Файл: Programming_in_Scala,_2nd_edition.pdf

scala> capital("New Zealand") res21: String = Wellington

You can create synchronized sets similarly to the way you create synchronized maps. For example, you could create a synchronized HashSet by mixing in the SynchronizedSet trait, like this:

import scala.collection.mutable

val synchroSet =

new mutable.HashSet[Int] with mutable.SynchronizedSet[Int]

Finally, if you are thinking of using synchronized collections, you may also wish to consider the concurrent collections of java.util.concurrent instead. Alternatively, you may prefer to use unsynchronized collections with Scala actors. Actors will be covered in detail in Chapter 32.

24.9 Concrete immutable collection classes

Scala provides many concrete immutable collection classes for you to choose from. They differ in the traits they implement (maps, sets, sequences), whether they can be infinite, and the speed of various operations. We’ll start by reviewing the most common immutable collection types.

Lists

Lists are finite immutable sequences. They provide constant-time access to their first element as well as the rest of the list, and they have a constant-time cons operation for adding a new element to the front of the list. Many other operations take linear time. See Chapters 16 and 22 for extensive discussions about lists.

Streams

A stream is like a list except that its elements are computed lazily. Because of this, a stream can be infinitely long. Only those elements requested will be computed. Otherwise, streams have the same performance characteristics as lists.

Cover · Overview · Contents · Discuss · Suggest · Glossary · Index

Whereas lists are constructed with the :: operator, streams are constructed with the similar-looking #::. Here is a simple example of a stream containing the integers 1, 2, and 3:

scala> val str = 1 #:: 2 #:: 3 #:: Stream.empty

str: scala.collection.immutable.Stream[Int] = Stream(1, ?)

The head of this stream is 1, and the tail of it has 2 and 3. The tail is not printed here, though, because it hasn’t been computed yet! Streams are required to compute lazily, and the toString method of a stream is careful not to force any extra evaluation.

Below is a more complex example. It computes a stream that contains a Fibonacci sequence starting with the given two numbers. A Fibonacci sequence is one where each element is the sum of the previous two elements in the series:

scala> def fibFrom(a: Int, b: Int): Stream[Int] = a #:: fibFrom(b, a + b)

fibFrom: (a: Int,b: Int)Stream[Int]

This function is deceptively simple. The first element of the sequence is clearly a, and the rest of the sequence is the Fibonacci sequence starting with b followed by a + b. The tricky part is computing this sequence without causing an infinite recursion. If the function used :: instead of #::, then every call to the function would result in another call, thus causing an infinite recursion. Since it uses #::, though, the right-hand side is not evaluated until it is requested.

Here are the first few elements of the Fibonacci sequence starting with two ones:

scala> val fibs = fibFrom(1, 1).take(7)

fibs: scala.collection.immutable.Stream[Int] = Stream(1, ?)

scala> fibs.toList

res22: List[Int] = List(1, 1, 2, 3, 5, 8, 13)

Vectors

Lists are very efficient when the algorithm processing them is careful to only process their heads. Accessing, adding, and removing the head of a list takes

Cover · Overview · Contents · Discuss · Suggest · Glossary · Index

only constant time, whereas accessing or modifying elements later in the list takes time linear in the depth into the list.

Vectors are a new collection type in Scala 2.8 that give efficient access to elements beyond the head. Access to any elements of a vector take only “effectively constant time,” as defined below. It’s a larger constant than for access to the head of a list or for reading an element of an array, but it’s a constant nonetheless. As a result, algorithms using vectors do not have to be careful about accessing just the head of the sequence. They can access and modify elements at arbitrary locations, and thus they can be much more convenient to write.

Vectors are built and modified just like any other sequence:

scala> val vec = scala.collection.immutable.Vector.empty vec: scala.collection.immutable.Vector[Nothing] = Vector()

scala> val vec2 = vec :+ 1 :+ 2

vec2: scala.collection.immutable.Vector[Int] = Vector(1, 2)

scala> val vec3 = 100 +: vec2

vec3: scala.collection.immutable.Vector[Int] = Vector(100, 1, 2)

scala> vec3(0) res23: Int = 100

Vectors are represented as broad, shallow trees. Every tree node contains up to 32 elements of the vector or contains up to 32 other tree nodes. Vectors with up to 32 elements can be represented in a single node. Vectors with up to 32 * 32 = 1024 elements can be represented with a single indirection. Two hops from the root of the tree to the final element node are sufficient for vectors with up to 215 elements, three hops for vectors with 220, four hops for vectors with 225 elements and five hops for vectors with up to 230 elements. So for all vectors of reasonable size, an element selection involves up to five primitive array selections. This is what we meant when we wrote that element access is “effectively constant time.”

