sciss / scalacolliderugens   1.21.4

Contributors Wanted GitHub

UGen database and code synthesizer to generate class files for ScalaCollider. Mirror of

Scala versions: 3.x 2.13 2.12 2.11 2.10
Scala.js versions: 1.x


Gitter Build Status Maven Central


Specification and base API of ScalaCollider UGens, as well as a core library of generated UGen classes.

This project is (C)opyright 2008–2022 by Hanns Holger Rutz. All rights reserved. All subprojects released under the GNU LGPL v2.1+, except for the specification which is released under a BSD-style license. All code comes with absolutely no warranties. To contact the author, send an e-mail to contact at

sub projects

  • spec contains an XML specification of the standard SuperCollider UGens, suitable for synthesising ScalaCollider UGen classes or other meta-data purposes
  • api contains the base classes for ugens, graph elements, ugen and synth graphs.
  • gen generates the source codes of the core project from the descriptions provided by the spec project
  • core contains ScalaCollider classes for the standard UGens


All artifacts are published to Maven Central, and are available as follows:

"de.sciss" %  "scalacolliderugens-spec" % v
"de.sciss" %% "scalacolliderugens-api"  % v
"de.sciss" %% "scalacolliderugens-core" % v
"de.sciss" %% "scalacolliderugens-plugins" % v

The current stable version v is "1.21.4".

The spec contains the XML meta-data, api contains basic types without specific UGens, core contains the standard UGens included with SuperCollider, and plugins will include the third-party plugins managed by the sc3-plugins project (still incomplete).


The project builds with sbt against Scala 2.13, 2.12, Dotty (JVM) and Scala 2.13 (JS). The last version to support Scala 2.11 was 1.19.5.

The synthetic UGen sources are inside directories core/gen and plugins/gen. Since v1.20.1, when the XML specifications are changes, these must be recreated, by running sbt gen/ugen! The synthetic sources are currently checked into git, thus you do not necessarily have to regenerate these files.

To compile all sources, run sbt compile.


Please see the file

generating additional UGen class files

The UGen descriptions reside in XML files. The spec subproject contains files for the standard UGens included with a plain SuperCollider installation. You will need to create additional XML files if you wish to compile sources for third party UGens.

To synthesize the source code for a given UGen description XML file, run as follows:

$ sbt
> project gen
> run -d path/to/scala/source/output path/to/descriptions.xml

The generated source files then need to be compiled against core.

If you want to contribute UGen descriptions, there are two existing sbt subprojects to consider:

  • core creates class files for the UGens included in the standard SuperCollider distribution. If you are not editing an existing XML file, you need to ensure a new XML file is included in UGenSpec.standardPlugins.
  • plugins creates class files for the UGens included in the sc3-plugins umbrella project. Again, if you are not editing an existing XML file, you need to ensure a new XML file is included, this time in the list UGenSpec.thirdPartyPlugins.
  • if you are planning to contribute the description of a plugin that is neither in the standard UGens nor in the sc3-plugins project, please consult first where those classes should be published. Most likely, we will add another sbt subproject similar to the plugins project. If you look at its code, you will see that you have to change the call to runUGenGenerator so that instead of args = "--plugins" :: Nil you provide an explicit list of paths to the XML files you wish to generate sources from.

Please have a good look at the current XML files and try to follow the documentation style used there. In particular

  • try to be precise about the technical description of the behaviour of the UGen and its arguments. Try to think of the typical case where you would consult the API doc of the UGen. Which are the behavioural aspects you would want to look up, such as initialisation conditions, precise triggering behaviour, interaction of parameters, differences to related UGens.
  • each argument should be described in terms of its behaviour, the physical units (if any), the allowed or expected value ranges, and corner cases.
  • ideally, each UGen should be documented along with a concise example (in the XML file).

format of ugen XML descriptions

There is no DTD yet. But the structure of the XML file is as follows:

<ugens revision="<num>">
  <ugen name="UGenName" [ ugenAttrs ]>
    <rate name="RateName" [ implied="true" [ method="MethodName" ]]>
      [ <arg ... /> ]
    <rate ... />
    [ <output ... /> ]
    <arg name="ArgumentName" [ argAttrs ]>
      [ <doc>Argument description</doc> ]
    <arg ... />
    [ <doc>
      <text>UGen description</text>
      [ <example name="description">
          Example code demonstrating the UGen
        <example ... /> ]
      [ <see>ugen.RelatedUGenName</see>
        <see ... /> ]
      </doc> ]
  [ <ugen ... /> ]

All UGens within one file are considered to be part of that particular .scx plugin. Their synthesized classes will also be grouped in a file by that name.

