What's the best way of grouping utility functions that don't belong in a class? In Ruby, I would have grouped them in a module. Should I use traits in Scala for the same effect, or objects?
Usually, I put utility functions that are semantically different into different traits and create an object for each trait, e.g.
trait Foo {
def bar = 1
}
object Foo extends Foo
That way I'm most flexible. I can import the utility functions via an import statement or via with in the class declaration. Moreover I can easily group different utility traits together into a new object to simplify the import statements for the most commonly used utility functions, e.g.
object AllMyUtilites extends Foo with Foo2
Package objects or just plain objects.
See, for instance, Scala.Predef and scala.math.
Traits if you want them to be mixed in with the classes that are going to use it. Objects if you want to only have to import them.
Related
I am new to Scala and I now started a project in Scala and I see similar to the following construct:
trait SomeTrait extends SomeOtherStuff with SomeOtherStuff2
object SomeTrait {
def someFunction():Unit = { ??? }
}
I understand that for a class, companion objects hold methods that are used in a "static", like Factory methods in Java or something alike, but what about traits, why not put these methods in traits?
The first style is called mixin, it used to be somewhat popular back in the days.
It could be replaced by the following code:
object SomeOtherStuff {
def someMethod(): String
}
object SomeObj {
import SomeOtherStuff._
//now someMethod is available
def otherMethod(): String = someMethod + "!"
}
object Caller {
import SomeObj._
import SomeOtherStuff._
//utility methods from both objects are available here
}
Pros of mixins:
If SomeTrait extends 10 other mixins then extending this trait would allow to scrap 10 import statements
Cons of mixins:
1) creates unnecessary coupling between traits
2) awkward to use if the caller doesn't extend the mixin itself
Avoiding mixins for business-logic code is a safe choice.
Although I'm aware of 2 legitimate usecases:
1) importing DSLs
e.g. ScalaTest code :
class SomeSuite extends FunSuite with BeforeAndAfter {...}
2) working (as a library author) with implicit parameters:
e.g. object Clock extends LowPriorityImplicits
(https://github.com/typelevel/cats-effect/blob/master/core/shared/src/main/scala/cats/effect/Clock.scala#L127)
Another perspective to this is the OOP principle Composition Over Inheritance.
Pros of companion objects (composition):
composition can be done at runtime while traits are defined at compile time
you can easily have multiple of them. You don't have to deal with the quirks of multiple inheritance: say you have two traits that both have a method with the name foo - which one is going to be used or does it work at all? For me, it's easier to see the delegation of a method call, multiple inheritance tends to become complex very fast because you lose track where a method was actually defined
Pros of traits (mixins):
mixins seem more idiomatic to reuse, a class using a companion object of another class is odd. You can create standalone objects though.
it's cool for frameworks because it adds the frameworks functionality to your class without much effort. Something like that just isn't possible with companion objects.
In doubt, I prefer companion objects, but it always depends on your environment.
I was trying to look into trait and object in scala when it seems like we can use trait and object to do a similar task.
What should be the guiding principles on when to use trait and when to use object?
Edit:
As many of you are asking for an example
object PercentileStats {
def addPercentile(df: DataFrame): DataFrame // implementation
}
trait PercentileStats {
def addPercentile(df: DataFrame): DataFrame // implementation
}
There is a Process class which can use the object
object Process {
def doSomething(df: DataFrame): DataFrame {
PercentileStats.addPercentile(df)
}
}
We can also make it use the trait
object Process with PercentileStats {
def doSomething(df: DataFrame): DataFrame {
addPercentile(df)
}
}
I think the real question here is Where do I put stand-alone functions?
There are three options.
