Beets

- Plugin-provided fields can now be used in queries. For example, if you use the `inline` plugin to define a field called `era`, you can now filter your library based on that field by typing something like `beet list era:goldenage`.
- Album-level flexible attributes and plugin-provided attributes can now be used in path formats (and other item-level templates).
- Date-based queries are now possible. Try getting every track you added in February 2014 with `beet ls added:2014-02` or in the whole decade with `added:2010..`. Thanks to Stig Inge Lea Bjørnsen.
- The modify-cmd command is now better at parsing and formatting fields. You can assign to boolean fields like `comp`, for example, using either the words "true" or "false" or the numerals 1 and 0. Any boolean-esque value is normalized to a real boolean. The update-cmd and write-cmd commands also got smarter at formatting and colorizing changes.

For developers, the short version of the story is that Item and Album objects provide _uniform access_ across fixed, flexible, and computed attributes. You can write `item.foo` to access the `foo` field without worrying about where the data comes from.

Unrelated new stuff:
- The importer has a new interactive option (_G_ for "Group albums"), command-line flag (`--group-albums`), and config option (group_albums) that lets you split apart albums that are mixed together in a single directory. Thanks to geigerzaehler.
- A new `--config` command-line option lets you specify an additional configuration file. This option _combines_ config settings with your default config file. (As part of this change, the `BEETSDIR` environment variable no longer combines---it _replaces_ your default config file.) Thanks again to geigerzaehler.
- `ihate`: The plugin's configuration interface was overhauled. Its configuration is now much simpler---it uses beets queries instead of an ad-hoc per-field configuration. This is _backwards-incompatible_---if you use this plugin, you will need to update your configuration. Thanks to BrainDamage.

Other little fixes:
- `echonest`: Tempo (BPM) is now always stored as an integer. Thanks to Heinz Wiesinger.
- Fix Python 2.6 compatibility in some logging statements in `chroma` and `lastgenre`.
- Prevent some crashes when things go really wrong when writing file metadata at the end of the import process.
- New plugin events: `item_removed` (thanks to Romuald Conty) and `item_copied` (thanks to Stig Inge Lea Bjørnsen).
- The `pluginpath` config option can now point to the directory containing plugin code. (Previously, it awkwardly needed to point at a directory containing a `beetsplug` directory, which would then contain your code. This is preserved as an option for backwards compatibility.) This change should also work around a long-standing issue when using `pluginpath` when beets is installed using pip. Many thanks to geigerzaehler.
- `web`: The `/item/` and `/album/` API endpoints now produce full details about albums and items, not just lists of IDs. Thanks to geigerzaehler.
- Fix a potential crash when using image resizing with the `fetchart` or `embedart` without ImageMagick installed.
- Also, when invoking `convert` for image resizing fails, we now log an error instead of crashing.
- `fetchart`: The `beet fetchart` command can now associate local images with albums (unless `--force` is provided). Thanks to brilnius.
- `fetchart`: Command output is now colorized. Thanks again to brilnius.
- The modify-cmd command avoids writing files and committing to the database when nothing has changed. Thanks once more to brilnius.
- The importer now uses the album artist field when guessing existing metadata for albums (rather than just the track artist field). Thanks to geigerzaehler.
- `fromfilename`: Fix a crash when a filename contained only a track number (e.g., `02.mp3`).
- `convert`: Transcoding should now work on Windows.
- `duplicates`: The `move` and `copy` destination arguments are now treated as directories. Thanks to Pedro Silva.
- The modify-cmd command now skips confirmation and prints a message if no changes are necessary. Thanks to brilnius.
- `fetchart`: When using the `remote_priority` config option, local image files are no longer completely ignored.
- `echonest`: Fix an issue causing the plugin to appear twice in the output of the `beet version` command.
- `lastgenre`: Fix an occasional crash when no tag weight was returned by Last.fm.
- `mpdstats`: Restore the `last_played` field. Thanks to Johann Klähn.
- The modify-cmd command's output now clearly shows when a file has been deleted.
- Album art in files with Vorbis Comments is now marked with the "front cover" type. Thanks to Jason Lefley.

This update brings new plugins for fetching acoustic metrics and listening statistics, many more options for the duplicate detection plugin, and flexible options for fetching multiple genres.

The "core" of beets gained a new built-in command: `beet write` updates the metadata tags for files, bringing them back into sync with your database. Thanks to Heinz Wiesinger.

