search

com-lihaoyi / sourcecode

Scala library providing "source" metadata to your program, similar to Python's __name__, C++'s __LINE__ or Ruby's __FILE__.
Other
521 stars 76 forks source link

readme

SourceCode Join the chat at https://gitter.im/lihaoyi/Ammonite Maven Central

sourcecode is a small Scala library that provides common "source code" context to your program at runtime, similar to Python's __name__, C++'s __LINE__ or Ruby's __FILE__. For example, you can ask for the file-name and line number of the current file, either through the () syntax or via an implicit:

val file = sourcecode.File()
assert(file.endsWith("/sourcecode/shared/src/test/scala/sourcecode/Tests.scala"))

val line = implicitly[sourcecode.Line]
assert(line == 16)

This might not be something you want to use for "business logic", but is very helpful for things like debugging, logging or providing automatic diagnostics for DSLs. This information is also available via an implicit, letting you write functions that automatically pull it in.

Using SourceCode on code dealing with lots of anonymous functions or anonymous classes can easily turn what you see in your debug printouts from this:

Before

To this:

After

By capturing source information you can use to give your objects and function meaningful names that tell you where they were defined, automatically without needing you to manually assign a string-ID to every anonymous function or anonymous class you define all over your code base.

If you like using Sourcecode, you might also enjoy this book by the author which teaches you Scala in a similarly simple and straightforward way:

Table of Contents

Overview

The kinds of compilation-time data that sourcecode provides are:

All these are available both via () and as implicits, e.g. sourcecode.File can be summoned via sourcecode.File() or implicitly[sourcecode.File].value. This also means you can define functions that pull in this information automatically:

def foo(arg: String)(implicit file: sourcecode.File) = {
  ... do something with arg ...
  ... do something with file.value ...
}

foo("hello") // the implicit sourcecode.File is filled in automatically

sourcecode does not rely on runtime reflection or stack inspection, and is done at compile-time using macros. This means that it is both orders of magnitude faster than e.g. getting file-name and line-numbers using stack inspection, and also works on Scala.js where reflection and stack inspection can't be used.

Download

Mill

sourcecode is published to Maven Central.

def ivyDeps = Agg(
  ivy"com.lihaoyi::sourcecode:0.4.2", // Scala-JVM
  ivy"com.lihaoyi::sourcecode::0.4.2" // Scala.js / Scala Native
)

sbt

"com.lihaoyi" %% "sourcecode" % "0.4.2" // Scala-JVM
"com.lihaoyi" %%% "sourcecode" % "0.4.2" // Scala.js / Scala Native

Examples

Here are a few examples of sourcecode's core functions being used in a variety of contexts. Hopefully they will give you an idea of how the various implicits behave:

package sourcecode

object Implicits {
  def implicitRun() = {
    val name = implicitly[sourcecode.Name]
    assert(name.value == "name")

    val fullName = implicitly[sourcecode.FullName]
    assert(fullName.value == "sourcecode.Implicits.fullName")

    val enclosing = implicitly[sourcecode.Enclosing]
    assert(enclosing.value == "sourcecode.Implicits.implicitRun enclosing")

    val pkg = implicitly[sourcecode.Pkg]
    assert(pkg.value == "sourcecode")

    val file = implicitly[sourcecode.File]
    assert(file.value.endsWith("/sourcecode/Implicits.scala"))

    val fileName = implicitly[sourcecode.FileName]
    assert(fileName.value == "Implicits.scala")

    val line = implicitly[sourcecode.Line]
    assert(line.value == 23)

    lazy val myLazy = {
      trait Bar{
        val name = implicitly[sourcecode.Name]
        assert(name.value == "name")

        val fullName = implicitly[sourcecode.FullName]
        assert(fullName.value == "sourcecode.Implicits.Bar.fullName")

        val file = implicitly[sourcecode.File]
        assert(file.value.endsWith("/sourcecode/Implicits.scala"))

        val fileName = implicitly[sourcecode.FileName]
        assert(fileName.value == "Implicits.scala")

        val line = implicitly[sourcecode.Line]
        assert(line.value == 40)

        val enclosing = implicitly[sourcecode.Enclosing]
        assert(
          (enclosing.value == "sourcecode.Implicits.implicitRun myLazy$lzy Bar#enclosing") ||
          (enclosing.value == "sourcecode.Implicits.implicitRun myLazy Bar#enclosing") // encoding changed in Scala 2.12
        )
      }
      val b = new Bar{}
    }
    myLazy
  }
}