UGen Attributes

UGen Attributes (ugenAttr) are boolean flags (all false by default) which can be set to characterize a UGen:

Attribute Name Meaning when value is "true" Example
reads-bus UGen reads from a bus In
writes-bus UGen writes to a bus Out
reads-buf UGen reads audio buffer data BufRd
writes-buf UGen overwrites audio buffer data BufWr
reads-fft UGen reads from an FFT buffer IFFT
writes-fft UGen writes to an FFT buffer FFT
done-flag UGen sets a 'done flag' Line
side-effect UGen has another side effect, for example causing a 'done-action', sending OSC commands, or printing to the console SendTrig
random UGen depends on random seeding WhiteNoise
indiv Each UGen is otherwise individual, even with identical inputs Demand UGens advance their inputs
helper A helper element that is not a genuine UGen itself Nyquist
optimized A UGen that might be optimized at runtime to other UGens (allows to skip rate specification) MulAdd
sourcecode Manually written source code is provided Nyquist
fragment UGen that cannot be fully represented in the spec. For example, it has hidden or currently not represented argument types LocalBuf
elem Client facing class name differing from UGen JPverbRaw

Part of this information is used by ScalaCollider when building the UGen graph. For example, subtrees which do not have any side effects are automatically removed. UGens which have side effects are those for which either of the following flags is set: writes-bus | writes-buf | writes-fft | side-effect. Furthermore, multiple occurrences of UGens which are functionally equivalent are collapsed. UGens are functionally not equivalent if either of the following flags is set: any of the side effects | any of the resource readers | random | indiv. That is to say, if there are two WhiteNoise UGens, they are functionally distinct by definition and will thus not be collapsed. The same is true for two Out UGens, even if their inputs are the same, as they have accumulative side effects on the bus to which they write. On the other hand, two SinOsc UGens with the same frequency and rate inputs are functionally equivalent and thus one can be replaced for the other.

More flags and meta data are planned in future version, e.g. oscillator signal ranges, filter coefficients.

UGen Rates

The possible rate names are "scalar", "control", "audio", and "demand". Each supported rate should have its own element. There are three extra attributes, implied, method, and method-alias.

implied says that the UGen not only has exactly one supported rate (an exception is thrown if you have a UGen with multiple rate elements and an implied attribute), but that this a natural precondition for the type of UGen. That way, the case class for that UGen does not carry a rate argument, but mixes in a trait which provides it. As a consequence, there is no argument for the rate when using pattern matching against that UGen. For example, K2A makes only sense at audio rate, A2K makes only sense at control rate, FreeVerb and Pitch make only sense at audio rate. Using this attribute, we have case class K2A(in: GE) (with mixin AudioRated) instead of the redundant case class K2A(rate: Rate, in: GE).

Be very careful with this attribute, it should not be used if another rate could be added in a future SuperCollider version, as this would break binary compatibility. This is why implied has been removed from DiskIn, for example (there is no reason, why DiskIn could not support control rate reading in the future).

The second attribute, method, builds up implied and requires that implied has been specified. It states that instead of the default method name in the companion object—ir for scalar rate, kr for control rate, ar for audio rate, and dr for demand rate—an alternative method name is used. The method name is typically apply, so that instead of, sig) you have to write FFT.apply(buf, sig) or short FFT(buf, sig), which is more convenient.

method-alias adds an additional method for the rate. An example is IFFT which specifies <rate name="audio" method-alias="apply"/>. This means the default method ar is created, plus an apply method as an alias.

