Beets

This release adds two new standard plugins to beets: one for synchronizing Last.fm listening data and one for integrating with Linux desktops. And at long last, imports can now create symbolic links to music files instead of copying or moving them. We also gained the ability to search for album art on the iTunes Store and a new way to compute ReplayGain levels.

The major new features are:
- A new `lastimport` lets you download your play count data from Last.fm into a flexible attribute. Thanks to Rafael Bodill.
- A new `freedesktop` creates metadata files for Freedesktop.org-compliant file managers. Thanks to kerobaros. 1056, 707
- A new link option in the `import` section creates symbolic links during import instead of moving or copying. Thanks to Rovanion Luckey. 710, 114
- `fetchart`: You can now search for art on the iTunes Store. There's also a new `sources` config option that lets you choose exactly where to look for images and in which order.
- `replaygain`: A new Python Audio Tools backend was added. Thanks to Francesco Rubino. 1070
- `embedart`: You can now automatically check that new art looks similar to existing art---ensuring that you only get a better "version" of the art you already have. See image-similarity-check.
- `ftintitle`: The plugin now runs automatically on import. To disable this, unset the `auto` config flag.

There are also core improvements and other substantial additions:
- The `media` attribute is now a _track-level field_ instead of an album-level one. This field stores the delivery mechanism for the music, so in its album-level incarnation, it could not represent heterogeneous releases---for example, an album consisting of a CD and a DVD. Now, tracks accurately indicate the media they appear on. Thanks to Heinz Wiesinger.
- Re-imports of your existing music now preserve its added date and flexible attributes. Thanks to Stig Inge Lea Bjørnsen.
- Slow queries, such as those over flexible attributes, should now be much faster when used with certain commands---notably, the `play`.
- `bpd`: Add a new configuration option for setting the default volume. Thanks to IndiGit.
- `embedart`: A new `ifempty` config option lets you only embed album art when no album art is present. Thanks to kerobaros.
- `discogs`: Authenticate with the Discogs server. The plugin now requires a Discogs account due to new API restrictions. Thanks to multikatt. 1027, 1040

And countless little improvements and fixes:
- Standard cover art in APEv2 metadata is now supported. Thanks to Matthias Kiefer. 1042
- `convert`: Avoid a crash when embedding cover art fails.
- `mpdstats`: Fix an error on start (introduced in the previous version). Thanks to Zach Denton.
- `convert`: The `--yes` command-line flag no longer expects an argument.
- `play`: Remove the temporary .m3u file after sending it to the player.
- The importer no longer tries to highlight partial differences in numeric quantities (track numbers and durations), which was often confusing.
- Date-based queries that are malformed (not parse-able) no longer crash beets and instead fail silently.
- `duplicates`: Emit an error when the `checksum` config option is set incorrectly.
- The migration from pre-1.1, non-YAML configuration files has been removed. If you need to upgrade an old config file, use an older version of beets temporarily.
- `discogs`: Recover from HTTP errors when communicating with the Discogs servers. Thanks to Dustin Rodriguez.
- `embedart`: Do not log "embedding album art into..." messages during the import process.
- Fix a crash in the autotagger when files had only whitespace in their metadata.
- `play`: Fix a potential crash when the command outputs special characters. 1041
- `web`: Queries typed into the search field are now treated as separate query components. 1045
- Date tags that use slashes instead of dashes as separators are now interpreted correctly. And WMA (ASF) files now map the `comments` field to the "Description" tag (in addition to "WM/Comments"). Thanks to Matthias Kiefer. 1043
- `embedart`: Avoid resizing the image multiple times when embedding into an album. Thanks to kerobaros. 1028, 1036
- `discogs`: Avoid a situation where a trailing comma could be appended to some artist names. 1049
- The output of the `stats` command is slightly different: the approximate size is now marked as such, and the total number of seconds only appears in exact mode.
- `convert`: A new `copy_album_art` option puts images alongside converted files. Thanks to Ángel Alonso. 1050, 1055
- There is no longer a "conflict" between two plugins that declare the same field with the same type. Thanks to Peter Schnebel. 1059 1061
- `chroma`: Limit the number of releases and recordings fetched as the result of an Acoustid match to avoid extremely long processing times for very popular music. 1068
- Fix an issue where modifying an album's field without actually changing it would not update the corresponding tracks to bring differing tracks back in line with the album. 856
- `echonest`: When communicating with the Echo Nest servers fails repeatedly, log an error instead of exiting. 1096
- `lyrics`: Avoid an error when the Google source returns a result without a title. Thanks to Alberto Leal. 1097
- Importing an archive will no longer leave temporary files behind in `/tmp`. Thanks to multikatt. 1067, 1091

This release has two big new chunks of functionality. Queries now support **sorting** and user-defined fields can now have **types**.

