop.UPower's OnBattery on systems with a battery
 - implemented the org.gnome.Shell.SearchProvider2 interface. The content
 of documents in directories indexed by the daemon can be searched through
 the Gnome 3 Shell.
UI :
 - removed the gtk2 UI, refreshed the gtk3 UI
 - the gtk3 UI is loaded from a Glade XML file
 - reduced direct reading of the daemon's index
 - reopen the daemon's index as per the IndexFlushEpoch property
 - fixed file imports and error reporting
 - stored queries' results may be sorted by file size
 - rationalized activation and behaviour of the Edit menuitems
 - pinot may be started on any new query with -q/--query TERMS
 - require gtkmm >= 3.24

2020/01/09 version_1_1_0
General:
 - updated gmime, textcat, exiv, glib, OpenSSL and Xapian dependencies

2015/06/11 version_1_0_9
Tokenize :
 - new JSON filter
 - streamlined the mbox filter a bit
 - pinot-index looks for filters in the same locations as the UI
IndexSearch :
 - sort documents by date, in ascending or descending order

2014/07/18 version_1_0_8
General :
 - don't install the xdgmime files
 - use libnotify if --enable-libnotify=yes is specified
 - fixed check on library symbols that could cause a crash, added
 support for LLVM-built filters
 - general clean up
SQL :
 - refactored the DB interface

2014/05/22 version_1_0_7
General:
 - compilation fixes
 - prefer off_t for file sizes, offsets, buffer lengths
Monitor :
 - try and reapply watches on directories that have just been removed
Tokenize :
 - refactored encoding conversion
IndexSearch :
 - fixed build with libexttextcat

2013/05/26 version_1_0_6
IndexSearch:
 - support boost 1.50's Spirit
SQL :
 - better handling of potential errors while stepping through results
Collect :
 - minor fixes to curl backend

2013/03/03 version_1_0_5
IndexSearch:
 - fixes to abstract generation, CJKV tokenization

2013/02/11 version_1_0_4
IndexSearch:
 - fix stripping of diacritics
 - stem subject terms

2013/01/14 version_1_0_3
IndexSearch :
 - fix a Unicode handling issue introduced in 1.01
UI :
 - updated French translation by Eliovir

2012/11/04 version_1_0_2
General :
 - turn memory pooling off by default to avoid issues with newer boost
UI :
 - new Czech translation by Zbyněk Schwarz
 - updated Japanese translation by Takafumi Arakaki
 - updated Brazilian Portuguese translation by Adriano Steffler

2012/08/27 version_1_0_1
General :
 - run rst2html on RST files, if available, and if RST files are detected as such.
 See http://code.google.com/p/pinot-search/issues/detail?id=12
Tokenize :
 - better mbox parts extraction
IndexSearch :
 - dropped unac in favour of own code, resulting in faster indexing.
 - pinot-index --override MIMETYPE:EXTENSION overrides MIME type detection
 based on files extensions

2012/06/16 version_1_0_0
General :
 - install headers and libraries
 - prefer default programs that don't support URIs to view local files
Deskbar :
 - dropped support for Deskbar since it's now dead
Tokenize :
 - better mbox parsing. The internal part numbering scheme has changed;
 reindexing email is recommended.
IndexSearch :
 - support for LibreOffice's libexttextcat v3.2
 - fixed the Google plugin
 - renamed the Freshmeat plugin to Freecode
UI :
 - updated Simplified Chinese translation by happymeng
 - updated German translation by Gena Haltmair
 - initial GTK+ 3 port. Enable with "./configure ... --enable-gtkmm3=yes"

2011/11/07 version_0_9_8
Tokenize :
 - new exiv2-based filter
 - new chmlib-based filter
IndexSearch :
 - support for LibreOffice's libexttextcat v3.1, and possibly v3.1.1
 - dropped plugins for Yahoo! REST API, Yahoo! BOSS, Google Code Search
  and RollYO
UI :
 - query results and view history are expired after 6 months
 - updated Dutch translation by Martijn Verstrate and Tico
 - updated German translation by Fitoschido
 - updated Russian translation by Nikolay Kachanov
 - updated Spanish translation by pkramerruiz

2011/01/09 version_0_9_7
General :
 - replaced custom memory pool class with Boost's
 - don't try and map more than 2Gb and use shared mappings
Deskbar :
 - install the module where DeskBar > 2.28 expects it to be
IndexSearch :
 - remove dots at the end of terms that don't look like acronyms
 - index components of acronyms and dot-separated terms on their own
Daemon :
 - index files as they are crawled, don't delegate indexing to other
  threads, unless PINOT_MAXIMUM_INDEX_THREADS > 1
 - fixed checking of symlinks against black-list
UI :
 - the maximum number of results returned by the Query field and used
  to initialize new stored queries follows PINOT_MAXIMUM_QUERY_RESULTS
 - when a spelling suggestion is available, don't show the same revised
  query multiple times
 - updated Simplified Chinese translation by mike2718
 - updated Dutch translation by Dirk Roos
 - updated Italian translation by Davide Vidal and Simone Sandri
 - updated Japanese translation by Mizuki-san
 - updated Brazilian Portuguese translation by feen
 - updated Portuguese translation by Almufadado
 - updated Russian translation by Alexander Zinin and Nikolay Kachanov
 - updated Spanish translation by Juan Miguel Boyero Corral, Matias Fonzo
  and Fitoschido