We added some plugins and overhauled some existing ones:
- The new `echonest` plugin can fetch a wide range of [acoustic attributes](http://developer.echonest.com/acoustic-attributes.html) from The Echo Nest, including the "speechiness" and "liveness" of each track. The new plugin supersedes an older version (`echonest_tempo`) that only fetched the BPM field. Thanks to Pedro Silva and Peter Schnebel.
- The `duplicates` plugin got a number of new features, thanks to Pedro Silva:
- The `keys` option lets you specify the fields used detect duplicates.
- You can now use checksumming (via an external command) to find duplicates instead of metadata via the `checksum` option.
- The plugin can perform actions on the duplicates it find. The new `copy`, `move`, `delete`, `delete_file`, and `tag` options perform those actions.
- The new `mpdstats` plugin collects statistics about your listening habits from [MPD](http://mpd.wikia.com/wiki/Music_Player_Daemon_Wiki). Thanks to Peter Schnebel and Johann Klähn.
- `lastgenre`: The new `multiple` option has been replaced with the `count` option, which lets you limit the number of genres added to your music. (No more thousand-character genre fields!) Also, the `min_weight` field filters out nonsense tags to make your genres more relevant. Thanks to Peter Schnebel and rashley60.
- `lyrics`: A new `--force` option optionally re-downloads lyrics even when files already have them. Thanks to Bitdemon.

As usual, there are also innumerable little fixes and improvements:
- When writing ID3 tags for ReplayGain normalization, tags are written with both upper-case and lower-case TXXX frame descriptions. Previous versions of beets used only the upper-case style, which seems to be more standard, but some players (namely, Quod Libet and foobar2000) seem to only use lower-case names.
- `missing`: Avoid a possible error when an album's `tracktotal` field is missing.
- `ftintitle`: Fix an error when the sort artist is missing.
- `echonest_tempo`: The plugin should now match songs more reliably (i.e., fewer "no tempo found" messages). Thanks to Peter Schnebel.
- `convert`: Fix an "Item has no library" error when using the `auto` config option.
- `convert`: Fix an issue where files of the wrong format would have their transcoding skipped (and files with the right format would be needlessly transcoded). Thanks to Jakob Schnitzer.
- Fix an issue that caused the id3v23 option to work only occasionally.
- Also fix using id3v23 in conjunction with the `scrub` and `embedart` plugins. Thanks to Chris Cogburn.
- `ihate`: Fix an error when importing singletons. Thanks to Mathijs de Bruin.
- The clutter option can now be a whitespace-separated list in addition to a YAML list.
- Values for the replace option can now be empty (i.e., null is equivalent to the empty string).
- `lastgenre`: Fix a conflict between canonicalization and multiple genres.
- When a match has a year but not a month or day, the autotagger now "zeros out" the month and day fields after applying the year.
- For plugin developers: added an `optparse` callback utility function for performing actions based on arguments. Thanks to Pedro Silva.
- `scrub`: Fix scrubbing of MPEG-4 files. Thanks to Yevgeny Bezman.

This release boasts a host of new little features, many of them contributed by beets' amazing and prolific community. It adds support for [Opus](http://www.opus-codec.org/) files, transcoding to any format, and two new plugins: one that guesses metadata for "blank" files based on their filenames and one that moves featured artists into the title field.

Here's the new stuff:
- Add [Opus](http://www.opus-codec.org/) audio support. Thanks to Rowan Lewis.
- `convert`: You can now transcode files to any audio format, rather than just MP3. Thanks again to Rowan Lewis.
- The new `fromfilename` plugin guesses tags from the filenames during import when metadata tags themselves are missing. Thanks to Jan-Erik Dahlin.
- The `ftintitle`, by Verrus, is now distributed with beets. It helps you rewrite tags to move "featured" artists from the artist field to the title field.
- The MusicBrainz data source now uses track artists over recording artists. This leads to better metadata when tagging classical music. Thanks to Henrique Ferreiro.
- `lastgenre`: You can now get multiple genres per album or track using the `multiple` config option. Thanks to rashley60 on GitHub.
- A new id3v23 config option makes beets write MP3 files' tags using the older ID3v2.3 metadata standard. Use this if you want your tags to be visible to Windows and some older players.

