Similarly to how configuration works in Durden, configuration, scripting and scene definitions in Safespaces work via a simple virtual filesystem, and all user actions point into this filesystem in one way or another. All of the paths except for the "System Windows" category are also shared with the vrviewer tool in Durden.
Conceptually, Safespaces is divided into one or many spaces where only one space can be active at any one time. Each space have zero or many layers.
To map to normal window management concepts, a space can be thought of as a more radical 'workspace', and a layer as a 'tag' or group of addressable windows.
/toggle_grab : when arcan is running in windowed mode, activating
this path toggles mouse grab on / off which might be
needed in som environments.
/control=name : enable ipc control channel for menu access
/shutdown : immediately kill all clients and terminate
/switch/ : collapse safespaces and move clients to a new appl
/mouse=
selected : input forwarded to selected model
view : input modifies camera orientation
scale : input modifies focus selected model scale
move : buttons set axis, motion moves model relative to current position
rotate : input modifies selected model orientation
ipd : input modifies eye- distance and distortion
parameters, repeated calls cycles which parameter
that is targeted.
If the control channel is opened via /control=name a socket will be created in ipc/name through which arcan-cfgfs or socat can be used to manipulate and explore VR state during runtime.
socat unix-client:/path/to/ipc/name -
ls /
write /hmd-ipd=10.0
exec /hmd/reset
This path has controls for dynamically tuning active HMD devices.
/hmd/reset : resets the orientation and marks your current position
as the "origo / looking straight ahead" state
/hmd/ipd=f : override the IPD setting (distance between eyes)
/hmd/step_ipd=f : adjust the IPD relative to its current value
/hmd/distortion=
none : disable distortion step
basic : shader based OpenHMD distortion model
This path is special, and populated by globbing the spaces subdirectory for matching profiles. To load, specify /space/spacename.lua and it will activate.
Layers provide a positional anchor to which you attach and group models. Only one layer can be selected at a time, with one model as input focus. A layer has a 'positioner' attached to it. This positioner determines how the different objects will be positioned.
/layers/grow_shrink=f : increment or decrement the layouter spacing
/layers/push_pull=f : move the anchor closer or further away
/layers/add=name : create a new layer with the specified (unique) name
/layers/layer_name : name comes from the add=name call and is a submenu
/layers/current : refers to the currently selected layer.
/layers/cycle_input=n : switch input n layers forward or backwards (< 0)
Each individual layer accessed through /layers/layer_name or /layers/current has the following possible submenus, actions and values:
focus : set the layer as having input focus
terminal : add a rectangle model tied to a new terminal group
media : add a rectangle model tied to a media feed
remoting : add a rectangle model tied to a remote desktop session
swap=i : swap model to the left (< 0) or to the right (> 0) into focus
cycle=i : rotate models n steps on the left (< 0) or right (> 0) side
destroy : delete the layer and all associated models
destroy_protected : delete the layer unless there are protected models,
otherwise delete only non-protected models
opacity=f : (0..1) set the layer relative opacity, default=1 (opaque)
hide_show : switch layer visibility on/off (input focus remains)
nudge=f : move the layer anchor relative to the current position
models/ : subdirectory of all attached models
settings/ : parameters that affect layout and model attachment
add_model (see Model Creation below)
These settings appear under /layers/layer_name/settings/ and control hints to the layer "layouter" that adjusts positioning.
depth=f : set the distance from the anchor to the viewer
radius=f : set the circle that windows will be positioned through
spacing=f : set the spacing between window along the longitude
vspacing=f : set the spacing between windows along the y axis
active_scale=f : set the scale factor for the selected window
inactive_scale=f : set the scale factor for the inactive windows
fixed=b : (true | false) disable (=true) or enable autolayouting
ignore : ignore this layer when stepping input focus
These paths are relative to layers/layer_name where name is what you specified when adding it (e.g. /layers/add=background) would become:
/layers/layer_background/add_model/primitive=
You can also specify them relative to the currently selected layer:
/layers/current/add_model/primitive=
Substitute primitive with one out of:
pointcloud, sphere, hemisphere, rectangle, cylinder, halfcylinder, cube
And a unique name in order to reference the model later, similarly to how layers were created and referenced. So for example:
/layers/current/add_model/rectangle=screen
Would create a model in the currently active layer with a curved rectangular surface. It will not be visible yet since nothing has been attached to it.
The models are accessed via a layer (/layers/current/ or /layers/layer_name) within the models subdirectory (/layers/current/models/model_n) either by the model name:
/layers/current/models/my_model_name/
Or through the currently selected:
/layers/current/models/selected/
From this path you have access to the following properties:
rotate=fff : set absolute rotation for the object around x, y and z axis
spin=fff : relative adjust the rotation for the model
scale=f : set the absolute scale value fore the model
opacity=f : force-toggle blending for the entire model (0..1)
curvature=f : for models that can be bent (rectangle), adjust curvature
flip=b : change the source mapping from UL origo to LL origo (y-axis flip)
source_inv=s : specify a resource image to map to the object (y-inv)
source=s : specify a resource image to map to the object
stereoscopic= : (none, sbs, sbs-rl, oau, oau-rl) treat the source material as stereoscopic
layout_block=b: block model from layouter or not
move=fff : set layer-anchor relative position
nudge=fff : like move=fff but relative its current position
protected=b : block model from being deleted on space swap
faces/ : like source= for objects with multiple faces (cubes, spheres, ...)
events/ : actions to run on certain events
connpoint/
And the following actions: destroy : destroy the model and kill any external connections merge_collapse : change vertical hierarchy children to horizontal cycle_stereo : step between the different stereoscopic modes child_swap=i : -n .. +n, swap hierarchy position with its children (positioner hint)
The latest created model in any layer can also be accessed through the shortcut: /latest
Any model can have its contents (map) be defined dynamically via a 'connection point' that an external client can access via ARCAN_CONNPOINT=name. This is activated via the model access (see above) connpoint subdirectory.
replace=s : the connection point contents will permanently map to the model
temporary=s : when the connection point closes, the previous map will return
reveal=s : auto hide/show the model when the connection is activated
reveal_focus=s : like reveal, but automatically set layer input focus to it