If you want to see all your songs in reverse chronological order, just type `beet list year-`. It couldn't be easier.

Flexible field types mean that some functionality that has previously only worked for built-in fields, like range queries, can now work with plugin- and user-defined fields too. For starters, the /plugins/echonest/ and `mpdstats` now mark the types of the fields they provide---so you can now say, for example, `beet ls liveness:0.5..1.5` for the Echo Nest "liveness" attribute. The `types` makes it easy to specify field types in your config file.

One upgrade note: if you use the `discogs`, you will need to upgrade the Discogs client library to use this version. Just type `pip install -U discogs-client`.

Other new features:
- `info`: Target files can now be specified through library queries (in addition to filenames). The `--library` option prints library fields instead of tags. Multiple files can be summarized together with the new `--summarize` option.
- `mbcollection`: A new option lets you automatically update your collection on import. Thanks to Olin Gay.
- `convert`: A new `never_convert_lossy_files` option can prevent lossy transcoding. Thanks to Simon Kohlmeyer.
- `convert`: A new `--yes` command-line flag skips the confirmation.

Still more fixes and little improvements:
- Invalid state files don't crash the importer.
- `lyrics`: Only strip featured artists and parenthesized title suffixes if no lyrics for the original artist and title were found.
- Fix a crash when reading some files with missing tags.
- `discogs`: Compatibility with the new 2.0 version of the [discogs_client](https://github.com/discogs/discogs_client) Python library. If you were using the old version, you wil need to upgrade to the latest version of the library to use the correspondingly new version of the plugin (e.g., with `pip install -U discogs-client`). Thanks to Andriy Kohut.
- Fix a crash when writing files that can't be read. Thanks to Jocelyn De La Rosa.
- The `stats` command now counts album artists. The album count also more accurately reflects the number of albums in the database.
- `convert`: Avoid crashes when tags cannot be written to newly converted files.
- Formatting templates with item data no longer confusingly shows album-level data when the two are inconsistent.
- Resuming imports and beginning incremental imports should now be much faster when there is a lot of previously-imported music to skip.
- `lyrics`: Remove `<script>` tags from scraped lyrics. Thanks to Bombardment.
- `play`: Add a `relative_to` config option. Thanks to BrainDamage.
- Fix a crash when a MusicBrainz release has zero tracks.
- The `--version` flag now works as an alias for the `version` command.
- `lastgenre`: Remove some unhelpful genres from the default whitelist. Thanks to gwern.
- `importfeeds`: A new `echo` output mode prints files' paths to standard error. Thanks to robotanarchy.
- `replaygain`: Restore some error handling when `mp3gain` output cannot be parsed. The verbose log now contains the bad tool output in this case.
- `convert`: Fix filename extensions when converting automatically.
- The `write` plugin event allows plugins to change the tags that are written to a media file.
- `zero`: Do not delete database values; only media file tags are affected.

This release of beets fixes all the bugs, and you can be confident that you will never again find any bugs in beets, ever. It also adds support for plain old AIFF files and adds three more plugins, including a nifty one that lets you measure a song's tempo by tapping out the beat on your keyboard. The importer deals more elegantly with duplicates and you can broaden your cover art search to the entire web with Google Image Search.

The big new features are:
- Support for AIFF files. Tags are stored as ID3 frames in one of the file's IFF chunks. Thanks to Evan Purkhiser for contributing support to [Mutagen](https://bitbucket.org/lazka/mutagen).
- The new `importadded` reads files' modification times to set their "added" date. Thanks to Stig Inge Lea Bjørnsen.
- The new `bpm` lets you manually measure the tempo of a playing song. Thanks to aroquen.
- The new `spotify` generates playlists for your [Spotify](https://www.spotify.com/) account. Thanks to Olin Gay.
- A new required configuration option for the importer skips matches that are missing certain data. Thanks to oprietop.
- When the importer detects duplicates, it now shows you some details about the potentially-replaced music so you can make an informed decision. Thanks to Howard Jones.
- `fetchart`: You can now optionally search for cover art on Google Image Search. Thanks to Lemutar.
- A new asciify-paths configuration option replaces all non-ASCII characters in paths.

