% dtas-player_protocol(7) dtas user manual
dtas-player_protocol - protocol for controling dtas-player
NOTE - NOTE - NOTE - NOTE - NOTE - NOTE - NOTE - NOTE
I'm considering a heavy revamp of this protocol. The "OK" responses for
a lot of commands may not be necessary since this is Unix sockets and
not TCP, and I may move away from a request-response model and towards
an entirely listen/notification model. I have little experience in
non-TCP-based application protocols, so this is an area of
experimentation for me.
This must stay over Unix sockets because filesystem permissions are
needed to enforce code execution permissions. dtas-player is really a
shell in disguise, after all.
Protocol feedback greatly appreciated email me at
This gives a specification of the dtas-player protocol over a local Unix
SOCK_SEQPACKET socket. The dtas-player protocol should NOT be
considered stable at this point and compatibility will break.
Inspiration is taken from MPRIS and MPRIS 2.0 specifications (e.g.
play_pause, play, pause), and there will be a proxy in the future to
support MPRIS/MPRIS 2.0 clients.
The DTAS_PLAYER_SOCK is the standard environment determining the control
socket for dtas-player(1). This defaults to $HOME/.dtas/player.sock if
Most low-level commands may be issued using the dtas-ctl(1) command.
Higher-level commands such as dtas-console(1), dtas-sourceedit(1),
dtas-sinkedit(1), and dtas-enq(1) also implement this protocol.
# ARGUMENT TYPES
- BOOLEAN - must be "true" or "false"
- INTEGER - a signed integer in decimal notation (base 10)
- UNSIGNED - an unsigned integer in decimal or hex notation
- ENVNAME - must be a suitable environment variable (setenv(3))
- ENVVALUE - must be a suitable environment variable (setenv(3))
- COMMAND, this may be quoted string passed to sh -c "",
variable/argument expansion will be performed by the shell
- SOURCENAME - "sox" or "av", more backends may be supported in the future
- TIMESTAMP - a time stamp formatted in HH:MM:SS.FRAC (for seeking)
- TRACKID - a unique unsigned integer in decimal (base-10) representing a
track in the tracklist
- FILENAME - an expanded pathname relative to / is recommended since
dtas-player and the client may run in different directories
# VARIABLE EXPANSION
For source and sink "command" arguments, the $SOXFMT and $ECAFMT
environment variables are exported automatically to source to ease
integration with sox(1) and ecasound(1).
Both $SOXFMT and $ECAFMT are based on the configured "format" of
the dtas-player (see below).
For all machines, $SOXFMT defaults to: -ts32 -c2 -r44100
For little-endian machines, $ECAFMT defaults to: -fs32_le,2,44100
Commands here should be alphabetized according to `LC_ALL=C sort'
* cd - change the current working directory of the player
* clear - clear current queue (current track/command continues running)
PENDING: this may be renamed to "queue clear" or "queue-clear"
* cue - display the index/offsets of the file based on the embedded
cue sheet, if any
* cue next - skip to the next section of the track based on the
embedded cue sheet. This may skip to the next track if there is
no embedded cue sheet or if playing the last (embedded) track
* cue prev - skip to the previous section of the track based on
the embedded cue sheet. This may just seek to the beginning
if there is no embedded cue sheet or if we are playing the first
* cue goto INTEGER - go to the embedded track with cue index denoted
by INTEGER (0 is first track as returned by "cue"). Negative
values of INTEGER allows selecting track relative to the last
track (-1 is the last track, -2 is the penultimate, and so on).
* current - output information about the currently-playing track/command
in YAML. The structure of this is unstable and subject to change.
* enq FILENAME - enqueue the given FILENAME for playback
An expanded (full) pathname relative to '/' is recommended, as
dtas-player and the client may be in different directories.
PENDING: this may be renamed to "queue add"
* enq-cmd "COMMAND" - run the following command for playback.
