Collections Framework : Cover the history of Collections frameworks back to the late 1970s or early 1980s, and how they were retrofitted to various languages. Discuss the advantage of a standard Collections framework (e.g., Java) vs. several competing frameworks (e.g., C++), and the advantage of collections built into the language (e.g., Ruby) vs. collections as class libraries. Give examples to illustrate the advantages you identify.

Introduction

In this wiki, we are going to talk about collections frameworks : their history, the advantages of standard collections frameworks (like Java), compared to that of C++, the advantages of collections built into languages like Ruby compared to those having collections as class libraries. We will illustrate the advantages and compare the differences with some examples.

What is a collections framework?

Firstly, what is a collection? A collection is a group of objects, and is essentially a container that groups multiple elements into a single unit. Since a collection is just a group of objects, some examples of a collection are : a stack of books, a dictionary, a list of names, a telephone directory. A collections framework is a system to represent and manipulate these collections.

Need for a collections framework

There are many needs for a collections framework. Some of them are :

It solves the problem of storing and retrieving data
It makes sure we write lesser code, and the code is more readable
It improves the performance, and makes sure the programmer doesn't have to bother with issues related to performance
Makes sure we have reusable data structures

These reasons led to the development of many collections frameworks.

History

In the 70's and 80's, a number of languages had collections built into them, unlike how it is now in say C++ where libraries have to be included for performing tasks on certain collections. Some of these languages are Smalltalk, Perl, Scheme, Miranda, and Haskell. Just to give a brief idea, below we have given a short note on Smalltalk and Perl.

Smalltalk

Smalltalk was object-oriented and was one of them that had collections. Collections in Smalltalk could be as indexed or non-indexed, indexed by explicit or implicit keys, ordered using an index or unordered, and fixed size or expandable. Smalltalk has Arrays, and the Array class is a subclass of ArrayedCollection. An array can hold any type of object, and each entry in the array can be a different object. The code snippet below creates an array object in Smalltalk :

a _ Array new: 5

And to place an object at a specific location in the array :

a at: 3 put: object

a at: 2 returns the object at position 2. Very similar to arrays, OrderedCollection is also used. The difference between them both is that all objects of Array type are fixed size, while those of OrderedCollection can be of varying size. Collections that hold key value pairs were done using Dictionaries, as shown by the code snippet below :

| dict | 
dict := Dictionary new. 
dict at: 'Apple' put: 'Fruit'. 
dict at: 'Onion' put: 'Vegetable'. 
dict at: 'Pork' put: 'Meat'.

Then, in order to get the value for key 'Onion' we simply say :

dict at 'Onion'

which gives us 'Vegetable'.

Perl

Perl is a very powerful scripting language, and has collections as arrays and associative arrays. Arrays in Perl are defined with an @ sign.

@mylist = ("Tom", 2);

$mylist[0] has the value "Tom", and $mylist[1] has the value 2. In Perl, associative arrays can be used for hashing, and are defined with a % sign.


%months = ( "January" , 1,
           "April" ,    4,
           "November",  11 );

In the example above, $months("April") would return the value 4.

Collections Frameworks today

Alexander Stepanov was the biggest contributor to the designing of Standard Template Library. In late 1970's he started working on Generic Programming. At that time there was not much support for Generic Programming in any Computer programming languages. A language called Ada was able to provide some amount of support. By late 1980's Alexander Stepanov and David Musser had worked on the Ada library and later developed and also published it. By that time, programmers were using C++ and it was more widely used and thought to provide a good support for generic programming even though it was still considered a new language. Stephanov realized that the model of computation in C and C++ had a lot of flexibility and turned to C++. After a lot of research and experimentation at AT&T and HP laboratories, Stepanov presented his ideas at ANSI/ISO meeting. With some more experimentation and collaboration, in 1994, the STL (Standard Template Library) became a part of the ANSI/ISO standard definition of the C++ language.

As mentioned before, to address the need for reusable collection data structures, several independent frameworks were developed. Doug Lea's collections package (released in 1995) was the first widely used collection. In 1996, a company called Object Space developed and released the Java Generic Library ,JGL, with the main goal of being consistent with C++ Standard Template Library.

The Java collections framework was first introduced in 1998, with Java 2 platform, Standard edition. version 1.2. Before collections framework was introduced, grouping of objects in Java was through the use of arrays, Vector and HashTable.

There are many other collections frameworks : the .NET framework has collection classes, for example. In this article, however, we will be talking about the Standard Library Template of C++, Java Collections Framework and Ruby to compare and analyze the advantages and disadvantages of each.

C++'s Standard Template Library vs Java Collections Framework

Before we talk about the difference between the Standard Template Library and Java Collections Framework, given below is the overview of each of them (in terms of architecture, types of collections, and examples).

C++ Standard Template Library (STL)

The C++ Standard Template Library (STL) is a powerful and versatile collection of classes and functions that provides an efficient, lightweight, and extensible framework for application development. STL also offers a sophisticated level of abstraction that promotes the use of generic data structures and algorithms.The Standard Template Library, is a C++ library of container classes, algorithms, and iterators. The STL provides many of the basic algorithms and data structures of computer science. It can be called a Generic library because most of the components in the STL are templates.