And the multitude of little improvements and fixes:
- Compatibility with the latest version of [Mutagen](https://bitbucket.org/lazka/mutagen), 1.23.
- `web`: Lyrics now display readably with correct line breaks. Also, the detail view scrolls to reveal all of the lyrics. Thanks to Meet Udeshi.
- `play`: The `command` config option can now contain arguments (rather than just an executable). Thanks to Alessandro Ghedini.
- Fix an error when using the modify command to remove a flexible attribute. Thanks to Pierre Rust.
- `info`: The command now shows audio properties (e.g., bitrate) in addition to metadata. Thanks Alessandro Ghedini.
- Avoid a crash on Windows when writing to files with special characters in their names.
- `play`: Playing albums now generates filenames by default (as opposed to directories) for better compatibility. The `use_folders` option restores the old behavior. Thanks to Lucas Duailibe.
- Fix an error when importing an empty directory with the `--flat` option.
- `mpdstats`: The last song in a playlist is now correctly counted as played. Thanks to Johann Klähn.
- `zero`: Prevent accidental nulling of dangerous fields (IDs and paths). Thanks to brunal.
- The remove command now shows the paths of files that will be deleted. Thanks again to brunal.
- Don't display changes for fields that are not in the restricted field set. This fixes write showing changes for fields that are not written to the file.
- The write command avoids displaying the item name if there are no changes for it.
- When using both the `convert` and the `scrub`, avoid scrubbing the source file of conversions. (Fix a regression introduced in the previous release.)
- `replaygain`: Logging is now quieter during import. Thanks to Yevgeny Bezman.
- `fetchart`: When loading art from the filesystem, we now prioritize covers with more keywords in them. This means that `cover-front.jpg` will now be taken before `cover-back.jpg` because it contains two keywords rather than one. Thanks to Fabrice Laporte.
- `lastgenre`: Remove duplicates from canonicalized genre lists. Thanks again to Fabrice Laporte.
- The importer now records its progress when skipping albums. This means that incremental imports will no longer try to import albums again after you've chosen to skip them, and erroneous invitations to resume "interrupted" imports should be reduced. Thanks to jcassette.
- `bucket`: You can now customize the definition of alphanumeric "ranges" using regular expressions. And the heuristic for detecting years has been improved. Thanks to sotho.
- Already-imported singleton tracks are skipped when resuming an import.
- `chroma`: A new `auto` configuration option disables fingerprinting on import. Thanks to ddettrittus.
- `convert`: A new `--format` option to can select the transcoding preset from the command-line.
- `convert`: Transcoding presets can now omit their filename extensions (extensions default to the name of the preset).
- `convert`: A new `--pretend` option lets you preview the commands the plugin will execute without actually taking any action. Thanks to Dietrich Daroch.
- Fix a crash when a float-valued tag field only contained a `+` or `-` character.
- Fixed a regression in the core that caused the `scrub` not to work in `auto` mode. Thanks to Harry Khanna.
- The write command now has a `--force` flag. Thanks again to Harry Khanna.
- `mbsync`: Track alignment now works with albums that have multiple copies of the same recording. Thanks to Rui Gonçalves.

This is primarily a bugfix release, but it also brings two new plugins: one for playing music in desktop players and another for organizing your directories into "buckets." It also brings huge performance optimizations to queries—your `beet ls` commands will now go much faster.

New features:
- The new `play` lets you start your desktop music player with the songs that match a query. Thanks to David Hamp-Gonsalves.
- The new `bucket` provides a `%bucket{}` function for path formatting to generate folder names representing ranges of years or initial letter. Thanks to Fabrice Laporte.
- Item and album queries are much faster.
- `ftintitle`: A new option lets you remove featured artists entirely instead of moving them to the title. Thanks to SUTJael.

And those all-important bug fixes:
- `mbsync`: Fix a regression in 1.3.5 that broke the plugin entirely.
- Shell completion now searches more common paths for its `bash_completion` dependency.
- Fix encoding-related logging errors in `convert` and `replaygain`.
- `replaygain`: Suppress a deprecation warning emitted by later versions of PyGI.
- Fix a crash when reading files whose iTunes SoundCheck tags contain non-ASCII characters.
- The `%if{}` template function now appropriately interprets the condition as false when it contains the string "false". Thanks to Ayberk Yilmaz.
- `convert`: Fix conversion for files that include a video stream by ignoring it. Thanks to brunal.
- `fetchart`: Log an error instead of crashing when tag manipulation fails.
- `convert`: Log an error instead of crashing when embedding album art fails.
- `convert`: Embed cover art into converted files. Previously they were embedded into the source files.
- New plugin event: `before_item_moved`. Thanks to Robert Speicher.

This is a short-term release that adds some great new stuff to beets. There's support for tracking and calculating musical keys, the ReplayGain plugin was expanded to work with more music formats via GStreamer, we can now import directly from compressed archives, and the lyrics plugin is more robust.

One note for upgraders and packagers: this version of beets has a new dependency in [enum34](https://pypi.python.org/pypi/enum34), which is a backport of the new [enum](https://docs.python.org/3.4/library/enum.html) standard library module.

The major new features are:
- Beets can now import zip, tar and rar archives. Just type `beet import music.zip` to have beets transparently extract the files to import.
- `replaygain`: Added support for calculating ReplayGain values with GStreamer as well the mp3gain program. This enables ReplayGain calculation for any audio format. Thanks to Yevgeny Bezman.
- `lyrics`: Lyrics should now be found for more songs. Searching is now sensitive to featured artists and parenthesized title suffixes. When a song has multiple titles, lyrics from all the named songs are now concatenated. Thanks to Fabrice Laporte and Paul Phillips.