The COMMAND is expected to output audio in the audio format matching
the current audio format of the player. This may be a shell pipeline
and include multiple commands. The $SOXFMT and $ECAFMT variables are
PENDING: this may be renamed to "queue add-cmd"
* env ENVTOSET=ENVVALUE ENVTOSET2=ENVVALUE2
Set environment variables. This affects all future source/sink
processes as well as helper commands dtas-player may spawn
(e.g. soxi(1)). Environment variables set this way are currently not
preserved across invocations of dtas-player(1), but may change in the
* env ENVTOUNSET1# ENVTOUNSET#
Unset the given environment variable.
PENDING: the '#' is ugly and inconsistent with the per-sink/source.
* format FORMATARGS - configure the format between source and sink
Changing this will affect the $SOXFMT and $ECAFMT environments passed
to source and sink commands. Changing this implies a "restart"
Changing rate to 48000 is probably useful if you plan on playing to some
laptop sound cards. In all cases where "bypass" is supported, it
removes the guarantee of gapless playback as the audio device(s)
will likely need to be restarted.
+ channels=(UNSIGNED|bypass) - (default: 2 (stereo)) - number of
channels to use internally. sox will internally invoke the remix
effect when decoding. This supports the value "bypass" (without
quotes) to avoid the automatic remix effect. Using "bypass" mode
removes the guarantee of gapless playback, as the audio device will
likely need to be restarted, introducing an audible gap.
+ endian=(big|little|swap) - (default: native) - there is probably no
point in changing this unless you output over a network sink to
a machine of different endianess.
+ bits=(UNSIGNED|bypass) - (default: implied from type) - sample precision
This may be pointless and removed in the future, since the sample
precision is implied from type. This supports the value of "bypass"
to avoid dither/truncation in later stages.
+ rate=(UNSIGNED|bypass) - (default: 44100) - sample rate of audio
Typical values of rate are 44100, 48000, 88200, 96000. Not all
DSP effects are compatible with all sampling rates/channels.
This supports the value of "bypass" as well to avoid introducing
software resamplers into the playback chain.
+ type=(s16|s24|s32|u16|u24|u32|f32|f64) - (default: s32)
change the raw PCM format. s32 currently offers the best performance
when only sox/play are used. f32 may offer better performance if
piping to/from non-sox applications (e.g. ecasound)
* pause - pause playback
Player pause state is preserved across dtas-player invocations.
* play - restart playback from pause. Playback sinks will yield
control of the audio playback device once no source is playing.
* play_pause - toggle the play/pause state. This starts playback if
paused, and pauses playback if playing.
* restart - restarts all processes in the current pipeline. Playback
will be momentarily interrupted while this change occurs. This is
necessary if one of the commands (e.g. sox or ecasound) or loaded
libraries (e.g. a LADSPA plugin) is upgraded. Use "source restart"
instead to only restart the source chain, leaving the sinks
* rg RGARGS - configure ReplayGain support
All FLOAT values may be adjusted via '+=' or '-=' instead of simple
assignment ('='). If RGARGS is empty, the current rg state of
non-default values will be dumped in YAML.
+ fallback_gain=FLOAT (-6.0) - dB value
Adjust the volume by this level (usually negative) for tracks
missing ReplayGain tags. This is useful when the queue contains
a mix of tracks with and without ReplayGain tags.
+ fallback_track=BOOLEAN (true)
When in album_gain mode, fallback to track_gain if the
REPLAYGAIN_ALBUM_GAIN metadata is missing.
This controls the ReplayGain tag to use. The *_norm options
are used for peak normalization and not commonly found in other
+ preamp=FLOAT (0) - dB value
Adjust the album_gain or track_gain amount by this value (in dB).
+ norm_level=FLOAT (1.0 == dBFS)
Controls the level to normalize to when using album_norm or track_norm.
* seek [+-]TIMESTAMP - seek the current track to a specified time.
This is passed directly as the first argument for the sox(1) "trim"
command. See the sox(1) manpage for details.
Seeking to a relative time is also supported by prefixing the time
with '+' or '-'
* skip - abort current track/command (via closing the output pipe)
Running the "clear" command before this will abort playback.
* sink ls - list names of current sinks
* sink cat SINKNAME - dump SINKNAME config in YAML
* sink rm SINKNAME - remove SINKNAME
* sink ed SINKNAME SINKARGS - create/edit SINKNAME
This currently does not restart running (active) sinks.
This will stop active sinks if active is set to false, and start
active sinks if active is set to true.