2010/07/12 version_0_9_6
General :
 - fixed "GIO can sniff PNG" program used at configure time
 - README clarifies that operators should be upper-case
 - builds with gmime-2.4 or 2.6
 - link with the library that has dlopen() and fix Debian bug #556062
 - merged Debian's patch for --as-needed support
 - merged FreeBSD build patch
 - dropped support for Xesam
SQL :
 - fixed prepared statements interface to work with insertion and deletion
 - sleep then retry operations if the database is busy
 - better transaction support
Tokenize :
 - the mbox filter now supports messages of type "message/external-body"
IndexSearch :
 - fixed possible crash at exit time when the textcat configuration file
  points to non-existing model files
Xapian :
 - prefer the Chert back-end if available. Applicable to Xapian >= 1.2.0.
Daemon :
 - check symlinks against black-list
 - the battery status can now be obtained from DeviceKit-power or upower
UI :
 - updated Simplified Chinese translation by Eleanor Chen
 - updated French translation by verdy_p and Fabrice Colin
 - updated Hebrew translation by Yaron
 - updated Brazilian Portuguese translation by andbelo

2009/11/14 version_0_9_5
General :
 - OpenBSD support, thanks to the work of Antoine Jacoutot
 - fixed build when HAVE_DBUS isn't set
SQL :
 - use prepared statements on most common queries, transactions on mass updates
Tokenize :
 - better handling of acronyms
IndexSearch :
 - updated Bing plugin
 - removed plugins for Exalead and IOI
Xapian :
 - fixed the "path:" operator. Reindexing may be necessary
 - rewrote Search This For feature
UI :
 - if gtkmm >= 2.16 is available, the Find button is replaced with an icon
 - updated Dutch translation by JW
 - updated French translation by Thierry Thomas
 - updated German translation by Fabian Affolter and Marco Jahn
 - updated Hebrew translation by Ddorda
 - updated Portuguese translation by Bernardo Lopes
 - updated Spanish translation by Jesus Tramullas and DiegoJ

2009/06/27 version_0_9_4
General :
 - set _FILE_OFFSET_BITS=64 and fix Debian bug #530572
 - merged Funda Wang's linkage patch for Mandriva
 - gmime 2.4 is required
Tokenize :
 - mbox filter now fully works with gmime 2.4
 - archives filter supports Debian packages
 - set the close-on-exec flag on document files
 - better MIME type detection removes superfluous calls to external
  uncompressor programs when dealing with archives
 - use file names as title for files attached to mbox messages
 - fixed "quashing" of results titles
 - fixed indexing of the last document's attachments in an mbox
IndexSearch :
 - new search filter "inurl" allows finding files from an mbox or archive
  at a given URL
 - pinot-index --check on an mbox or archive will return the ID of the
  first nested document
 - pinot-search shows an estimate of the total number of results
 - the Bing plugin replaces the MSN plugin
 - fixed Freshmeat plugin
Daemon :
 - send a IndexFlushed signal over D-Bus when the index changes on disk
 - fixed restoring of user-set metadata
UI :
 - better MIME type detection fixes cases where documents nested in
  archives couldn't be open and viewed
 - reopen the index upon receiving the IndexFlushed signal
 - show properties of external indices' documents read-only
 - on exit, delete temporary files created for viewing some documents
 - fixed More Like This on Web results

2009/04/13 version_0_9_3
Tokenize :
 - moved the first 5Mb limit from the terms generator to the tokenizer
Daemon :
 - fixed major bug that caused the daemon to reindex all files on each run,
  unless started in full scan mode

2009/04/10 version_0_9_2
General :
 - fixed successive initialization and cleanup of libxml2 that could lead to
  a crash with libxml2 2.7.3
IndexSearch :
 - work around invalid charset declarations in documents
 - fixed pinot-index handling of black-listed documents
 - redesigned how documents nested in other documents (eg mbox...) are indexed
 - removed MozDex plugin
Tokenize :
 - new filter for tar files and ISO images based on libarchive >= 2.6.2 to
  index the content of those archive formats. Enable with "./configure ...
  --enable-libarchive=yes"
Daemon :
 - major changes to try and minimize memory usage
 - PINOT_MAXIMUM_INDEX_THREADS sets the daemon's number of indexing threads and
  defaults to 4
 - indexes created with version < 0.92 will be automatically upgraded
UI :
 - can open/view files within indexed archives

2009/03/07 version_0_9_1
General :
 - patch by Adel Gadllah for gcc 4.4
 - removed obsolete Encoding field from .desktop files
Tokenize :
 - new HTML filter based on Xapian Omega's HTML parser
 - prevent rpm from choking on files with the ".rpm" extension that are not RPMs
 - look for the ROBOTS metatag in remote documents only
 - only consider the first 5Mb of documents
IndexSearch :
 - pinot-search can run stored queries created by the UI
 - pinot-index can deal with relative paths, index directories and their contents,
  open My Web Pages, My Documents or other UI-configured index by name
Daemon :
 - fixed issue where symlinks would get unindexed every second run
 - fixed memory leak in time to timestamp conversions
 - fixed memory leak when reapplying user-set metadata
 - fixed memory leak when skipping the download of a local file
 - don't flush the index while files are being indexed
UI :
 - fixed spelling suggestions on the live query
 - the Import URL option can import local directories and their contents in
  My Web Pages
 - updated German translation by Gena Haltmair
 - updated Portuguese translation by Flávio Martins