In particular, a full complement of features for supporting musical keys are new in this release:
- A new `initial_key` field is available in the database and files' tags. You can set the field manually using a command like `beet modify initial_key=Am`.
- The `echonest` plugin sets the `initial_key` field if the data is available.
- A new `keyfinder` runs a command-line tool to get the key from audio data and store it in the `initial_key` field.

There are also many bug fixes and little enhancements:
- `echonest`: Truncate files larger than 50MB before uploading for analysis.
- `fetchart`: Fix a crash when the server does not specify a content type. Thanks to Lee Reinhardt.
- `convert`: The `--keep-new` flag now works correctly and the library includes the converted item.
- The importer now logs a message instead of crashing when errors occur while opening the files to be imported.
- `embedart`: Better error messages in exceptional conditions.
- Silenced some confusing error messages when searching for a non-MusicBrainz ID. Using an invalid ID (of any kind—Discogs IDs can be used there too) at the "Enter ID:" importer prompt now just silently returns no results. More info is in the verbose logs.
- `mbsync`: Fix application of album-level metadata. Due to a regression a few releases ago, only track-level metadata was being updated.
- On Windows, paths on network shares (UNC paths) no longer cause "invalid filename" errors.
- `replaygain`: Fix crashes when attempting to log errors.
- The modify-cmd command can now accept query arguments that contain = signs. An argument is considered a query part when a : appears before any =s. Thanks to mook.

This release brings a hodgepodge of medium-sized conveniences to beets. A new `config` command manages your configuration, we now have bash completion, and the `modify` command can delete attributes. There are also some significant performance optimizations to the autotagger's matching logic.

One note for upgraders: if you use the fetchart plugin, it has a new dependency, the [requests](http://www.python-requests.org/) module.

New stuff:
- Added a `config` command to manage your configuration. It can show you what you currently have in your config file, point you at where the file should be, or launch your text editor to let you modify the file. Thanks to geigerzaehler.
- Beets now ships with a shell command completion script! See completion. Thanks to geigerzaehler.
- The `modify` command now allows removing flexible attributes. For example, `beet modify artist:beatles oldies!` deletes the `oldies` attribute from matching items. Thanks to brilnius.
- Internally, beets has laid the groundwork for supporting multi-valued fields. Thanks to geigerzaehler.
- The importer interface now shows the URL for MusicBrainz matches. Thanks to johtso.
- smartplaylist: Playlists can now be generated from multiple queries (combined with "or" logic). Album-level queries are also now possible and automatic playlist regeneration can now be disabled. Thanks to brilnius.
- echonest: Echo Nest similarity now weights the tempo in better proportion to other metrics. Also, options were added to specify custom thresholds and output formats. Thanks to Adam M.
- Added the `after_write` plugin event.
- lastgenre: Separator in genre lists can now be configured. Thanks to brilnius.
- We now only use "primary" aliases for artist names from MusicBrainz. This eliminates some strange naming that could occur when the languages config option was set. Thanks to Filipe Fortes.
- The performance of the autotagger's matching mechanism is vastly improved. This should be noticeable when matching against very large releases such as box sets.
- The `import` command can now accept individual files as arguments even in non-singleton mode. Files are imported as one-track albums.

Fixes:
- Error messages involving paths no longer escape non-ASCII characters (for legibility).
- Fixed a regression that made it impossible to use the `modify` command to add new flexible fields. Thanks to brilnius.
- echonest: Avoid crashing when the audio analysis fails. Thanks to Pedro Silva.
- duplicates: Fix checksumming command execution for files with quotation marks in their names. Thanks again to Pedro Silva.
- Fix a crash when importing with both of the `group_albums` and `incremental` options enabled. Thanks to geigerzaehler.
- Give a sensible error message when `BEETSDIR` points to a file. Thanks again to geigerzaehler.
- Fix a crash when reading WMA files whose boolean-valued fields contain strings. Thanks to johtso.
- fetchart: The plugin now sends "beets" as the User-Agent when making scraping requests. This helps resolve some blocked requests. The plugin now also depends on the [requests](http://www.python-requests.org/) Python library.
- The `write` command now only shows the changes to fields that will actually be written to a file.
- duplicates: Spurious reports are now avoided for tracks with missing values (e.g., no MBIDs). Thanks to Pedro Silva.
- The default replace sanitation options now remove leading whitespace by default. Thanks to brilnius.
- importfeeds: Fix crash when importing albums containing `/` with the `m3u_multi` format.
- Avoid crashing on Mutagen bugs while writing files' tags.
- convert: Display a useful error message when the FFmpeg executable can't be found.