trackfs
is a read-only FUSE filesystem that splits audio files that contain full albums into individual FLAC files per track.
trackfs
supports three flavors of album files:
- FLAC files with embedded cuesheets (FLAC+CUE)
- FLAC files with accompanying cuesheet
- WAVE files with accompanying cuesheet
Accompanying cuesheets must use the fileextension .cue
and have the same basename as album file.
The recommended way to use trackfs
is using the docker image andresch/trackfs
. In case you want to use trackfs
without docker see section Manual Installation below.
You can directly run trackfs
on any Linux system with Docker and FUSE installed.
The simplest way to get familiar with trackfs
is to just launch it from the command-line:
docker run --rm \
--name=trackfs \
--device /dev/fuse \
--cap-add SYS_ADMIN \
--security-opt apparmor:unconfined \
-v /path/to/yourmusiclibrary:/src:ro \
-v /path/to/yourmountpoint:/dst:rshared \
andresch/trackfs \
--root-allowed
Replace /path/to/yourmusiclibrary
with the root directory where trackfs
scans for your album files and /path/to/yourmountpoint
with the directory that you want to use as mount point for the trackfs
-filesystem. Ideally the mount point already exists, if not, docker will create the directory (but then with root as owner)
Once started you will find all directories and files from your music library also in the trackfs
-filesystem. Only album files got replaced: Instead of a single album file you will find individual FLAC files for each track found in the cue sheet. The track-files will have the following names:
{album-file}.#-#.{tracknumber}.{track-title}.flac
While the tracks can be used like regular files, they don't exist in the physical file system on your machine. Instead trackfs
creates them on the fly whenever an application starts loading any of the track files. This usually takes (depending on your system) a few seconds.
In case you're not familiar with docker, a quick explanation on the used docker arguments:
-v /path/to/yourmusiclibrary:/src:ro
: make your music library accessible for trackfs by mounting it to /src in read-only mode inside your docker container-v /path/to/yourmountpoint:/dst:rshared
: share the trackfs filesystem (/dst
inside the container) accessible under your mount pointandresch/trackfs
: the name of thetrackfs
docker image on docker hub.--device
,--cap-add
--security-opt
: With those arguments you grant the docker container the privileges required to mount FUSE filesystems. You can try to leave out the--security-opt
option as it is not required on all systems. There is onging discussion if docker containers should allow mounting FUSE filesystems, by just using the--device
option, but for now this is not the case.--rm
: remove the orphaned container after termination
Please refer to the docker run documentation for more details.
While the above is working just fine, it is not the recommended way to use trackfs
as it runs trackfs
inside the docker container as user root. Running as root does allow trackfs
to access any file in your music library, irrespective of its underlying file permissions. If we would have omitted the --root-allowed
argument, trackfs
would have terminated with a corresponding error message.
Instead it is recommended to let trackfs
run as a regular user. For that to work we need a few changes:
-
Make sure that in your host system the file
/etc/fuse.conf
has the optionuser_allow_other
enabled, e.g. by calling from your command linesudo echo "user_allow_other" >> /etc/fuse.conf
-
Make sure that your mount point already exists and is owned by the user that is supposed to run
trackfs
. -
Use the docker run option
--user
to define the user that will runtrackfs
E.g. the following docker command would run
trackfs
with the current user:docker run --rm \ --name=trackfs \ --device /dev/fuse \ --cap-add SYS_ADMIN \ --security-opt apparmor:unconfined \ --user $(id -u):$(id -g) \ -v /path/to/yourmusiclibrary:/src:ro \ -v /path/to/yourmountpoint:/dst:rshared \ andresch/trackfs
trackfs
provides a few options that allow you to tweak its default behavior:
-e EXTENSION
,--extension EXTENSION
(default: "(\.flac|\.wav)") : A regular expression matching file extension of album files in the music library-s SEPARATOR
,--separator SEPARATOR
(default: ".#-#."): The separator used inside the name of the track-files. Must never occur in regular filenames-i IGNORE
,--ignore-tags IGNORE
(default: "CUE_TRACK.*|COMMENT"): A regular expression matching all tags in the FLAC+CUE file that will not be copied over to the track FLACs-k
,--keep-album
: Keep the source album file (FLAC+CUE or WAVE) in thetrackfs
filesystem in addition to the individual tracks-t TITLE_LENGTH
,--title-length TITLE_LENGTH
(default: 20): Nr. of characters of the track title in filename of track--root-allowed
: Allow runningtrackfs
as with root permissions; Neither necessary nor recommended. Use only when you know what you are doing-v
,--verbose
: Activate info-level logging-d
,--debug
: Activate debug-level logging
You can use -h
, --help
to get a list of all all options. Keep in mind that the parameters root
and mount
are already defined with the two -v
options to docker run
and are implicitly set by the docker container.
