process

This package handles the execution of external processes. The contents of this package can be divided in three groups, according to their responsibilities:

• Indicating what to run and how to run it.
• Handling a process input and output.
• Running the process.

For simple uses, the only group that matters is the first one. Running an external command can be as simple as "ls".!, or as complex as building a pipeline of commands such as this:

import scala.sys.process._
"ls" #| "grep .scala" #&& "scalac *.scala" #|| "echo nothing found" lines

We describe below the general concepts and architecture of the package, and then take a closer look at each of the categories mentioned above.

Concepts and Architecture

The underlying basis for the whole package is Java's Process and ProcessBuilder classes. While there's no need to use these Java classes, they impose boundaries on what is possible. One cannot, for instance, retrieve a process id for whatever is executing.

When executing an external process, one can provide a command's name, arguments to it, the directory in which it will be executed and what environment variables will be set. For each executing process, one can feed its standard input through a java.io.OutputStream, and read from its standard output and standard error through a pair of java.io.InputStream. One can wait until a process finishes execution and then retrieve its return value, or one can kill an executing process. Everything else must be built on those features.

This package provides a DSL for running and chaining such processes, mimicking Unix shells ability to pipe output from one process to the input of another, or control the execution of further processes based on the return status of the previous one.

In addition to this DSL, this package also provides a few ways of controlling input and output of these processes, going from simple and easy to use to complex and flexible.

When processes are composed, a new ProcessBuilder is created which, when run, will execute the ProcessBuilder instances it is composed of according to the manner of the composition. If piping one process to another, they'll be executed simultaneously, and each will be passed a ProcessIO that will copy the output of one to the input of the other.

What to Run and How

The central component of the process execution DSL is the scala.sys.process.ProcessBuilder trait. It is ProcessBuilder that implements the process execution DSL, that creates the scala.sys.process.Process that will handle the execution, and return the results of such execution to the caller. We can see that DSL in the introductory example: #|, #&& and #!! are methods on ProcessBuilder used to create a new ProcessBuilder through composition.

One creates a ProcessBuilder either through factories on the scala.sys.process.Process's companion object, or through implicit conversions available in this package object itself. Implicitly, each process is created either out of a String, with arguments separated by spaces -- no escaping of spaces is possible -- or out of a scala.collection.Seq, where the first element represents the command name, and the remaining elements are arguments to it. In this latter case, arguments may contain spaces. One can also implicitly convert scala.xml.Elem and java.lang.ProcessBuilder into a ProcessBuilder. In the introductory example, the strings were converted into ProcessBuilder implicitly.

To further control what how the process will be run, such as specifying the directory in which it will be run, see the factories on scala.sys.process.Process's object companion.

Once the desired ProcessBuilder is available, it can be executed in different ways, depending on how one desires to control its I/O, and what kind of result one wishes for:

• Return status of the process (! methods)
• Output of the process as a String (!! methods)
• Continuous output of the process as a Stream[String] (lines methods)
• The Process representing it (run methods)

Some simple examples of these methods:

import scala.sys.process._

// This uses ! to get the exit code
def fileExists(name: String) = Seq("test", "-f", name).! == 0

// This uses !! to get the whole result as a string
val dirContents = "ls".!!

// This "fire-and-forgets" the method, which can be lazily read through
// a Stream[String]
def sourceFilesAt(baseDir: String): Stream[String] = {
  val cmd = Seq("find", baseDir, "-name", "*.scala", "-type", "f")
  cmd.lines
}

We'll see more details about controlling I/O of the process in the next section.

Handling Input and Output

In the underlying Java model, once a Process has been started, one can get java.io.InputStream and java.io.OutpuStream representing its output and input respectively. That is, what one writes to an OutputStream is turned into input to the process, and the output of a process can be read from an InputStream -- of which there are two, one representing normal output, and the other representing error output.

This model creates a difficulty, which is that the code responsible for actually running the external processes is the one that has to take decisions about how to handle its I/O.

This package presents an alternative model: the I/O of a running process is controlled by a scala.sys.process.ProcessIO object, which can be passed _to_ the code that runs the external process. A ProcessIO will have direct access to the java streams associated with the process I/O. It must, however, close these streams afterwards.

Simpler abstractions are available, however. The components of this package that handle I/O are:

• scala.sys.process.ProcessIO: provides the low level abstraction.
• scala.sys.process.ProcessLogger: provides a higher level abstraction for output, and can be created through its object companion
• scala.sys.process.BasicIO: a library of helper methods for the creation of ProcessIO.
• This package object itself, with a few implicit conversions.

Some examples of I/O handling:

import scala.sys.process._

// An overly complex way of computing size of a compressed file
def gzFileSize(name: String) = {
  val cat = Seq("zcat", "name")
  var count = 0
  def byteCounter(input: java.io.InputStream) = {
    while(input.read() != -1) count += 1
    input.close()
  }
  cat ! new ProcessIO(_.close(), byteCounter, _.close())
  count
}

// This "fire-and-forgets" the method, which can be lazily read through
// a Stream[String], and accumulates all errors on a StringBuffer
def sourceFilesAt(baseDir: String): (Stream[String], StringBuffer) = {
  val buffer = new StringBuffer()
  val cmd = Seq("find", baseDir, "-name", "*.scala", "-type", "f")
  val lines = cmd lines_! ProcessLogger(buffer append _)
  (lines, buffer)
}

Instances of the java classes java.io.File and java.net.URL can both be used directly as input to other processes, and java.io.File can be used as output as well. One can even pipe one to the other directly without any intervening process, though that's not a design goal or recommended usage. For example, the following code will copy a web page to a file:

import java.io.File
import java.net.URL
import scala.sys.process._
new URL("http://www.scala-lang.org/") #> new File("scala-lang.html") !

More information about the other ways of controlling I/O can be looked at in the scaladoc for the associated objects, traits and classes.

Running the Process

Paradoxically, this is the simplest component of all, and the one least likely to be interacted with. It consists solely of scala.sys.process.Process, and it provides only two methods:

• exitValue(): blocks until the process exit, and then returns the exit value. This is what happens when one uses the ! method of ProcessBuilder.
• destroy(): this will kill the external process and close the streams associated with it.