问题
What are package objects, not so much the concept but their usage?
I've tried to get an example working and the only form I got to work was as follows:
package object investigations {
val PackageObjectVal = "A package object val"
}
package investigations {
object PackageObjectTest {
def main(args: Array[String]) {
println("Referencing a package object val: " + PackageObjectVal)
}
}
}
Observations I've made so far are:
package object _root_ { ... }
is disallowed (which is reasonable),
package object x.y { ... }
is also disallowed.
It seems that a package object must be declared in the immediate parent package and, if written as above, the brace delimited package declaration form is required.
Are they in common use? If so, how?
回答1:
Normally you would put your package object in a separate file called package.scala
in the package that it corresponds to. You can also use the nested package syntax but that is quite unusual.
The main use case for package objects is when you need definitions in various places inside your package as well as outside the package when you use the API defined by the package. Here is an example:
// file: foo/bar/package.scala
package foo
package object bar {
// package wide constants:
def BarVersionString = "1.0"
// or type aliases
type StringMap[+T] = Map[String,T]
// can be used to emulate a package wide import
// especially useful when wrapping a Java API
type DateTime = org.joda.time.DateTime
type JList[T] = java.util.List[T]
// Define implicits needed to effectively use your API:
implicit def a2b(a: A): B = // ...
}
Now the definitions inside that package object are available inside the whole package foo.bar
. Furthermore the definitions get imported when someone outside of that package imports foo.bar._
.
This way you can prevent to require the API client to issue additional imports to use your library effectively - e.g. in scala-swing you need to write
import swing._
import Swing._
to have all the goodness like onEDT
and implicit conversions from Tuple2
to Dimension
.
回答2:
While Moritz's answer is spot on, one additional thing to note is that package objects are objects. Among other things, this means you can build them up from traits, using mix-in inheritance. Moritz's example could be written as
package object bar extends Versioning
with JodaAliases
with JavaAliases {
// package wide constants:
override val version = "1.0"
// or type aliases
type StringMap[+T] = Map[String,T]
// Define implicits needed to effectively use your API:
implicit def a2b(a: A): B = // ...
}
Here Versioning is an abstract trait, which says that the package object must have a "version" method, while JodaAliases and JavaAliases are concrete traits containing handy type aliases. All of these traits can be reused by many different package objects.
回答3:
You could do worse than to go straight to the source. :)
https://lampsvn.epfl.ch/trac/scala/browser/scala/trunk/src/library/scala/package.scala
https://lampsvn.epfl.ch/trac/scala/browser/scala/trunk/src/library/scala/collection/immutable/package.scala
回答4:
The main use case for package objects is when you need definitions in various places inside your package as well as outside the package when you use the API defined by the package.
Not so with Scala 3, scheduled to be released mid-2020, based on Dotty, as in here:
Toplevel Definitions
All kinds of definitions can be written on the toplevel.
Package objects are no longer needed, will be phased out.
package p
type Labelled[T] = (String, T)
val a: Labelled[Int] = ("count", 1)
def b = a._2
def hello(name: String) = println(i"hello, $name)
来源:https://stackoverflow.com/questions/3400734/package-objects