ScaFi (Scala Fields)

ScaFi is a Scala-based library and framework for Aggregate Programming. It implements a variant of the Higher-Order Field Calculus (HOFC) operational semantics, which is made available as a usable domain-specific language (DSL), and provides a platform and API for simulating and executing Aggregate Computing systems and applications.

Please refer to the ScaFi main website and ScaFi Documentation the for further information.

Overview

Notes for application developers

• NOTE: scafi-core and scafi-simulator cross-target both the JVM and the JavaScript Platform (via Scala.js)

Import via Maven/sbt/Gradle

Add the dependency to scafi in your project, e.g., via sbt

// build.sbtvalscafi_version="1.1.5"valscafi_core="it.unibo.scafi"%%"scafi-core"% scafi_version
valscafi_simulator="it.unibo.scafi"%%"scafi-simulator"% scafi_version
valscafi_simulator_gui="it.unibo.scafi"%%"scafi-simulator-gui"% scafi_version
valscafi_platform="it.unibo.scafi"%%"scafi-distributed"% scafi_version
libraryDependencies ++=Seq(scafi_core, scafi_simulator, scafi_platform)

or Gradle

// build.gradle.kts
dependencies {
implementation("it.unibo.scafi:scafi-core_2.13:1.1.5")
}

Hello, ScaFi

• Consider the following repository: https://github.com/scafi/hello-scafi

As another example, consider the following steps.

Step 1: Import or define an incarnation (a family of types), from which you can import types like AggregateProgram

packageexperiments// Method #1: Use an incarnation which is already defined// (Note: BasicSimulationIncarnation is defined in module 'scafi-simulator')importit.unibo.scafi.incarnations.BasicSimulationIncarnation.AggregateProgram// Method #2: Define a custom incarnation and import stuff from itobjectMyIncarnationextends it.unibo.scafi.incarnations.BasicAbstractIncarnationimportMyIncarnation._

Step 2: Define an AggregateProgram which expresses the global behaviour of an ensemble.

// An "aggregate program" can be seen as a function from a Context to an Export// The Context is the input for a local computation: includes state // from previous computations, sensor data, and exports from neighbours.// The export is a tree-like data structure that contains all the information needed// for coordinating with neighbours. It also contains the output of the computation.objectMyAggregateProgramextendsAggregateProgramwithStandardSensorNames{
// Main program expression driving the ensemble// This is run in a loop for each agent// According to this expression, coordination messages are automatically generated// The platform/middleware/simulator is responsible for coordinationoverridedefmain() = gradient(isSource)
// The gradient is the (self-adaptive) field of the minimum distances from source nodes// `rep` is the construct for state transformation (remember the round-by-round loop behaviour)// `mux` is a purely functional multiplexer (selects the first or second branch according to condition)// `foldhoodPlus` folds over the neighbourhood (think like Scala's fold)// (`Plus` means "without self"--with plain `foldhood`, the device itself is folded)// `nbr(e)` denotes the values to be locally computed and shared with neighbours// `nbrRange` is a sensor that, when folding, returns the distance wrt each neighbourdefgradient(source: Boolean):Double=
rep(Double.PositiveInfinity){ distance =>
mux(source) { 0.0 } {
foldhoodPlus(Double.PositiveInfinity)(Math.min)(nbr{distance}+nbrRange)
}
}
// A custom local sensordefisSource= sense[Boolean]("source")
// A custom "neighbouring sensor"defnbrRange= nbrvar[Double](NBR_RANGE)
}

Step 3: Use ScaFi's internal simulator and GUI (modules scafi-simulator and scafi-simulator-gui, respectively) to run the program on a predefined network of devices.

importit.unibo.scafi.simulation.gui.{Launcher, Settings}
objectSimulationRunnerextendsLauncher {
Settings.Sim_ProgramClass="experiments.MyAggregateProgram"Settings.ShowConfigPanel=true
launch()
}

Alternatively, you can (a) implement your integration/middleware layer, or (b) leverages integration with the Alchemist simulator, for more sophisticated simulations. This and much more is described in the ScaFi Documentation Page.

ScaFi Architecture

From a deployment perspective, ScaFi consists of the following modules:

• scafi-commons: provides basic entities (e.g., spatial and temporal abstractions)
• scafi-core: represents the core of the project and provides an implementation of the ScaFi aggregate programming DSL, together with its standard library
• scafi-simulator: provides a basic support for simulating aggregate systems
• scafi-simulator-gui: provides a GUI for visualising simulations of aggregate systems
• spala: provides an actor-based aggregate computing middleware
• scafi-distributed: ScaFi integration-layer for spala

The modules to be imported (e.g., via sbt or Gradle) depend on the use case:

• Development of a real-world aggregate application. Bring scafi-core in for a fine-grained integration. For more straightforward distributed system setup, take a look at scafi-distributed.
• Play, exercise, and experiment with aggregate programming. Bring scafi-core in for writing aggregate programs as well as scafi-simulator-gui to quickly render an executing system.
• Set up sophisticated simulations Bring scafi-core in for writing aggregate programs and either (A) leverage the basic machinery provided by scafi-simulator, or (B) leverage the ScaFi support provided by Alchemist.

People

Main Researchers and Authors

• Roberto Casadei
• Gianluca Aguzzi
• Mirko Viroli

Research Collaborators

• Ferruccio Damiani
• Giorgio Audrito
• Danilo Pianini

Contributing

Please refer to the ScaFi Developer Manual.

License

scafi is Open Source and available under the Apache 2 License.