2009/01/29 version_0_9_0
General :
 - builds with MingW
 - patch by Martin Michlmayr to fix gcc 4.4 build errors
Tokenize :
 - skip mbox messages flagged deleted by Evolution
 - patch by Adel Gadllah to build with gmime 2.4
 - catch conversions errors to/from unsupported charsets
 - fixed handling of Unicode space and punctuation code points
Collect :
 - look for the extended attribute user.mime_type on local files
 - HTTP downloaders can do POST
Monitor :
 - check /proc/sys/fs/inotify/max_user_watches and set aside 1k watches for
  other applications if possible
 - patch by Adrian Bunk to fix inotify support with recent kernel headers
Search :
 - better rebasing of results' URLs
 - dropped A9 and BitTorrent plugins
 - added plugin for the Internet Open Index
 - fixed various issues with getting more than one results page from Web engines
 - support for HTML extracts in OpenSearch Response
Xapian :
 - some terms were not always highlighted in the abstract
 - support for the "path:" operator
 - fix to always return the total results estimate
 - don't index the title without prefix as if it were in the text body, but
  let queries search across both text body and title. The "title" filter
  still allows searching titles exclusively. Requires Xapian >= 1.0.4.
 - always add a term for "dir:/"
 - MIME type terms don't include any charset specification, as intended
 - remove the original query's terms, stopwords, infrequent terms or similar
  terms if the stemming language is set from More Like queries
 - close all databases in an orderly manner
Search :
 - pinot-search has a "sort by date first" mode
Scripts :
 - pinot-cd.sh implements a "tagged cd"
 - pinot-check-file.sh simplifies determining if a file is in My Documents
Deskbar :
 - pinot-module.py uses the new Query method, supports deskbar v2.24 snippets
Daemon :
 - user-set metadata, including labels, is saved and restored when reindexing
 - D-Bus method RenameLabel obsoleted
 - D-Bus method Query replaces SimpleQuery and allows querying the same engines
  as the UI
 - skip symlinks that refer to places that have been crawled or will be crawled
 - if the daemon was interrupted while crawling, some files would never be indexed
 - indexes created with version < 0.90 will be automatically upgraded
UI :
 - spelling suggestions are shown inline, above results. Upon selecting one and
  clicking the Yes button, a new query based on the selected suggestion will be
  created.
 - restructured menus
 - queries' Index Results option updates documents already in My Web Pages. This
  doesn't apply to those in My Documents
 - on More Like This, selected documents are indexed to My Web Pages if necessary
 - both results and index lists can be exported to CSV or XML
 - Open Parent opens the directory a file is in
 - say "No results" instead of showing a blank results list
 - when viewing a document and GIO is used, make sure we consider the default
  application(s) first
 - support for the "path:" operator
 - preferences can be open independantly with "pinot -p" or "pinot-prefs"
 - smoother status window
 - the List Contents Of menu wasn't refreshed after editing an index' name
 - automatic migration of pre-0.90 configuration
 - larger default blacklist
 - updated Simplified Chinese translation by rainofchaos
 - updated Dutch translation by JW
 - updated French translation by Frédéric Grosshans
 - updated German translation by Gena Haltmair
 - updated Japanese translation by Takeo Mizuki
 - updated Brazilian Portuguese translation by Henrique P. Machado
 - updated Portuguese translation by _PN_boy
 - updated Swedish translation by Daniel Nylander
 - new Hebrew translation by Yaron

2008/09/20 version_0_8_9
Xapian :
 - indexing and searching are now diacritics insensitive by default, thanks
  to Unac 1.7.0 by Loic Dachary
 - support for removal of stopwords at query time. Language specific lists
  should be installed in $PREFIX/share/pinot/stopwords and be named
  stopwords.language_code
 - better abstracts for short queries
Daemon :
 - fixed indexing of plain text and XML files, following changes made in 0.88
 - indexes created with version < 0.89 will be automatically upgraded
UI :
 - fixed boolean operators in spelling suggestions, broken in 0.88
 - spelling suggestion doesn't suggest the same thing over and over again
 - dehyphen queries on line breaks, useful with text pasted from an external
  document
 - updated Simplified Chinese translation by Aron Xu
 - updated Brazilian Portuguese translation by André Gondim

2008/08/30 version_0_8_8
General :
 - replace xdgmime with GIO if it can sniff PNG at configure time
 - with gcc 4.x, set symbol visibility to hidden by default
Tokenize :
 - for unknown text formats, don't be too quick to fall back on the plain
  text filter
 - the output of external filters can be scanned if need be
 - some support for OpenXML formats
Search :
 - obsolete Google API engine now built as a dynamic backend
 - backends provide slightly more information
 - updated several plugins
 - removed the CreativeCommons plugin
Index :
 - pinot-index --showinfo shows which actions are associated with a MIME type
Xapian :
 - only support boolean operators in upper-case. This helps fixing issues
  with queries made of text pasted from elsewhere.
Daemon :
 - extended GetStatistics to return the flags "low disk space", "on battery"
  and "crawling"
UI :
 - on first runs, create useful stored queries
 - display extended status in the Status window
 - updated Simplified Chinese translation by rainofchaos
 - updated German translation by Gena Haltmair
 - updated Brazilian Portuguese translation by Rafael Porto Rodrigues
 - updated Swedish translation by Daniel Nylander