See dtas-sinkedit(1) for an example of using this.
+ command=COMMAND - change the command-line used for playback
+ active=BOOLEAN - whether or not the sink will be in use (default: false)
+ env.ENVNAME=ENVVALUE - set ENVNAME to ENVVALUE for the sink process
+ env#ENVNAME - unset ENVNAME in the sink process (only)
+ prio=INTEGER - priority of the sink, lower values run first
+ nonblock=BOOLEAN - drop audio data to avoid holding back other sinks
+ pipe_size=UNSIGNED - set the size of the pipe for the sink (Linux-only)
* source cat SOURCENAME - dump the current source command and env in YAML
* source ed SOURCENAME SOURCEARGS - edit the source parameters.
This changes here will immediately restart the source process.
See the code for dtas-sourceedit(1) for an example of using this.
+ command=COMMAND - change the command-line used to decode audio
+ env.ENVNAME=ENVVALUE - set ENVNAME to ENVVALUE for the source process
+ env#ENVNAME - unset ENVNAME in the source process (only)
+ tryorder=INTEGER - lower values are tried first
PENDING: tryorder here is wrong and may be removed or changed.
We need to account for at least two variables input file:
1. input type (flac/opus/mp3/etc)
2. transport protocol (local FS/http/ftp/sftp/etc)
* source ls - dump the names of sources sorted by tryorder
* source restart - restart the current source command
This can be useful if the source file is changed during playback
and the current player process is holding onto an unlinked inode.
This is advantageous over a full "restart" as there is no audible
gap on most systems.
* state dump [FILENAME]
Immediately dump the state of the player. If a FILENAME is specified,
the state is written to that file. Otherwise, the default state file
(configured via DTAS_PLAYER_STATE environment variable, defaulting
to ~/.dtas/player_state.yml) is written to. This does not use fsync(2),
users requiring fsync should open(2) that file and fsync(2) it
themselves if necessary.
* tl add FILENAME [TRACKID [BOOLEAN]] - add files to the tracklist
With one arg, adds FILENAME to the head of the tracklist.
If TRACKID is specified, FILENAME is added immediately after TRACKID
on the existing tracklist. The final BOOLEAN argument replaces the
currently playing track with the newly-added one.
Returns the TRACKID of the newly added track
* tl current - display the pathname to the currently playing track
* tl current-id - display the TRACKID of the currently playing track
* tl remove TRACKID - remove the track with the given TRACKID from
the track list
* tl get [TRACKIDS]
returns a list of TRACKIDS mapped to shell-escaped filenames.
* tl goto TRACKID [TIMESTAMP] - plays the given TRACKID
An optional timestamp may be added to prevent playing the
same part(s) repeated.y
* tl next - jump to the next track in the tracklist
* tl prev - jump to the previous track in the tracklist
* tl repeat [BOOLEAN|1] - show/or change repeat status of the tracklist.
With no args, this will show "true", "false", or "1"
If set to "1", dtas-player will repeat the current track.
* tl tracks
returns a list of all TRACKIDS in the tracklist
* trim [off|TBEG [TLEN]]
Limits playback of all tracks in the tracklist to the time starting
at TBEG and ending after TLEN has elapsed. Not specifying TLEN will
cause sox. Like the sox "trim" effect, prefixing TLEN with a '='
will be interpreted as an absolute stop time.
Both TBEG and TLEN should be specified in seconds, not sample counts.
Specifying "off" (without quotes) disables trim.
This feature is intended to allow users to "zoom-in" on a particular
portion of a track to tweak parameters (either with
dtas-sourceedit(1) or via playback of splitfx YAML files) and often
combined with looping the tracklist (via "tl repeat").
* watch - adds the client to the passive watch list for notifications.
It is recommended clients issue no further commands and open
another client socket to issue non-watch commands.
All feedback welcome via plain-text mail to: <firstname.lastname@example.org>\
Mailing list archives available at <http://80x24.org/dtas-all/> and
No subscription is necessary to post to the mailing list.
Copyright 2013-2015 all contributors <email@example.com>.\
License: GPLv3 or later <http://www.gnu.org/licenses/gpl-3.0.txt>