Vectors are immutable, so you cannot change an element of a vector in place. However, with the updated method you can create a new vector that differs from a given vector only in a single element:

scala> val vec = Vector(1, 2, 3)

vec: scala.collection.immutable.Vector[Int] = Vector(1, 2, 3)

Cover · Overview · Contents · Discuss · Suggest · Glossary · Index

scala> vec updated (2, 4)

res24: scala.collection.immutable.Vector[Int] = Vector(1, 2, 4)

scala> vec

res25: scala.collection.immutable.Vector[Int] = Vector(1, 2, 3)

As the last line above shows, a call to updated has no effect on the original vector vec. Like selection, functional vector updates are also “effectively constant time.” Updating an element in the middle of a vector can be done by copying the node that contains the element, and every node that points to it, starting from the root of the tree. This means that a functional update creates between one and five nodes that each contain up to 32 elements or subtrees. This is certainly more expensive than an in-place update in a mutable array, but still a lot cheaper than copying the whole vector.

Because vectors strike a good balance between fast random selections and fast random functional updates, they are currently the default implementation of immutable indexed sequences:

scala> collection.immutable.IndexedSeq(1, 2, 3) res26: scala.collection.immutable.IndexedSeq[Int]

= Vector(1, 2, 3)

Immutable stacks

If you need a last-in-first-out sequence, you can use a Stack. You push an element onto a stack with push, pop an element with pop, and peek at the top of the stack without removing it with top. All of these operations are constant time.

Here are some simple operations performed on a stack:

scala> val stack = scala.collection.immutable.Stack.empty stack: scala.collection.immutable.Stack[Nothing] = Stack()

scala> val hasOne = stack.push(1)

hasOne: scala.collection.immutable.Stack[Int] = Stack(1)

scala> stack

res27: scala.collection.immutable.Stack[Nothing] = Stack()

scala> hasOne.top res28: Int = 1

Cover · Overview · Contents · Discuss · Suggest · Glossary · Index

scala> hasOne.pop

res29: scala.collection.immutable.Stack[Int] = Stack()

Immutable stacks are used rarely in Scala programs because their functionality is subsumed by lists: A push on an immutable stack is the same as a :: on a list, and a pop on a stack is the same a tail on a list.

Immutable queues

A queue is just like a stack except that it is first-in-first-out rather than last-in- first-out. A simplified implementation of immutable queues was discussed in Chapter 19. Here’s how you can create an empty immutable queue:

scala> val empty = scala.collection.immutable.Queue[Int]() empty: scala.collection.immutable.Queue[Int] = Queue()

You can append an element to an immutable queue with enqueue:

scala> val has1 = empty.enqueue(1)

has1: scala.collection.immutable.Queue[Int] = Queue(1)

To append multiple elements to a queue, call enqueue with a collection as its argument:

scala> val has123 = has1.enqueue(List(2, 3)) has123: scala.collection.immutable.Queue[Int]

= Queue(1, 2, 3)

To remove an element from the head of the queue, use dequeue:

scala> val (element, has23) = has123.dequeue element: Int = 1

has23: scala.collection.immutable.Queue[Int] = Queue(2, 3)

Note that dequeue returns a pair consisting of the element removed and the rest of the queue.

Ranges

A range is an ordered sequence of integers that are equally spaced apart. For example, “1, 2, 3” is a range, as is “5, 8, 11, 14.” To create a range in Scala, use the predefined methods to and by. Here are some examples:

Cover · Overview · Contents · Discuss · Suggest · Glossary · Index

scala> 1 to 3

res30: scala.collection.immutable.Range.Inclusive with scala.collection.immutable.Range.ByOne

= Range(1, 2, 3)

scala> 5 to 14 by 3

res31: scala.collection.immutable.Range = Range(5, 8, 11, 14)

If you want to create a range that is exclusive of its upper limit, use the convenience method until instead of to:

scala> 1 until 3

res32: scala.collection.immutable.Range

with scala.collection.immutable.Range.ByOne = Range(1, 2)

Ranges are represented in constant space, because they can be defined by just three numbers: their start, their end, and the stepping value. Because of this representation, most operations on ranges are extremely fast.

Hash tries

Hash tries4 are a standard way to implement immutable sets and maps efficiently. Their representation is similar to vectors in that they are also trees where every node has 32 elements or 32 subtrees, but selection is done based on a hash code. For instance, to find a given key in a map, you use the lowest five bits of the hash code of the key to select the first subtree, the next five bits the next subtree, and so on. Selection stops once all elements stored in a node have hash codes that differ from each other in the bits that are selected so far. Thus, not all the bits of the hash code are necessarily used.

Hash tries strike a nice balance between reasonably fast lookups and reasonably efficient functional insertions (+) and deletions (-). That’s why they underlie Scala’s default implementations of immutable maps and sets. In fact, Scala has a further optimization for immutable sets and maps that contain less than five elements. Sets and maps with one to four elements are stored as single objects that just contain the elements (or key/value pairs in the case of a map) as fields. The empty immutable set and empty immutable map is in each case a singleton object—there’s no need to duplicate storage for those because an empty immutable set or map will always stay empty.