2008/07/20 version_0_8_7
General :
 - install the Amazon API plugin with other plugins
Collect :
 - obey META REFRESH if set
Search :
 - basic Xesam back-end based on xesam-glib. Enable with configure's option
  --enable-xesam-glib
 - in plugins, setting a value to "EDIT:description" makes it editable and
  allows to assign it a value at search time
 - pinot-search can set editable parameters with -e/--seteditable
 - plugin for the new Yahoo! BOSS API
Xapian :
 - skip very short non-CJKV terms when expanding queries.
 - fixed mangling of some CJKV queries
 - abstract generation is less skewed towards common terms
UI :
 - support for drag-n-drop to the stored queries list. Dropping a file will
  create a query to look for similar documents. Dropping text will create a new
  query set to that text.
 - preferences let the user edit all editable parameters defined in the plugins.
  They are saved to the configuration file.
 - extracts can be selected, copied and dropped onto the queries list to create
  new queries
 - let foreground threads run for a minute max
 - search-only backends (such as Xesam) will appear in the Current User channel

2008/06/21 version_0_8_6
General :
 - make manuals generates the manuals with help2man
 - dropped date parser for curl's/neon's
Tokenize :
 - decode emails subject lines properly
 - the HTML filter skips HTDig's no_index block
SQL :
 - query history can keep more than one results set
Collect :
 - use Last-Modified header as document's date
Search :
 - don't run queries consisting exclusively of spaces
 - fixed A9 plugin, removed Accoona
Xapian :
 - don't attempt offering suggestions for CJKV terms
Daemon :
 - in ignore-version mode, reapply labels too
 - documents from directories removed from indexing/monitoring should now be
  unindexed on full scans
 - SIGTERM wasn't caught !
UI :
 - use buttons on notebook tabs
 - the live query text field doesn't offer suggestions for filters and ranges
 - after a query edit, lists of documents are refreshed correctly
 - fixed date displayed when viewing query history
 - query history keeps the last two sets
 - .desktop file was missing Japanese and Simplified Chinese comments
 - SIGTERM wasn't caught !

2008/05/11 version_0_8_5
Build :
 - removed reference to m4 directory
General :
 - synced with gtk+'s xdgmime
Tokenize :
 - new libexif-based filter to extract image metadata
 - better conversion of mbox messages and HTML documents to UTF-8
 - tweaks to the CJKV tokenizer
SQL :
 - more abstract database interface
Search :
 - fixed CJKV queries on Web engines
 - fixed repetition of CJKV characters in abstracts
Xapian :
 - mixed CJKV queries should be processed correctly. See README.
 - fixed repetition of CJKV characters in abstracts
 - if the document specifies a language, double check it's valid
Daemon :
 - check whether there's already a daemon process running and exit if there is
 - fixed concurrency issues
 - fixed crawling and monitoring of new directories
UI :
 - if the global configuration file can't be open, don't reset the configuration
 - new Japanese translation by Takeo Mizuki

2008/03/27 version_0_8_4
Build :
 - patch by Adel Gadllah to fix gcc 4.3 build errors
General :
 - updated FAQ with how to compact the index
Monitor :
 - patch by Michael Biebl for inotify on m68k, mips, mipsel and hppa
Search :
 - new plugin for UNdata
Index :
 - fixed possible crash when pinot-index exits
Xapian :
 - faster CJKV indexing
 - the spelling database is populated with CJKV terms too
Daemon :
 - fixed possible crash when pinot-dbus-daemon exits
 - added option --ignore-version to deal with compacted indexes
UI :
 - fixed possible crash when pinot exits
 - fixed issue with signaling between crawler and indexer
 - stored queries can index all, or only new, results
 - in Preferences, patterns can be reset to default values
 - the Status window shows whether the daemon was stopped by, or
  disconnected from D-Bus
 - new simplified Chinese translation by Ashlee Ma

2008/02/28 version_0_8_3
Build :
 - fixed build errors with gcc 4.3, thanks to Adel Gadllah
 - fixed backend and non-backend flags mismatch
General :
 - dropped deprecated Encoding keys in .desktop files, as pointed out by
  David Paleino
Tokenize :
 - filters definition in external-filters.xml can specify what charset
  the text output is in
 - convert documents into UTF-8 prior to indexing
Search :
 - pinot-search supports option "--stemming LANGUAGE_NAME"
 - updated results parsing in Google.src
Xapian :
 - initial support for CJKV. See README for details
 - consider stemmed terms when building extracts
Daemon :
 - fixed options parsing
 - don't stop the directory crawler thread after 5 minutes
UI :
 - fixed extract display, broken in the previous release
 - tabs can be reordered, notebook is scrollable
 - make sure the Status window doesn't miss crawler errors
 - updated Spanish translation by Jesus Tramullas

2008/01/26 version_0_8_2
Build :
 - don't link to unnecessary libraries
Search :
 - removed the WiseNut plugin
 - fixed the Sherlock plugin parser's handling of input items, thanks to
  Claudio Bustos Navarrete
 - support for Xesam RC1
 - don't build the Xesam UL parser if Spirit is not available, thanks to
  Reuben Thomas
Xapian :
 - back-end moved into a dynamic library
 - fixed several issues with query stemming
 - generate terms for the MIME class
Daemon :
 - log an error when there's no inotify watch left
 - export HasDocument over D-Bus
UI :
 - when the index needs updating, tell the user on every run until he clicks
  the "Don't warn me again" checkbox
 - stemming is now configured separately and not driven by the "lang" filter
 - don't correct spelling of auto-generated and previously corrected queries
 - defer importing to the main window
 - Status window shows which engines are available
 - larger default blacklist
 - viewed documents are added to the list of recently used files. Requires
  gtkmm >= 2.10
 - updated Dutch translation by JW
 - updated Spanish translation by Jesus Tramullas
 - updated Swedish translation by Zirro

