Powerful configuration management for Scala (JSON, properties, command-line arguments, and environment variables)

3.4.16

In any case where there are existing libraries that accomplish a task it is worthwhile to document the justification for creating yet another library. This is beneficial both for users to understand how it is differentiated and for the developers to clarify if there is valid purpose in the endeavor.

In the Scala configuration arena the most popular offering is that of Typesafe Config (https://github.com/typesafehub/config). While this is a powerful and useful library it is more complicated to work with and less flexible than we'd like. One of the very specific problems with it is the lack of support for Scala.js, but the larger issue is the distinction that arises from considering configuration coming from files vs environment variables vs command-line arguments vs any other origin of configuration that any modern application may want to utilize.

Our goal is primarily simplicity. A configuration library should do the necessary work and get out of the way of the developer and let them get their job done. To this end we support a unified configuration merging command-line arguments, environment variables, system properties, and configuration files to provide maximum flexibility of defining, defaulting, and overriding configuration in your application.

• JSON
• Properties
• YAML
• HOCON
• XML

Profig is published to Sonatype OSS and synchronized to Maven Central supporting JVM and Scala.js on 2.11, 2.12, 2.13, and Scala 3.x:

libraryDependencies += "com.outr" %% "profig" % "3.4.16"   // Scala
libraryDependencies += "com.outr" %%% "profig" % "3.4.16"  // Scala.js / Cross-Build

Whether you are using this in JVM or JS you need one import to access everything:

import profig._

This brings some implicits on specific platforms (for example, loading URLs, Files, Sources, etc. in the JVM) but the only class you really need be concerned with is Profig.

As of version 3.0, you now need to initialize Profig in order to fully utilize it:

Profig.init()

While Profig has a lot of very powerful features, the most useful feature is its ability to provide multiple levels of configuration. Instead of locking yourself to only file configuration, environment variables, or command-line arguments, Profig allows you hierarchical overrides to be able to specify configuration at any level.

Configuration is loaded level-by-level with configuration at the lower-level overriding that at a higher level:

• File Configuration
• Environment Variables
• Command-Line arguments

For example: If the application is executed with a config.json of:

{
  "server": {
    "port": 8888
  }
}

Profig("server.port").as[Int]
// res2: Int = 8888

However, if there were an environment variable SERVER_PORT=8889:

Profig("server.port").as[Int]
// res4: Int = 8889

Finally, if a command-line property were set, for example:

runCommand -server.port 8890

Profig("server.port").as[Int]
// res6: Int = 8890

So, to summarize, command-line arguments take the highest priority, followed by environment variables, and finally configuration files.

The easiest way (explained in greater detail below) to properly initialize Profig:

Profig.initConfiguration()  // Loads everything except the command-line arguments
Profig.merge(args)          // This can be a `List[String]` or `Array[String]` depending on your application

When your application starts it is reasonable to want to allow execution of the application to override existing configuration via the command-line. In order to effectively do this we can simply invoke Profig.merge(args) within our main method. This will merge all command-line arguments into Profig.

Note that the signature of merge is def merge(json: Json, ``type``: MergeType = MergeType.Overwrite): Unit. If you set the type to MergeType.Add, existing configuration will not be overwritten. This is useful for default configuration loading.

Profig supports many configuration formats and can look in the classpath as well as the filesystem to find configuration to load. Of course, this is only supported on the JVM, but to load a file simply call:

Profig.loadFile(new java.io.File("config.json"))

This will look for config.json on the filesystem, load it as JSON, and merge it into the configuration. The signature of loadFile is: Profig.loadFile(file: File, mergeType: MergeType = MergeType.Overwrite, errorHandler: Option[Throwable => Unit] = None)

However, if your application doesn't need very explicit files to be loaded you can load defaults instead:

Profig.loadConfiguration()

This will look for any standardized configuration file in the classpath and filesystem and load it into the system.

You can also use Profig.initConfiguration() to initialize and load configuration in a single call.

Finally, you can use Profig.initConfigurationBlocking() if you want initialization and loading to block before continuing with your application.

As stated above, system properties and environment variables are automatically loaded into the configuration. So if we wanted to access the system property "java.version" we can easily do so:

val javaVersion = Profig("java.version").as[String]
// javaVersion: String = "22.0.2"

You can also load from a higher level as a case class to get more information. For example:

import fabric.rw._

case class JVMInfo(version: String, specification: Specification)

object JVMInfo {
  implicit val rw: RW[JVMInfo] = RW.gen
}

case class Specification(vendor: String, name: String, version: String)

object Specification {
  implicit val rw: RW[Specification] = RW.gen
}

val info = Profig("java").as[JVMInfo]
// info: JVMInfo = JVMInfo(
//   version = "22.0.2",
//   specification = Specification(
//     vendor = "Oracle Corporation",
//     name = "Java Platform API Specification",
//     version = "22"
//   )
// )

Configuration files will automatically be loaded from config.json, config.conf, configuration.json, configuration.conf, application.conf, and application.json if found in the application path or in the classpath.

If default values or Option values are defined in the case class they will be used if the value is not available in the config. However, if any required parameters are missing an exception will be thrown when attempting to read.

Adding values at runtime is almost exactly the same as reading values. For example, if we want to store a basic configuration:

import fabric._

case class MyConfig(path: String = "/my/application",
                    timeout: Long = 1000L,
                    username: String = "root",
                    password: String = "password")
                    
object MyConfig {
  implicit val rw: RW[MyConfig] = RW.gen
}

val json: Json = MyConfig(path = "/another/path").json
// json: Json = {"path": "/another/path", "timeout": 1000, "username": "root", "password": "password"}
Profig.merge(json)

If you prefer to merge in an object without overwriting existing values you can use defaults instead of merge:

val myConfig: Json = MyConfig(path = "/another/path").json
// myConfig: Json = {"path": "/another/path", "timeout": 1000, "username": "root", "password": "password"}
Profig.merge(json, MergeType.Add)

This only scratches the surface of the features and functionality Profig provides. For additional information read the ScalaDocs and the specs: https://github.com/outr/profig/blob/master/core/shared/src/test/scala/spec/ProfigSpec.scala