Note that in "normal" usage you would not directly call implicitly to summon up sourcecode values; rather, you would add implicit parameters of these types to your functions. That would make these values automatically available to your functions without needing to manually keep passing them in. Apart from summoning them via implicits, you can also use the apply method on each type to pull them in using the () syntax:

package sourcecode

object Implicits {
  def implicitRun() = {
    val name = implicitly[sourcecode.Name]
    assert(name.value == "name")

    val fullName = implicitly[sourcecode.FullName]
    assert(fullName.value == "sourcecode.Implicits.fullName")

    val enclosing = implicitly[sourcecode.Enclosing]
    assert(enclosing.value == "sourcecode.Implicits.implicitRun enclosing")

    val pkg = implicitly[sourcecode.Pkg]
    assert(pkg.value == "sourcecode")

    val file = implicitly[sourcecode.File]
    assert(file.value.endsWith("/sourcecode/Implicits.scala"))

    val fileName = implicitly[sourcecode.FileName]
    assert(fileName.value == "Implicits.scala")

    val line = implicitly[sourcecode.Line]
    assert(line.value == 23)

    lazy val myLazy = {
      trait Bar{
        val name = implicitly[sourcecode.Name]
        assert(name.value == "name")

        val fullName = implicitly[sourcecode.FullName]
        assert(fullName.value == "sourcecode.Implicits.Bar.fullName")

        val file = implicitly[sourcecode.File]
        assert(file.value.endsWith("/sourcecode/Implicits.scala"))

        val fileName = implicitly[sourcecode.FileName]
        assert(fileName.value == "Implicits.scala")

        val line = implicitly[sourcecode.Line]
        assert(line.value == 40)

        val enclosing = implicitly[sourcecode.Enclosing]
        assert(
          (enclosing.value == "sourcecode.Implicits.implicitRun myLazy$lzy Bar#enclosing") ||
          (enclosing.value == "sourcecode.Implicits.implicitRun myLazy Bar#enclosing") // encoding changed in Scala 2.12
        )
      }
      val b = new Bar{}
    }
    myLazy
  }
}

By default, the various implicits all ignore any synthetic <init>, <local Foo> or $anonfun methods that might be present:

package sourcecode

object NoSynthetic {
  def run() = {
    class EnumValue(implicit name: sourcecode.Name){
      override def toString = name.value
    }
    object Foo extends EnumValue

    assert(Foo.toString == "Foo")

    object Bar{
      assert(sourcecode.Name() == "Bar")
      assert(sourcecode.FullName() == "sourcecode.NoSynthetic.Bar")
      assert(sourcecode.Enclosing() == "sourcecode.NoSynthetic.run Bar")
    }
    Bar
  }
}

If you want these synthetic methods to be shown, use the .Machine versions of each of these instead:

package sourcecode

object Synthetic {
  def run() = {
    class EnumValue(implicit name: sourcecode.Name.Machine){
      override def toString = name.value
    }
    object Foo extends EnumValue

    assert(Foo.toString == "<init>")

    object Bar{
      assert(sourcecode.Name.Machine() == "<local Bar>", sourcecode.Name())
      assert(sourcecode.FullName.Machine() == "sourcecode.Synthetic.Bar.<local Bar>")
      assert(sourcecode.Enclosing.Machine() == "sourcecode.Synthetic.run Bar.<local Bar>")
    }
    Bar
  }
}

Hopefully this has given you a reasonable feel for *what* sourcecode does. You may still be wondering why* we would want any of this: what could we possibly use these things for? Why would we want to write code that depends on our package paths or variable names? The section below will provide use cases that you will hopefully be able to relate to.

Use Cases

At first it might seem strange to make use of these source-level details in your program: shouldn't a program's meaning not change under re-formatting and re-factoring?

It turns out that there are a number of entirely valid use cases for this sort of information that is both extremely handy, and also would not be surprising at all to a developer using your API. Here are a few example use cases:

Logging

You can use sourcecode.File and sourcecode.Line to define log functions that automatically capture their line number and file-name

def log(foo: String)(implicit line: sourcecode.Line, file: sourcecode.File) = {
  println(s"${file.value}:${line.value} $foo")
}

log("Foooooo") // sourcecode/shared/src/test/scala/sourcecode/Tests.scala:86 Fooooo

This can be handy for letting you see where the log lines are coming from, without tediously tagging every log statement with a unique prefix. Furthermore, this happens at compile time, and is thus orders of magnitude faster than getting this information by generating stack traces, and works on Scala.js where stack-inspection does not. Lastly, if you want additional information such as method names, class names, or packages to be provided to your logging function, you can easily do so by asking for the sourcecode.Name or sourcecode.FullName or sourcecode.Pkg implicits.