In the package
You can put stand-alone functions in the outer package scope. This makes them immediately available to the whole package but the name has to be meaningful across the whole package.
def addPercentile(df: DataFrame): DataFrame // implementation
In an object
You can group stand-alone functions in an object to provide a simple namespace. This means that you have to use the name of the object to access the functions, but it keeps them out of the global namespace and allows the names to be simpler:
object PercentileStats {
def add(df: DataFrame): DataFrame // implementation
}
In a trait
You can group stand-alone functions in a trait. This also removes them from the package namespace, but allows them to be accessed without a qualifier from classes that have that trait. But this also makes the method visible outside the class, and allows them to be overridden. To avoid this you should mark them protected final:
trait PercentileStats {
protected final def addPercentile(df: DataFrame): DataFrame // implementation
}
Which is best?
The choice really depends on how the function will be used. If a function is only to be used in a particular scope then it might make sense to put it in a trait, otherwise the other options are better. If there are a number of related function then grouping them in an object makes sense. One-off functions for general use can just go in the package.
Object - is a class that has exactly one instance. It is created lazily when it is referenced, like a lazy val.
As a top-level value, an object is a singleton.
Traits - are used to share interfaces and fields between classes.
Classes and objects can extend while traits cannot be instantiated and therefore have no parameters.
So, it means that if you prefer singleton type implementation with no new instance happen then use Object but if you want to inherit implementation to other class or objects then you can use trait.
Traits: are equivalent to interfaces in Java. So you can use it to define public contracts like interfaces in Java. In addition, a trait can be used to share values (beside methods) between classes extends the trait.
Objects in Scala is actually quite flexible. Example use cases include:
singletons: If you think that your objects are singletons (exactly
one instance exists in the program), you can use object.
factory: for instance, companion object of a class can be used as factory for creating instances of the class.
to share static methods: for example, common utilities can be declared in one object.
You also have to consider how you would want to use / import it.
trait Foo {
def test(): String
}
object Bar extends Foo
import Bar._
Objects enable you to import rather than mix in your class.
It is a life saver when you want to mock - with scalamock - a class that mixes a lot of traits and expose more than 22 methods that you don't really need exposed in the scope.
I want to understand the difference between these 2 codes.
Lets look at the following 2 pieces of code in Scala.
trait Expressions {
// Bunch of functions
}
CODE 1:
trait Scheduling{
val IR : Expressions
import IR._
// Bunch of functions
}
CODE 2:
trait Scheduling extends Expressions {
// Same Bunch of functions
}
What is the difference between those 2 codes?
Thanks
Your first code snippet requires that any class that extends the trait has a variable named IR of type Expressions. All members of that variable are then imported so they can be used within your "bunch of functions".
The Expression instance is aggregated.
In your second example you're inheriting the trait. This requires that the class that implements your trait must also implement the Expressions trait.
Both example result in the same semantic: A implementation of all members of the Expressions trait is available to your "Bunch of methods".
The only difference is that aggregation is bit more flexible since you can use any object that implements the Expressions.
Your second example can be seen as equivalent to the first one if your trait implementation definesval IR = this.
Here are some of the key differences between the two versions of Scheduling:
Objects that inherit from Scheduling in CODE1...
Must provide a val IR of type Expressions
Have an extra member IR that is visible from outside the object
May use a sub-class of Expressions with different implementations of the methods in Expressions
Can access all the methods of Expressions as IR.<method>
Objects that inherit from Scheduling in CODE2..
Have the type Expressions
Can access all the methods of Expressions directly
I would like to define an implicit conversion from Iterator[T] to a class that I have defined: ProactiveIterator[A].
The question isn't really how to do it but how to do it properly, i.e. where to place the method, so that it is as transparent and unobtrusive as possible. Ideally it should be as the implicit conversion from String to StringOps in scala.Predef If the conversion was from a class in the library to some other class, then it could be defined inside that class, but AFAIK that's not possible here.
So far I have considered to add an object containing these conversions, similarly to JavaConversions, but better options may be possible.
You don't really have much of a choice. All implicits must be contained within some sort of object, and imported with a wildcard import (you could import them individually, but I doubt you want that).