4“Trie” comes from the word "retrieval" and is pronounced tree or try.

Cover · Overview · Contents · Discuss · Suggest · Glossary · Index

Red-black trees

Red-black trees are a form of balanced binary trees where some nodes are designated “red” and others “black.” Like any balanced binary tree, operations on them reliably complete in time logarithmic to the size of the tree.

Scala provides implementations of sets and maps that use a red-black tree internally. You access them under the names TreeSet and TreeMap:

scala> val set = collection.immutable.TreeSet.empty[Int] set: scala.collection.immutable.TreeSet[Int] = TreeSet()

scala> set + 1 + 3 + 3

res33: scala.collection.immutable.TreeSet[Int] = TreeSet(1, 3)

Red-black trees are also the standard implementation of SortedSet in Scala, because they provide an efficient iterator that returns all elements of the set in sorted order.

Immutable bit sets

A bit set represents a collection of small integers as the bits of a larger integer. For example, the bit set containing 3, 2, and 0 would be represented as the integer 1101 in binary, which is 13 in decimal.

Internally, bit sets use an array of 64-bit Longs. The first Long in the array is for integers 0 through 63, the second is for 64 through 127, and so on. Thus, bit sets are very compact so long as the largest integer in the set is less than a few hundred or so.

Operations on bit sets are very fast. Testing for inclusion takes constant time. Adding an item to the set takes time proportional to the number of Longs in the bit set’s array, which is typically a small number. Here are some simple examples of the use of a bit set:

scala> val bits = scala.collection.immutable.BitSet.empty bits: scala.collection.immutable.BitSet = BitSet()

scala> val moreBits = bits + 3 + 4 + 4

moreBits: scala.collection.immutable.BitSet = BitSet(3, 4)

scala> moreBits(3) res34: Boolean = true

Cover · Overview · Contents · Discuss · Suggest · Glossary · Index

scala> moreBits(0) res35: Boolean = false

List maps

A list map represents a map as a linked list of key-value pairs. In general, operations on a list map might have to iterate through the entire list. Thus, operations on a list map take time linear in the size of the map. In fact there is little usage for list maps in Scala because standard immutable maps are almost always faster. The only possible difference is if the map is for some reason constructed in such a way that the first elements in the list are selected much more often than the other elements.

scala> val map = collection.immutable.ListMap( 1 -> "one", 2 -> "two")

map: scala.collection.immutable.ListMap[Int,java.lang.String] = Map((1,one), (2,two))

scala> map(2)

res36: java.lang.String = two

24.10Concrete mutable collection classes

Now that you’ve seen the most commonly used immutable collection classes that Scala provides in its standard library, take a look at the mutable collection classes.

Array buffers

You’ve already seen array buffers in Section 17.1. An array buffer holds an array and a size. Most operations on an array buffer have the same speed as an array, because the operations simply access and modify the underlying array. Additionally, array buffers can have data efficiently added to the end. Appending an item to an array buffer takes amortized constant time. Thus, array buffers are useful for efficiently building up a large collection whenever the new items are always added to the end. Here are some examples:

Cover · Overview · Contents · Discuss · Suggest · Glossary · Index

scala> val buf = collection.mutable.ArrayBuffer.empty[Int] buf: scala.collection.mutable.ArrayBuffer[Int]

= ArrayBuffer()

scala> buf += 1

res37: buf.type = ArrayBuffer(1)

scala> buf += 10

res38: buf.type = ArrayBuffer(1, 10)

scala> buf.toArray

res39: Array[Int] = Array(1, 10)

List buffers

You’ve also already seen list buffers in Section 17.1. A list buffer is like an array buffer except that it uses a linked list internally instead of an array. If you plan to convert the buffer to a list once it is built up, use a list buffer instead of an array buffer. Here’s an example:5

scala> val buf = collection.mutable.ListBuffer.empty[Int] buf: scala.collection.mutable.ListBuffer[Int]

= ListBuffer()

scala> buf += 1

res40: buf.type = ListBuffer(1)

scala> buf += 10

res41: buf.type = ListBuffer(1, 10)

scala> buf.toList

res42: List[Int] = List(1, 10)

String builders

Just like an array buffer is useful for building arrays, and a list buffer is useful for building lists, a string builder is useful for building strings. String builders are so commonly used that they are already imported into the default namespace. Create them with a simple new StringBuilder, like this:

5The “buf.type” that appears in the interpreter responses in this and several other examples in this section is a singleton type. As will be explained in Section 29.6, buf.type means the variable holds exactly the object referred to by buf.

Cover · Overview · Contents · Discuss · Suggest · Glossary · Index