Argument Attributes

Attribute Name Value Example
default Default expression for the argument. This can be either a number literal or a special string such "nyquist" or "doNothing" (see below) 440, 1.0, inf
type Argument type when it is not ge (generic graph element). (See below) PanAz, Poll
rate Constrains the supported rate for this argument. The only values presently recognized are "ugen" which means the argument is required to run at the same rate as the UGen, or a rate name such as "audio" which enforces that particular rate. Please see also the section below about rate specific argument settings. DiskOut (in)

The following table lists the allowed type values, and corresponding ways of defining default values. If the default value is unambiguous, the type is automatically inferred, e.g. using default="high" implies a type="trig". If the type and default value are incompatible, the parser will throw an exception.

Type name Description Example defaults
ge (default) Generic graph element -1.0, 440.0
ge-int Graph element used as integer -1, 18
ge-string String converted to variadic float constants "poll"
bus Bus index no default allowed
buf Buffer identifier no default allowed
fft FFT buffer phase chain signal no default allowed
trig Trigger signal (transition <= 0 to >0) low, high
switch Off/On signal (zero versus non-zero) false, true
gate Gating signal (open above zero) closed, open
mul Synthetic multiplier input 1.0
action Done action freeSelf, doNothing
done-flag UGen which sets a done flag no default allowed
int Static integer (no graph element) -1, 18

A special default value "nyquist" can be used which is understood as Note that expressions such as "60.midicps" have been currently disallowed for simplicity and language neutrality.

The following three argument attributes have boolean values, and are "false" by default:

Attribute Name Meaning when value is "true" Example
ugen-in Forces an Int type argument to be used as actual UGen input and not just auxiliary type. MFCC
variadic Indicates an argument which expands over multiple UGen inputs. RecordBuf (in), Dseq (seq)
prepend-size Must be combined with variadic: the variadic arg's size is prepended as additional UGen input. PackFFT

Arguments should be chosen careful not to conflict with methods available on GEOps. This is the reason, why various arguments which are named rate in SCLang have been renamed for example to speed, freq etc. It is recommended to take a look at the naming of the arguments in the default plugins (rather than relying on the naming in SCLang which is often unreflected and irregular) and try to reuse them whenever possible, and to be as consistent as possible with abbreviations. Care is also needed with the default values. There are some default values in SCLang which are insensible, while other useful defaults are missing. The aim is not to provide default values for every possible argument, but to require to fill in arguments for which defaults do not make sense.

Argument Positions

For some UGens, the actual positions of the arguments as they are coded in the underlying Plugin are either unintuitive (e.g. with respect to argument priority), inconvenient (e.g. with respect to default values), or irregular (e.g. different from an otherwise very similar other UGen). In those cases you are permitted to change the argument order as it is presented to the ScalaCollider user. To make sure the arguments are properly wired in the resulting SynthDef, explicit argument positions must be given.

If a UGen's arguments do not have pos attributes, they are considered in the order in which they appear in the XML file. Otherwise, the order of appearance in the XML file corresponds with the order in the underlying Plugin, whereas the values of the pos attributes specify the positions as presented to the user (counting from zero). Please read the previous sentence very carefully, as a common mistake is to falsely believe the correspondence to be the other way around.

As an example, consider XOut which has the unintuitive argument order of bus, followed by cross-fade level, followed by input signal. Compare this to Out which has the two arguments of bus, followed by input signal. We decided to make the XOut arguments appear to the user in the order of bus, then input signal (just like Out), then followed by the distinguishing parameter of the cross-fade level. Thus, we assign pos="1" to the in argument and pos="2" to the xfade argument, so they switch their positions. To minimize mistakes, ScalaCollider-UGens requires that we also add pos="0" to the bus argument, even if that does not affect its final position. The whole UGen specification thus becomes:

<ugen name="XOut" writes-bus="true">
   <rate name="audio">
      <arg name="in" rate="ugen"/>
   <rate name="control"/>
   <arg name="bus" pos="0"/>
   <arg name="xfade" pos="2"/>
   <arg name="in" variadic="true" pos="1"/>
   <doc warn-pos="true"/>

Note how the attribute warn-pos="true" was added to the doc element. This makes Scaladoc add an extra note to alert the user of the change in argument order. This is particularly important, as it may create confusion when coming from SCLang. It is recommended to apply argument reordering only after careful consideration, and to abstain from them when in doubt.

