Release notes for Recoll 1.17.x

Caveats

Installing over an older version: 1.17 is mostly compatible with 1.16 and 1.15 indexes, except for a few differences for weird terms containing punctuation signs. Also, file/document size computations have been improved in 1.17 and you'll need an index reset if you want to enjoy the new goodness. Always reset the index if installing over an older version (1.14 and older). The simplest way to do this is to quit all recoll programs and just delete the index directory (rm -rf ~/.recoll/xapiandb), then start recoll or recollindex. recollindex -z will do the same in most cases.

Some new, auxiliary, features do require a full reindex to work:

• The file size filtering functions.
• The anchored search feature if the index was created by release 1.15 or older.

Notes to the kind packagers:

• you need to either depend on libqt4-webkit or run configure --disable-webkit (1st option preferred if possible of course).
• The default build now installs the Python extension module. This means that it needs the Python development package. You can revert to the old behaviour by using configure --disable-python-module.

Changes

Recoll 1.17 is an incremental improvements release over 1.16, with quite a few changes under the hood (like the introduction of a webkit-based result list), but relatively few obvious changes for the casual user.

1.17.1 was released only for a small function needed for supporting the Unity Lens. If you don't need it, you can stay with 1.17.0.

1.17.2 fixes a few bugs and adds a small feature for handling characters that should not be accented in your language (ie: å in swedish). See unac_except_transx in the manual configuration section. Also a new rcldia filter for Dia files.

1.17.3 fixes a nasty indexing bug in email processing code. This was triggered most probably my messages containing Microsoft mimehtml attachments, but possibly by other ones too, and caused an indexing crash. Also fixes small glitches in result list paging.

Recoll 1.17.0 changes:

• GUI/config: new set of dialogs to set up the indexing schedule, either as automatic real time indexing started with the user session or as discrete cron entries. This will take care of the details of autostart files or crontab editing.
• Search: Recoll now supports filtering on file sizes. This is accessible from the advanced search panel, or from the query langage (ie: "size>10k size<1m"). For documents stored inside archives, the file size stored in the index is now the actual one, not the archive's as was previously the case. Because of this, you will need a full reindex to completely enable this feature.
• Query language: the "dir:" directive now accepts non-absolute paths. For example dir:dir1/dir2 will select any file the path of which includes "dir1/dir2".
• Query language: tilde is now expanded inside a "dir:" clause.
• General: the Python extension module is now installed by default. This will make it much easier to use small recoll integration and extension hacks like the Unity Lens.
• Desktop: there is a new Ubuntu Unity Recoll Lens, it is not installed by default, but quite easy to install, see the Wiki page.
• GUI: it is now possible to cancel a query that takes too long. Because of limitations in Xapian, this forces a program exit, but it's probably better than staring at a hung Recoll.
• GUI/advanced search: the dialog gets new sections for filtering on file dates and sizes, and has been split into two tabs: for search terms, and for filtering.
• GUI: use Shift+PageUp instead of Shift+Home to rewind the result list to the first page. Shift+Home is useful in the search entry.
• GUI: the preview function will now check that the index is up to date before fetching the preview data, and propose to update the index if necessary. This is especially important for mail folders. It was previously possible that the preview would display a message unrelated to the current result.
• GUI: it is now possible to control the date format displayed by the results (go to Query Preferences).
• GUI: when a search fails to retrieve results, the spelling suggestions are now presented as links inside the result list area. Clicking a link will replace the appropriate word inside the search entry.
• GUI: thumbnails will be displayed in the result list if found in the standard freedesktop thumbnails directory (previously created by gimp or some file manager). This Recoll release will not create thumbnails.
• GUI: it is now possible to Copy/Paste from the term explorer result list.
• GUI: what used to be the document category filters (media/text/etc.) are now defined as query language fragments inside mimeconf. This means that you can redefine them freely (see the comments for the [guifilters] section in the default mimeconf).
• GUI: it is now possible to use a Qt style sheet to modify many aspects of Recoll appearance. A style sheet file can be selected from the GUI query configuration dialog. A sample skeleton style sheet is to be found in the share/recoll/examples directory.
• GUI: the result list is now based on the WebKit widget instead of QTextBrowser. This means better HTML/CSS support, and even Javascript ... You can force the use the old QTextBrowser instead by using configure --disable-webkit. See a few CSS examples in the result list customisation examples page
• GUI: added menu entry to show all the mime types actually indexed.
• GUI/indexing: the indexing thread has been removed. All indexing is now performed by a separate recollindex process. This has multiple advantages, the thread was just a very bad idea. Among others, exiting the GUI has no influence on an ongoing indexing.
• GUI/indexing: new menu entry to force a full reindex.
• GUI/config: the indexing configuration now only shows the directories customized by the user, not those which come from the standard configuration (ie: ~/.thunderbird). Showing the standard directories was more often confusing than useful.
• GUI: the default result list paragraph now displays the ipath: this is very useful for archive members and not too disturbing for other kinds.
• Command line query: can now select a result slice (first result rank and count of results). New option for easy parsing of results by another program.
• Indexing: handle the html5 charset meta tag.
• Indexing: added support for Okular notes and Gnumeric.
• Real time indexer: restart automatically when the configuration changes.
• Fixed bugs:
  • Indexing configuration: deleting one customized directory would sometimes delete the whole list (depending on Qt version I think).
  • Indexing configuration: adding a customized directory would not make it the active/current one (so that another click was needed to choose it before making modifications).
  • The GUI preview now performs an up to date check before displaying data from a subdocument, and proposes to reindex the container if it changed. This is mostly useful for mailbox containers, where preview would sometimes display the wrong message (if the folder had been compacted).
  • Fix a bug in html charset detection. The wrong charset was sometimes used for GUI preview.
  • The chm filter can now process files lacking a Topics node.
  • The default slack for proximity searches from the query language (ie: "word1 word2"p) had become 0. It was restored to 10.