And some fixes:
- `fetchart`: Better error message when the image file has an unrecognized type.
- `mbcollection`: Detect, log, and skip invalid MusicBrainz IDs (instead of failing with an API error).
- `info`: Fail gracefully when used erroneously with a directory.
- `echonest_tempo`: Fix an issue where the plugin could use the tempo from the wrong song when the API did not contain the requested song.
- Fix a crash when a file's metadata included a very large number (one wider than 64 bits). These huge numbers are now replaced with zeroes in the database.
- When a track on a MusicBrainz release has a different length from the underlying recording's length, the track length is now used instead.
- With per_disc_numbering enabled, the `tracktotal` field is now set correctly (i.e., to the number of tracks on the disc).
- `scrub`: The `scrub` command now restores album art in addition to other (database-backed) tags.
- `mpdupdate`: Domain sockets can now begin with a tilde (which is correctly expanded to `$HOME`) as well as a slash. Thanks to Johann Klähn.
- `lastgenre`: Fix a regression that could cause new genres found during import not to be persisted.
- Fixed a crash when imported album art was also marked as "clutter" where the art would be deleted before it could be moved into place. This led to a "image.jpg not found during copy" error. Now clutter is removed (and directories pruned) much later in the process, after the `import_task_files` hook.
- `missing`: Fix an error when printing missing track names. Thanks to Pedro Silva.
- Fix an occasional KeyError in the update-cmd command introduced in 1.3.0.
- `scrub`: Avoid preserving certain non-standard ID3 tags such as NCON.

Albums and items now have **flexible attributes**. This means that, when you want to store information about your music in the beets database, you're no longer constrained to the set of fields it supports out of the box (title, artist, track, etc.). Instead, you can use any field name you can think of and treat it just like the built-in fields.

For example, you can use the `modify` command to set a new field on a track:


$ beet modify mood=sexy artist:miguel


and then query your music based on that field:


$ beet ls mood:sunny


or use templates to see the value of the field:


$ beet ls -f '$title: $mood'


While this feature is nifty when used directly with the usual command-line suspects, it's especially useful for plugin authors and for future beets features. Stay tuned for great things built on this flexible attribute infrastructure.

One side effect of this change: queries that include unknown fields will now match _nothing_ instead of _everything_. So if you type `beet ls fieldThatDoesNotExist:foo`, beets will now return no results, whereas previous versions would spit out a warning and then list your entire library.

There's more detail than you could ever need [on the beets blog](http://beets.radbox.org/blog/flexattr.html).

This is a bugfix release. We're in the midst of preparing for a large change in beets 1.3, so 1.2.2 resolves some issues that came up over the last few weeks. Stay tuned!

The improvements in this release are:
- A new plugin event, `item_moved`, is sent when files are moved on disk. Thanks to dsedivec.
- `lyrics`: More improvements to the Google backend by Fabrice Laporte.
- `bpd`: Fix for a crash when searching, thanks to Simon Chopin.
- Regular expression queries (and other query types) over paths now work. (Previously, special query types were ignored for the `path` field.)
- `fetchart`: Look for images in the Cover Art Archive for the release group in addition to the specific release. Thanks to Filipe Fortes.
- Fix a race in the importer that could cause files to be deleted before they were imported. This happened when importing one album, importing a duplicate album, and then asking for the first album to be replaced with the second. The situation could only arise when importing music from the library directory and when the two albums are imported close in time.

This release introduces a major internal change in the way that similarity scores are handled. It means that the importer interface can now show you exactly why a match is assigned its score and that the autotagger gained a few new options that let you customize how matches are prioritized and recommended.

The refactoring work is due to the continued efforts of Tai Lee. The changes you'll notice while using the autotagger are:
- The top 3 distance penalties are now displayed on the release listing, and all album and track penalties are now displayed on the track changes list. This should make it clear exactly which metadata is contributing to a low similarity score.
- When displaying differences, the colorization has been made more consistent and helpful: red for an actual difference, yellow to indicate that a distance penalty is being applied, and light gray for no penalty (e.g., case changes) or disambiguation data.

There are also three new (or overhauled) configuration options that let you customize the way that matches are selected:
- The `ignored` setting lets you instruct the importer not to show you matches that have a certain penalty applied.
- The `preferred` collection of settings specifies a sorted list of preferred countries and media types, or prioritizes releases closest to the original year for an album.
- The `max_rec` settings can now be used for any distance penalty component. The recommendation will be downgraded if a non-zero penalty is being applied to the specified field.

And some little enhancements and bug fixes:
- Multi-disc directory names can now contain "disk" (in addition to "disc"). Thanks to John Hawthorn.
- `web`: Item and album counts are now exposed through the API for use with the Tomahawk resolver. Thanks to Uwe L. Korn.
- Python 2.6 compatibility for the `beatport` plugin, `missing`, and `duplicates`. Thanks to Wesley Bitter and Pedro Silva.
- Don't move the config file during a null migration. Thanks to Theofilos Intzoglou.
- Fix an occasional crash in the `beatport` when a length field was missing from the API response. Thanks to Timothy Appnel.
- `scrub`: Handle and log I/O errors.
- `lyrics`: The Google backend should now turn up more results. Thanks to Fabrice Laporte.
- `random`: Fix compatibility with Python 2.6. Thanks to Matthias Drochner.