2007/11/24 version_0_8_1
Build :
 - misc fixes
General :
 - updated FAQ and README
 - fixed Icon field in desktop files
Index :
 - workaround for broken shared-mime-info rules that identify HTML files as
  Mozilla bookmarks
 - files whose name includes a question mark were not indexed correctly
 - pinot-label would loop forever if the supplied file name wasn't in the index
 - don't build the spelling table if the env var PINOT_SPELLING_DB is set to NO
Deskbar :
 - new plugin compatible with Deskbar 2.20
Daemon :
 - stop crawling and indexing if the partition on which the index resides is
  getting full. By default, that means less than 50 Mb. This can be overriden
  with the env var PINOT_MINIMUM_DISK_SPACE, eg PINOT_MINIMUM_DISK_SPACE=100
  for 100 Mb
 - stop crawling when the system goes on battery and restart when on AC. This
  requires support for the freedesktop.org's Power Management spec, or pre-spec
  gnome-power-manager.
UI :
 - fixed build against libsigc++ 2.1
 - updated Portuguese translation by Tiago Silva
 - updated Swedish translation by Daniel Nylander

2007/11/01 version_0_8_0
Build :
 - SMP builds, thanks to Gabriel C
Index :
 - unknown document types can be indexed if one of their parent types is known
 - new pinot-label tool to get, set and list labels on indexed files from the
  command-line
 - limit external programs to 5 minutes of CPU time
Search :
 - support for date (year, month, day), time (hours, minutes, seconds) and
  size (in bytes) ranges
 - attempt to correct the spelling of index queries that don't match anything
 - log how long queries take
 - keep connection to remote databases alive
 - better query expansion
 - the Yahoo! plugin was replaced with the Yahoo! API plugin
 - pass queries to Web engines unmodified, without attempting to filter results
  based on a filter or a range used in the query
Daemon :
 - fixed Reload method
 - new D-Bus methods to manage labels
 - new --reindex option
UI :
 - Search This For menu to search in results
 - suggest spelling corrections for index queries that don't match anything
 - all indices can be browsed
 - inline URL completion in the import dialog box, based on previous results
 - better support for user-specific MIME settings, thanks to Lee Marks
 - reload MIME settings when they are edited
 - results of stored queries can be sorted by relevance (default) or by date
 - new History button to show previous results for a stored query
 - documents' properties are updated in the background
 - documents' terms can be saved to a file
 - Status window shows description of errors
 - send a Reload to the daemon only when the relevant preferences are modified
 - updated Dutch translation by JW
 - updated Brazilian Portuguese translation by Leonardo Melo

2007/08/23 version_0_7_6
Build :
 - also look for textcat.h in libtextcat
Monitor :
 - a deletion would deadlock the monitor and prevent from processing any
  further event
 - unindex directories' contents when deleted and update when moved
Collect :
 - try to open files with NO_ATIME if possible
 - Neon-based downloader had not been brought up to date
Index :
 - replaced na(t)ive tokenizer with Xapian's TermGenerator. No effort is
  made to convert text to UTF-8 yet, so this depends on document formats
  and encodings
 - preliminary support for spelling corrections, without user feedback
 - index directories are tagged as cache directories so that they are skipped
  by "tar --exclude-caches"
 - preserve documents' title if possible, use user-specified title on import
Search :
 - fixed Sherlock parser for boost 1.34
 - support for gSOAP 2.7.9e
 - when searching an index, don't resort to OR'ing all terms if the original
  query doesn't match anything, this only confused users
Deskbar :
 - removed unnecessary shebang in script
Daemon :
 - new D-Bus method Reload, that enables to reload the configuration and act
  upon it whenever modified by the UI
 - SimpleQuery still resorts to OR'ing all terms if the original query doesn't
  match anything
UI :
 - fixed some minor cosmetic bugs, tweaked a few things
 - user-specific MIME settings in ~/.local have priority over system settings
 - queries with at least a start date can be run, and filter a documents list
 - the label specified on import was ignored
 - new traditional Chinese translation by Yung-Chung Lin
 - updated Portuguese translation by _PN_boy
 - updated Swedish translation by Daniel Nylander

2007/07/28 version_0_7_5
General :
 - install Dijon's ChangeLog
Index :
 - files with nested documents (eg mbox) could sometimes not be fully indexed
  and/or and the wrong MIME type was reported
Search :
 - updated Xesam Query Language parser to reflect current spec
 - updated Sherlock plugin for Exalead, removed Ask
Daemon :
 - mbox files are no longer configured separately : those found during
  a crawl are indexed and monitored automatically
 - patterns list can be used as a blacklist (default) or whitelist
UI :
 - fixed a bug where filtering a documents list with an empty query would
  prevent from viewing the list without or without a filter query
 - avoid a crash when viewing the properties of several documents one after
  the other
 - updated Dutch translation by Balaam's Miracle
 - updated Portuguese translation by _PN_boy