Enums

You can use sourcecode.Name to define an enumeration-value factory function that automatically assigns names to the enum values based on the name of the val that it is assigned to

package sourcecode

object EnumExample {
  def run() = {
    case class EnumValue(name: String){
      override def toString = name
    }
    class Enum{
      def value(implicit name: sourcecode.Name) = EnumValue(name.value)
    }
    object MyEnum extends Enum{
      val firstItem = value
      val secondItem = value
    }
    assert(MyEnum.firstItem.toString == "firstItem")
    assert(MyEnum.secondItem.toString == "secondItem")
  }
}

This is very handy, and this functionality is used in a number of libraries such as FastParse and Scalatags to provide a boilerplate-free experience while still providing good debuggability and convenience.

Sometimes you want to make sure that different enum values in differently named enums (or even an enum of the same name in a different package!) are given unique names. In that case, you can use sourcecode.FullName or sourcecode.Enclosing to capture the full path e.g. "com.mypkg.MyEnum.firstItem" and "com.mypkg.MyEnum.secondItem":

package sourcecode

object EnumFull {
  def run() = {
    case class EnumValue(name: String){
      override def toString = name
    }
    class Enum{
      def value(implicit name: sourcecode.FullName) = EnumValue(name.value)
    }
    object MyEnum extends Enum{
      val firstItem = value
      val secondItem = value
    }
    assert(MyEnum.firstItem.toString == "sourcecode.EnumFull.MyEnum.firstItem")
    assert(MyEnum.secondItem.toString == "sourcecode.EnumFull.MyEnum.secondItem")
  }
}

You can also use sourcecode.Name in an constructor, in which case it'll be picked up during inheritance:

class EnumValue(implicit name: sourcecode.Name){
  override def toString = name.value
}
object Foo extends EnumValue
println(Foo.toString)
assert(Foo.toString == "Foo")

Debug Prints

How many times have you written tedious code like

object Bar{
  def foo(arg: String) = {
    println("Bar.foo: " + arg)
  }
}

Where you have to prefix every print statement with the name of the enclosing classes, objects or functions to make sure you can find your print output 2-3 minutes later? With source.Enclosing, you can get this for free:

def debug[V](value: sourcecode.Text[V])(implicit enclosing: sourcecode.Enclosing) = {
  println(enclosing.value + " [" + value.source + "]: " + value.value)
}

class Foo(arg: Int){
  debug(arg) // sourcecode.DebugRun.main Foo [arg]: 123
  def bar(param: String) = {
    debug(arg -> param)
  }
}
new Foo(123).bar("lol")  // sourcecode.DebugRun.main Foo#bar [arg -> param]: (123,lol)

You can easily vary the amount of verbosity, e.g. by swapping the sourcecode.Enclosing for a sourcecode.Name if you think it's too verbose:

def debug[V](value: sourcecode.Text[V])(implicit name: sourcecode.Name) = {
  println(name.value + " [" + value.source + "]: " + value.value)
}

class Foo(arg: Int){
  debug(arg) // Foo [arg]: 123
  def bar(param: String) = {
    debug(param -> arg)
  }
}
new Foo(123).bar("lol")  // bar [param]: lol

Or leaving it out entirely:

def debug[V](value: sourcecode.Text[V]) = {
  println("[" + value.source + "]: " + value.value)
}

class Foo(arg: Int){
  debug(arg) // [arg]: 123
  def bar(param: String) = {
    debug(param -> arg)
  }
}
new Foo(123).bar("lol")  // [param]: lol

Thus you can easily configure how much information your debug helper method needs, at its definition, without having to hunt all over your codebase for the various debug call-sites you left lying around and manually tweaking the verbosity of each one. Furthermore, if you want additional information like sourcecode.Line or sourcecode.File, that's all just one implicit away.