So you'll have some sort of implicits object:
package foo.bar
object Implicits {
implicit class ProactiveIterator[A](i: Iterator[A]) {
...
}
}
Then you must explicitly import it wherever you use it:
import foo.bar.Implicits._
In my opinion, this is a good thing. Someone reading the code might not understand where your pimped methods are coming from, so the explicit import is very helpful.
You can similarly place your implicits within a package object. You would have to import them the same way into other namespaces, but they would be available to classes within the same package.
For example, using the following, anything within foo.bar will have this implicit class available:
package foo
package object bar {
implicit class ProactiveIterator[A](i: Iterator[A]) {
...
}
}
Elsewhere you would import foo.bar._ (which may or may not be as clean, depending on what's in bar).
Ok, I'll explain why I ask this question. I begin to read Lift 2.2 source code these days.
It's good if you happened to read lift source code before.
In Lift, I found that, define inner class and inner trait are very heavily used.
object Menu has 2 inner traits and 4 inner classes. object Loc has 18 inner classes, 5 inner traits, 7 inner objects.
There're tons of codes write like this. I wanna to know why the author write like this.
Is it because it's the author's
personal taste or a powerful use of
language feature?
Is there any trade-off for this kind
of usage?
Before 2.8, you had to choose between packages and objects. The problem with packages is that they cannot contain methods or vals on their own. So you have to put all those inside another object, which can get awkward. Observe:
object Encrypt {
private val magicConstant = 0x12345678
def encryptInt(i: Int) = i ^ magicConstant
class EncryptIterator(ii: Iterator[Int]) extends Iterator[Int] {
def hasNext = ii.hasNext
def next = encryptInt(ii.next)
}
}
Now you can import Encrypt._ and gain access to the method encryptInt as well as the class EncryptIterator. Handy!
In contrast,
package encrypt {
object Encrypt {
private[encrypt] val magicConstant = 0x12345678
def encryptInt(i: Int) = i ^ magicConstant
}
class EncryptIterator(ii: Iterator[Int]) extends Iterator[Int] {
def hasNext = ii.hasNext
def next = Encrypt.encryptInt(ii.next)
}
}
It's not a huge difference, but it makes the user import both encrypt._ and encrypt.Encrypt._ or have to keep writing Encrypt.encryptInt over and over. Why not just use an object instead, as in the first pattern? (There's really no performance penalty, since nested classes aren't actually Java inner classes under the hood; they're just regular classes as far as the JVM knows, but with fancy names that tell you that they're nested.)
In 2.8, you can have your cake and eat it too: call the thing a package object, and the compiler will rewrite the code for you so it actually looks like the second example under the hood (except the object Encrypt is actually called package internally), but behaves like the first example in terms of namespace--the vals and defs are right there without needing an extra import.
Thus, projects that were started pre-2.8 often use objects to enclose lots of stuff as if they were a package. Post-2.8, one of the main motivations has been removed. (But just to be clear, using an object still doesn't hurt; it's more that it's conceptually misleading than that it has a negative impact on performance or whatnot.)
(P.S. Please, please don't try to actually encrypt anything that way except as an example or a joke!)
Putting classes, traits and objects in an object is sometimes required when you want to use abstract type variables, see e.g. http://programming-scala.labs.oreilly.com/ch12.html#_parameterized_types_vs_abstract_types
It can be both. Among other things, an instance of an inner class/trait has access to the variables of its parent. Inner classes have to be created with a parent instance, which is an instance of the outer type.
In other cases, it's probably just a way of grouping closely related things, as in your object example. Note that the trait LocParam is sealed, which means that all subclasses have to be in the same compile unit/file.
sblundy has a decent answer. One thing to add is that only with Scala 2.8 do you have package objects which let you group similar things in a package namespace without making a completely separate object. For that reason I will be updating my Lift Modules proposal to use a package object instead of a simple object.