2007/06/24 version_0_7_4
General :
 - make uninstall actually uninstalls all files
Index :
 - can detect Hungarian, Romanian and Turkish with libtextcat 2.2 and stem
  with Xapian 1.0
 - adopted Xapian 1.0's new indexing strategy
Search :
 - basic support for the Xesam Query and User Language in pinot-search
Daemon :
 - history database is separate from the UI's
UI :
 - in index list tabs, replaced labels filtering with stored queries filtering,
  so that one can find out which and how many documents in the index being
  shown match a query (the query's maximum number of results is ignored)
 - Hungarian, Romanian and Turkish are valid document languages

2007/05/23 version_0_7_3
Tokenize :
 - mbox filter was broken in previous release
 - TagLib filter returns the file's name as title if no tag is found
SQL :
 - don't force opening and closing the database on every request
Index :
 - overwrite index on upgrades, it's faster than deleting all its documents
 - support for Xapian 1.0 API
Search :
 - stemming wasn't activated when a language is set
 - support for Xapian 1.0 API
Daemon :
 - don't scan for deleted files on every run
UI :
 - fixed crash when unindexing documents
 - updated Dutch translation by JW
 - updated Italian translation by Marco Bazzani

2007/04/28 version_0_7_2
General :
 - synced with gtk+'s xdgmime
Index :
 - add extra term to documents so that filter "dir:/" can be applied
Search :
 - results lists can be saved as CSV or OpenSearch response XML/RSS
Daemon :
 - indexes created with version < 0.72 will be automatically upgraded
 - fixed problem where too much time spent upgrading would cause the daemon
  to be killed
UI :
 - open documents on HTTP/HTTPS with the default Web browser
 - updated German translation by Christian Dywan
 - updated Italian translation by Vincenzo Consales

2007/03/31 version_0_7_1
Tokenize :
 - HTML filter wasn't always properly initialized, which could cause a crash
Monitor :
 - file creation wasn't acted upon
 - moves to unmonitored locations are treated as deletions
Collect :
 - fixed building with a version of curl that doesn't depend on OpenSSL
 - dropped pinot-collect tool
Index :
 - added versioning
 - label names, file names, directory names and URLs are escaped
 - directories are indexed as separate documents
Search :
 - filters with spaces or control characters should be double-quoted, eg :
  file:"Cats & Dogs.txt"
Daemon :
 - extended D-Bus methods Set and GetDocumentInfo
 - fixed off-by-one error in numbering of mail messages parts
 - blacklist wasn't applied to files for which the monitor reports events
 - an index created with Pinot < 0.71 is automatically upgraded
Deskbar :
 - use deskbar.Utils.url_show() if available
UI :
 - Status window is now live
 - View on a plain text message shows the mail headers
 - added proxy support for Web engines queries and collection of documents on HTTP
 - better work-around for desktop files that attempt setting env variables in Exec
 - use Gtk::ComboboxText where appropriate
 - German translation by Christian Dywan
 - Italian translation by Michele Angrisano
 - Portuguese translation by _PN_boy

2007/03/06 version_0_7_0
Tokenize :
 - now use Dijon's filters (http://dijon.berlios.de/)
Index :
 - support for remote indexes served by xapian-progsrv+ssh
 - index to use by pinot-index specified with --db
Search :
 - date range filtering applies to index searches
 - limit the number of results returned by pinot-search with --max
Daemon :
 - log the daemon's PID
 - sped up unindexing of documents after a directory is deleted
 - caught up with changes in D-Bus 1.0
UI :
 - stored queries can do date range filtering. If dates don't make sense
  (eg From >= To), they are ignored
 - under the Session menu, Status shows various bits of information about
  the indexes and the daemon's crawler
 - prompt for command to use to open documents of a type for which no
  application is defined
 - revamped configuration dialog for external indexes
 - Index > Properties now shows a document's size and number of unique terms
 - most operations that involve peeking at the index are done in the background
 - work-around for desktop files that attempt setting env variables in Exec
 - caught up with changes in D-Bus 1.0
 - Brazilian Portuguese translation by Leonardo Melo
 - Russian translation by Sergey Vostrikov

2006/12/21 version_0_6_5
Build :
 - added option --enable-debug to configure, --enable-soap replaces --with-soap
 - complain bitterly if libtextcat header is not found
 - Pthreads may be provided by a library other than libpthread
 - install configuration files in sysconfdir, libraries in libdir
Tokenize :
 - optimized Ogg/MP3 filter
Index :
 - fixed extraction of language and size from document data
 - store date terms for future date range filtering
 - fixed argument checking in pinot-index
Daemon :
 - new DBus method SetDocumentsLabels to relabel several documents at once
 - process DBus messages in a separate thread, not in the main thread
UI :
 - fixed expansion of .desktop's Exec
 - initialize D-Bus ! Not sure why this didn't cause problems before.
 - fixes for when the locale is not UTF-8
 - on first run, open the Preferences box and show the Indexing tab
 - for stored queries set to index and label results, only apply the new label
  to results that already are in one of the indexes, don't do a full update
 - refresh labelled documents list correctly after properties are changed

2006/12/05 version_0_6_4
Collect :
 - don't needlessly load files that are going to be handled by an helper application
Index :
 - store documents size and file extension, if any
 - fixed concurrency bug that could thrash the index on SMP systems
 - very long capitalized terms could lead to document loss
 - file names were always lower-cased
Search :
 - regenerated the Google SOAP API stubs with gsoap 2.7.8c
UI :
 - added filter on file extension
 - fixed crash on SMP systems when listing an index
 - fixed deadlock when indexing a query's results. Oddly enough, it seems it
  happened only on FreeBSD !
 - Dutch translation by Tikkel
 - Swedish translation by Daniel Nylander
Daemon :
 - set a lower scheduling priority
 - queue events in the database, not in memory
 - clean exit when signalled/stopped while crawling