The PPrint library provides a pprint.log method that does exactly this: prints out the value provided (in this case pretty-printing it with colors and nice formatting & indentation) together with the enclosing context and line number, so you can easily distinguish your individual prints later:

scala> class Foo{
     |   def bar(grid: Seq[Seq[Int]]) = {
     |     // automatically capture and print out source context
     |     pprint.log(grid, tag="grid")
     |   }
     | }
defined class Foo

scala> new Foo().bar(Seq(0 until 10, 10 until 20, 20 until 30))
pkg.Foo#bar "grid":12
List(
  Range(0, 1, 2, 3, 4, 5, 6, 7, 8, 9),
  Range(10, 11, 12, 13, 14, 15, 16, 17, 18, 19),
  Range(20, 21, 22, 23, 24, 25, 26, 27, 28, 29)
)

pprint.log is itself defined as

def log[T: PPrint](value: T, tag: String = "")
                  (implicit cfg: Config = Config.Colors.PPrintConfig,
                   path: sourcecode.Enclosing,
                   line: sourcecode.Line) = ...

Using sourcecode.Enclosing and sourcecode.Line to provide the context to be printed. You can, or course, define your own log method in the same way, customizing it to print or not-print exactly what you want to see via the implicits that sourcecode provides!

sourcecode.Args can be used to access all parameters that where provided to a method:

def debug(implicit name: sourcecode.Name, args: sourcecode.Args): Unit = {
  println(name.value + args.value.map(_.map(a => a.source + "=" + a.value).mkString("(", ", ", ")")).mkString(""))
}

def foo(bar: String, baz: Int)(p: Boolean): Unit = {
  debug
}

foo("baz", 42)(true) // foo(bar=baz, baz=42)(p=true)

Embedding Domain-Specific Languages

Scala is a popular language in which to embed domain-specific languages: that means that you start with some external language, e.g. this MathProg example

param m;
param n;
param l;

set I := 1 .. m;
set J := 1 .. n;
set K := 1 .. l;

param c{J};
param d{K};
param a{I, J};

var x{J} integer, >= 0;
var y{K} >= 0;

The linked slides has more detail about what exactly this language does (it describes mathematical optimization problems). For a variety of reasons, you may prefer to write this as part of a Scala program instead: for example you may want Scala's IDE support, or its ability to define functions that help reduce boilerplate, or maybe you like the way the compiler provides type errors when you do the wrong thing.

A first attempt at converting this to Scala may look like this:

val m = param("m")
val n = param("n")
val l = param("l")

val I = set("I") := 1 to m
val J = set("J") := 1 to m
val K = set("K") := 1 to m

val c = param("c", J)
val d = param("d", K)
val a = param("a", I, J)

val x = xvar("x", J).integer >= 0
val y = xvar("y", K) >= 0

There's a bunch of duplication around the names of the vals: each val has its name repeated in a string that gets passed to the expression on the right. This is for the program to use the name of the val later: for example when printing error messages, or the results of the computation, you want to see which vals are involved! Thus you end up duplicating the names over and over and over.

With sourcecode, you can easily define param set and xvar as taking implicit sourcecode.Names, thus eliminating all the boilerplate involved in duplicating names:

val m = param
val n = param
val l = param

val I = set := 1 to m
val J = set := 1 to m
val K = set := 1 to m

val c = param(J)
val d = param(K)
val a = param(I, J)

val x = xvar(J).integer >= 0
val y = xvar(K) >= 0

The popular FastParse parser-combinator library uses sourcecode for exactly this use case

import fastparse.all._
val A = P( "aa" )
val B = P( "bb" )
val C = P( (A | B).rep(1) )

C.parse("aabb") // Success((), 4)
C.parse("X") // Failure((A | B):1:1 ..."X")

As you can see, the names of the rules A and B are embedded in the error messages for parse failures. This makes debugging parsers far easier, while saving you the effort of duplicating the name of the parser in possibly hundreds of rules in a large parser. In this case, it is the P(...) function which takes an implicit sourcecode.Name that does this work:

def P[T](p: => Parser[T])(implicit name: sourcecode.Name): Parser[T] =
    parsers.Combinators.Rule(name.value, () => p)

And forwards the name on to the actual Rule object, which can make use of it in its .toString method.

Version History

0.4.2

0.4.1

0.4.0

0.3.1

0.3.0

0.2.7

0.2.6

0.2.5

0.2.2

0.1.9

0.1.8

0.1.7

0.1.5

0.1.4

0.1.3

0.1.2

0.1.1

0.1.0