Rate Specific Argument Settings

Sometimes it is necessary to change the default value or the description of an argument with respect to the rate at which the UGen runs. And sometimes the rate constraint for an argument only applies to particular rates at which the UGen runs. In this case, it is permitted to embed an auxiliary arg element inside the rate element. This auxiliary arg element must have a corresponding element in the outer scope (inside the ugen element). Their correspondence is established by using the same name attribute, and the auxiliary element may provide an additional default or rate attribute and may contain an additional doc element.

As an example for different default values, here is the full text of LeakDC:

<ugen name="LeakDC">
   <rate name="control">
      <arg name="coeff" default="0.9"/>    <!-- provide a default value for the `kr` method -->
   <rate name="audio">
      <arg name="coeff" default="0.995"/>  <!-- provide a default value for the `ar` method -->
   <arg name="in" rate="ugen"/>
   <arg name="coeff"/>  <!-- the outer argument must still be provided -->

An example of restricting the argument's rate only in certain cases is Out:

<ugen name="Out" writes-bus="true">
   <rate name="audio">
      <arg name="in" rate="ugen"/>
   <rate name="control"/>
   <rate name="scalar"/>
   <arg name="bus"/>
   <arg name="in" variadic="true"/>

Here, the "outer" definition of argument in says that the argument is a multichannel argument, but it does not enforce a particular rate. Only for the case that Out is run at audio rate, the auxiliary entry for in enforces that in in this case must run at the same rate as the UGen (thus audio rate, too).


By default, the UGen is considered to have one monophonic output. All other UGens must explicitly contain either a <no-outputs/> element, or one or more <output ... /> elements. An output element may have a name and type attribute, and one element may have a variadic="<id>" attribute, where <id> is the name of the input argument determining the number of channels. A <doc> element may be nested inside a <output> node. Examples:

Example UGen Note
<no-outputs/> Out
<output name="left"/><output name="right"/> Pan2
<output variadic="in"/> Demand in is a GE type input
<output variadic="numChannels"/> DiskIn numChannels is an Int input
<output name="chain" type="fft"/> PV_MagShift


The description text for arguments is the text inside the argument's <doc></doc> element. The description text for the UGen is inside the <text></text> element inside the <doc></doc> element. In each case, standard Scaladoc formatting is allowed. Cross-links are provided through any number of <see></see> elements.

Please follow carefully the style of the descriptions used for the standard UGens. They adhere mostly to Javadoc style practice, and not so much to the more colloquial style of SCLang docs. The purpose here is not to include lengthy examples, but to be technically precise in the meanings of the argument values and the exact functioning of the UGen, if possible covering corner cases, providing details about underlying formulas, phase behavior of oscillators, typical ranges and scale.

Whenever the argument order has been significantly changed from the SCLang counterpart, the UGen's doc element should contain the attribute warn-pos="true" which will create a special highlight in the Scala docs to alert the reader of this change.

Adjunct types

In order to register all necessary deserialization factories, UGens may define adjunct types. An adjunct is defined by a reader class that implements ProductReader, and a number of product prefixes that are returned from the reader. For instance, the BinaryOpUGen and UnaryOpUGen each define their own Op types which have to be serialized and deserialized. In these cases, there is one reader and one returned type:

<adjunct reader="UnaryOpUGen.Op" self="true"/>

The reader object is UnaryOpUGen.Op and the prefixes supported solely consists of itself (it will be automatically translated to UnaryOpUGen$Op). Any additional types deserialized can be added within the adjunct element as prefix element with name attributes:

<adjunct reader="Env.Curve">
    <prefix name="Env$Curve$Apply" />
    <prefix name="Env$Curve$Const" />