Most tags of the album file will be set in the track files too. There are only two exceptions:
- Tags that contain multi-line values (like the
CUESHEET
-tag) - Any tag whose name matches the regular expression of the
--ignore-tags
option (default:"CUE_TRACK.*|COMMENT"
)
In addition trackfs
does the following modifications to tags:
- If the album file contains an
ARTIST
tag but noALBUMARTIST
tag, then anALBUMARTIST
tag will be created with the value ofARTIST
tag. - If the album file contains a
TITLE
tag, but noALBUM
tag, then anALBUM
tag will be created with the value of theTITLE
tag. - If the cue sheet contains a
TITLE
tag for a given track, it overwrites theTITLE
tag from the album file - If the cue sheet contains a
PERFORMER
tag for a given track, it overwrites theARTIST
tag from the album file - If the cue sheet contains a
SONGWRITER
tag for a given track, it overwrites theCOMPOSER
tag from the album file - Any other information from the cue sheet will only be used to add their missing corresponding tags from the audio file, but never to overwrite an existing tag.
Cue sheet entries of type PERFORMER
, SONGWRITER
will be split at ";" characters and create multiple tags in the track file.
In case the album has embedded pictures, the first picture will be available in the track file. Alternatively trackfs
will scan the directory for a jpg-file with the same basename as the album file or a file named folder.jpg
In case you want/have to run trackfs
on some linux system without docker you can also install the python package trackfs
manually. Please refer to the homepage of the trackfs python package for further information.
trackfs
is currently in an early stage. While it runs stable on the author's NAS, it has not been tested in other environments, esp. on various Linux distributions with different kernels/FUSE versions. Using the dockerized version should remove some of the difficulties, but given the dependencies on FUSE, some my still remain.
Also keep in mind that this is the author's first python project, so don't expect that the source code matches professional quality criteria of experienced python coders.
There are a few ideas for additional improvements
- Find out if there is a way to extract tracks from the FLAC+CUE file without re-encoding the track. This should allow to increase the performance when starting to read a track massively
Allow encoding in other audio-formats (esp. mp3). While you can create a FUSE chain, by using mp3fs with the trackfs filesystem as source, the performance of that approach is not very compelling and a unified solution might provide bette results.(ffmpegfs has picked up the idea of cuesheets for splitting and transcoding; currently only in dev branch, but soon on master. No need to reinvent the wheel)
When trackfs
doesn't get properly terminated, then your system might still have an orphaned mount point. When you then restart trackfs
this will fail with a corresponding error message.
In that case you have to first unmount the orphaned mount point by calling:
sudo umount /path/to/yourmountpoint
In case the path to your mountpoint contains a symbolic link the above might not work as umount
expects the real path of the mount point. In that case use
mount -t fuse
to find the path that umount
expects.
In case trackfs
hangs (should not happen, but just in case) you might have to explicitly kill it. For that we use the docker stop
command (which give the container a chance to currently shutdown before killing it). This requires the container name or container id as parameter.
If you have defined a container name (e.g. to "trackfs") you can just use
docker stop trackfs
otherwise your first have to find the id of the container that runs the andresch/trackfs image:
docker stop $(docker ps | awk '/andresch\/trackfs/{print $1}')
trackfs
began its live as a clone of FLACCue. While FLACCue is designed for the usage with the Plex media server, the underlying idea of both projects is the same. Although there is little unmodified code of FLACCue left in trackfs
, the project would most likely not have been started without the ideas in this groundwork. Kudos go to acenkos!
trackfs
is licensed under the terms of the GNU Lesser General Public License v3.0