2006/11/18 version_0_6_3
Collect :
 - watch out for NULL characters in data
Tokenize :
 - fixed memory leak. Temporary documents were not deleted most of the time
UI :
 - prettified results list
 - better abstract highlighting
 - fixed clipboard copy of results list and abstract
Daemon :
 - autostart the daemon process
 - with dbus < 0.70, close the connection

2006/11/04 version_0_6_2
General :
 - query shared-mime-info prefix, so that the applications database can be
  loaded even when Pinot is installed under a different prefix
 - copyright notice was missing in source
Index :
 - detect and support libtextcat 3.0 peculiarities
 - can skip files based on glob pattern
Search :
 - fixed issue where label and directory filters were not applied correctly
  when the filter doesn't start with an upper-case letter. Directory filters
  starting with a non-alphanumeric character only work with Xapian >= 0.9.8.
 - fixed A9, Accoona and Exalead plugins
UI :
 - file patterns to skip can be set in Preferences, Indexing
 - columns showing a timestamp were sorted alphabetically
 - refresh index lists correctly when exiting Preferences
Daemon :
 - fixed major bug where the daemon would loop endlessly reindexing mp3/ogg
  files. When notified that a writable file was closed, check the file was
  actually modified before reindexing it.
 - fixed D-Bus warning about closing the connection when exiting

2006/10/18 version_0_6_1
General :
 - switched to gtk+'s version of xdgmime
Index :
 - can now run queries like "type:text/html and lang:en and (tcp near ip)".
  See README for more information about the syntax and a list of filters.
Search :
 - don't reject MozSearch plugins
 - added plugin for Google code search service
 - filters 'site' and 'file' (host name, file name) apply to Web engines
 - fixed abstract generation
UI :
 - revamped the stored queries editor to allow any number of terms and filters
 - preferences relative to My Documents and My Web Pages gathered under the
  same tab
 - hide the stored queries and engines lists by default to avoid scaring
  people used to Beagle too much ;-)
 - mail accounts configuration wasn't always saved correctly
 - updating a document from the My Web Pages index messed the abstract
Daemon :
 - index attachements in mbox files
 - documents deleted since last crawl were not removed from the index if the
  corresponding location wasn't monitored

2006/09/25 version_0_6_0
Build :
 - modifications for building on Cygwin, contributed by Reini Urban
Monitor :
 - inotify monitor more flexible
SQL :
 - retry if the database is busy
 - save the daemon's crawler history
Index :
 - able to open remote indexes that were not initially available
 - indexes are open in write mode only when necessary
 - prefer Xapian's Flint back-end to Quartz
 - terms starting with an upper-case letter are not ignored by terms suggestion
 - refresh terms generated from the previous title, location etc... when updating a document
 - MIME type and directory hierarchy are stored as terms
 - store documents last modification date in a way compatible with Xapian Omega
Search :
 - fixed abstract generation with remote indexes, and queries with OR'ed terms !
 - brought MozDex plugin back
UI :
 - show results scores
 - fixed language shown in the properties box when language is unknown
  for one or more of the documents selected
 - better at queuing document indexing
 - My Email index is replaced with the index managed by the daemon
 - search terms suggestion, More Like This and the index status icon on results
  rely on both internal indexes
 - only URLs can be imported now as the daemon takes care of local files
Daemon :
 - crawls and indexes user-defined locations on the filesystem, mbox files included
 - recursively monitors these locations for changes and updates the index
 - provides a DBus interface suitable for querying and document inspection
 - no dependency on gtkmm
Deskbar Applet :
 - plugin for Deskbar Applet that searches documents indexed by the daemon

2006/07/05 version_0_5_0
Tokenize :
 - switched to pdftotext, which seems to support more files than pdftohtml
SQL :
 - remove history items older than one month at startup
Search :
 - can search for documents with language "Unknown" (ie those for which
  language detection failed)
UI :
 - multiple user-configurable cache services. They are defined in
  $PREFIX/share/pinot/globalconfig.xml and listed under the View Cache menu.
 - allow to change the language of several documents at once
 - after editing a query, don't change selection in the queries list
 - More Like This extracts representative terms from the selected documents
  (if they are indexed) and creates a "More Like ..." query
 - set the indexed status icon when results are indexed
 - view results on double clicks
 - show a message if there is no application defined for a file's type
 - most tree columns can be sorted at last ! :-)
 - work-around for bizarre bug that causes a segfault when creating a query
  that indexes and labels results based on a language filter

2006/06/11 version_0_4_9
General :
 - able to find the default application for a given MIME type
Tokenize :
 - better HTML tokenizer
UI :
 - titles in documents lists are no longer truncated
 - if import fails, the progress bar displays the error message
 - dropped the internal viewer, default applications are launched on View
 - View Cache no longer dependant on Google SOAP API. It just points to the
  Google Cache for http[s] results
 - remember whether the queries list is expanded

2006/05/25 version_0_4_8
Monitor :
 - replaced FAM/Gamin with inotify
Search :
 - fixed BitTorrent source
UI :
 - fixed menuitems inconsitencies
 - always use user-provided language on update !
 - all query terms are highlighted in the extract !
 - threads that finish while the import dialog is up are processed when the
  dialog box is closed

2006/05/12 version_0_4_7
General :
 - all programs have man pages and support --help and --version
Search :
 - the Google API engine is no longer unnecessarily picky about queries parameters
 - updated Creative Commons plugin, based on the one shipped with Firefox
UI :
 - merged channel Web Services with The Web
 - the state of engine channels is saved and restored
 - query terms are highlighted in the extract field
 - allow editing the language of documents. A subsequent update would use the
  given language to stem terms.

2006/04/22 version_0_4_6
Search :
 - resurrected support for the Google API, enabled with "./configure --with-soap=yes".
  This requires the gsoap development toolkit.
 - tweaked extraction of results' extract and URL
 - renamed pinot_search to pinot-search
 - (temporarily hopefully) removed MozDex plugin
Index :
 - changed URL hash algorithm for compatibility with omindex. Reindexing documents
  with URLs > 128 characters is necessary
Tokenize :
 - new tokenizer for MP3, Vorbis and FLAC files that depend on the TagLib library
 - new XML tokenizer
 - new tokenizer for OpenDocument/StarOffice files (unzip required)
 - use xdgmime to query shared-mime-info for better MIME type detection
HTML browser :
 - can build against Mozilla or Firefox, set with "./configure --with-gecko=mozilla|firefox"
UI :
 - discovered and fixed pretty stupid bug that would cause a crash when indexing
  search results while the My Documents tab isn't opened
 - internationalized desktop file

2006/03/25 version_0_4_5
Build :
 - various fixes for building on Fedora Core 5
Index :
 - listing a label's documents displays the correct documents count in the status bar
  and navigation buttons are enabled or disabled accordingly
Search :
 - identified and fixed cause of crash when searching several engines simultaneously
 - Teoma and Ask Jeeves plugins replaced by Ask.com plugin
 - dropped Altavista and Lycos
 - added RollYO's Top News, Exalead and Creative Commons sources
 - fixed results extraction with Topix
 - better URL extraction for those search engines that use redirectors, for
  instance http://rds.yahoo.com/_ylt=.../**http%3a//some.website.com/here/index.html
 - tweaked abstract generation
UI :
 - when no email boxes were being monitored, Pinot could crash at exit time
  depending on which version of FAM was used
 - suggestion of query terms can be disabled
 - new results are now shown with a background colour of the user's choice
 - since index searches are multi-stepped, live queries' terms are now ANDed together

2006/03/12 version_0_4_4
Build :
 - fixed building of PDF, RTF and MS Word tokenizer libraries
 - HTTP library can be selected with "./configure --with-http=curl|neon"
Collect :
 - added curl-based downloader, used by default instead of Neon
Index :
 - dropped dependency on OTS, abstracts are now generated at search time
Search :
 - fixed page browsing with Lycos plugin
UI :
 - fixed potential deadlock, eg between saving preferences and listing the index
 - labels no longer have colours; selecting a label on an index only lists
  those documents that have the label

2006/02/25 version_0_4_3
Build :
 - dropped patch for libxml++ v0.26 support
 - use autotools
Search :
 - include pinot_search to query search engines from the command-line
 - better parsing of Sherlock plugins with foreign tags
 - URLs of results returned by Sherlock engines were incorrectly lower-cased
 - fix for Topix and Acoona
Index :
 - changed term prefixes to conform to conventions used by Omega and other
  Xapian-based tools. Unfortunately, users will have to update documents and
  reapply labels !
UI :
 - UTF-8 fixes on stored queries dates, results extracts and locale catalogs;
   the Spanish and French catalogs are usable now !
 - Better error handling in worker threads and at startup

2006/01/31 version_0_4_2
Search :
 - support for OpenSearch Description, Query and Response
 - replaced Koders and Omega Sherlock plugins with their OpenSearch equivalent
 - added MozDex
Index :
 - queries on an index that cannot be locked no longer loop
UI :
 - when no email boxes were being monitored, Pinot wouldn't exit right away
 - results extract field can be resized
 - Spanish translation by Jesús Tramullas (jesus at tramullas dot com)

2006/01/20 version_0_4_0
UI and SQLite :
 - dropped ActionHistory
Search :
 - fixed issues with documents and queries language
 - detect encoding of results pages
 - fixed AskJeeves source
Index :
 - fixed issues with documents language
 - limit amount of text parsed by summarization and language guessing
 - limit terms length
 - canonicalize URLs
 - make sure index is always unlocked properly
Tokenize :
 - modified tokenizer plugins interface
 - new RTF tokenizer (requires unrtf)
UI :
 - standard About box
 - better documents importing
 - better charset conversion
 - completion on query field, based on terms in the documents index
 - all notebook tabs are open on a need-to basis and can be closed
 - fixed several UI inconsistencies
 - catch signals and signal threads to ensure clean exits
 - save language names in English, load in current locale
 - requires gtkmm v2.6

2005/12/18 version_0_3_5
Xapian + SQLite back-end :
 - dropped unnecessary tables, moved labels and properties into the index
Neon downloader :
 - fail on HTTP errors
 - setup OpenSSL callbacks for multi-threading support
Search :
 - can filter documents by label
 - multi-step index search : exact matches first, then with stemming, then with ignoring the operators
HTML browser :
 - initialize NSS and NSPR to get rid of crash when visiting https sites
UI :
 - on double clicks in the index tab, check a document is selected before opening the properties box
 - copy & paste should now work as expected
 - modified index browser
Threads :
 - handle select() errors

2005/12/08 version_0_3_0
 - first public release