Architecture of STL

The diagram image (Source : [15]) below gives an idea of the architecture of the Standard Template Library in STL :

Architecture of STL - Image Source : http://www.cs.brown.edu/~jak/proglang/cpp/stltut/tut.html

Types of Collections

The STL (Standard Template Library) is organized around three fundamental components namely containers - to hold objects, algorithms ( perform manipulation of elements in containers) and iterators. In addition, supporting them there are three additional components: allocators, adaptors and function objects.

Containers manage collections of objects.

There are a variety of containers, and each offers unique advantages. Standard containers include vector, list, deque, set, map, and so on.

Algorithms process elements in containers, or more specifically, in ranges. The STL offers a wide variety of algorithms to search, sort, and operate on elements in containers.

Iterators are generalized pointer-like objects that are used to step through the elements of a container.

Iterators can work on an entire collection of objects or on a subset of objects.

STL distinguishes general algorithms from the more specialized data and methods encapsulated by ordinary abstract data types. Through this, complex and powerful algorithms can be implemented independently of the data to which they are applied, allowing for generalization and reuse of these algorithms.

Examples

Vectors: In C++, a typical vector implementation consists of a pointer referring to a dynamically allocated array, and data members holding the capacity and size of the vector. The size of the vector refers to the number of elements, where as the capacity refers to the size of the internal array. As new elements are inserted, and if the new size of the vector becomes larger than its capacity, reallocation occurs. This causes the vector to allocate a new location for storage followed by moving the previously held elements to a new location of storage, and freeing the old space. The example below illustrates this.


#include <vector>

int main()
{
  std::vector<int> v(1); // Creating a Vector.
 
  int& first = *v.begin(); // Making a reference to the first element.
 
  v.insert(v.end(), v.capacity(), 0); // Adding more elements to the Vector thereby forcing re-allocation.
  
}

HashMap: The hash_map class uses an associative array in which the key is stored according to some hashing scheme.

#include <hash_map>
    
     int main()
    {
         hash_map<const char*, int, hash<const char*>, eqstr> fruits;
  
  fruits["Apple"] =  1;
  fruits["Orange"] = 2;
  fruits["Grapes"] = 3;
  
  cout << "Apple   -> " << months["Apple"] << endl;
  
}

Java Collection Framework (JCF)

Architecture of JCF

The diagram image (Source : [2]) below shows the interface hierarchy in the collections framework in Java.

The interfaces map and collection are both distinct in the hierarchy in Java. This distinction is because of the way maps and collection are used in Java libraries.

The collections framework in Java consists of:

Collection Interfaces : These are needed for representing different types of collections, like sets, lists and maps.

General-purpose Implementations : They are are the most commonly used implementations, designed for day to day use.

Legacy Implementations: The collection classes that were a part of Java 1.1, namely Vector and Hashtable, are retrofitted to implement the collection interfaces that are mentioned above.

Special-purpose Implementations : These are implementations designed for special situations usage.

Concurrent Implementations : These are implementations designed for highly concurrent use.

Wrapper Implementations : These implementations add functionality, like synchronization, to other implementations.

Convenience Implementations : These are High-performance implementations which are in a way smaller implementations of the collection interfaces.

Abstract Implementations : These are partial implementations of the collection interfaces required to facilitate custom implementations.

Algorithms : These are static methods that perform useful functions on collections.

Infrastructure : These are interfaces that provide essential support for all of the collection interfaces.

Array Utilities : These are utility functions for arrays of primitives and reference objects.

Types of Collections

Most collections in Java are derived from the java.util.Collection interface.

The main types of collections in Java are :

Lists : There are two classes, java.util.LinkedList and java.util.ArrayList that implement the java.util.List interface in the JCF framework. Lists in Java are ordered, and may contain duplicates.
Sets : This is defined by the java.util.Set interface. It is implemented by java.util.HashSet, etc. A set cannot have duplicates by definition, and it doesn't have an order.
Maps : The java.util.Map interface is used to define maps. Maps are data structures that associate or map a key to a value. Hashmaps are used as they make the time of searching for a key in O(1), a constant time complexity order.

Examples

As an example, let us consider the List interface. It is part of the java.util package. List interface can add value elements by add(value) method.

import java.util.Iterator;
import java.util.List;
import java.util.ArrayList;

public class MyList {

    public static void main(String[] args) {

   
        List<String> ml=new ArrayList<String>();

        ml.add("Milk");
        ml.add("Eggs");
        ml.add("Cereal");
      
        Iterator it=ml.iterator();

        while(it.hasNext())
        {
          String output=(String)it.next();

          System.out.println("My Grocery List :"+output);
        }
    }
}

As expected, the output of the above code snippet would simply print out the List of Groceries :

My Grocery List : Milk
My Grocery List : Eggs
My Grocery List : Cereal

Comparison between STL and JCF

Prior to Java 2, arrays, vectors and HashTable had different syntaxes for accessing members. For example, the arrays uses square bracket symbol [ ], the HashTable uses 'get and put' method and as for Vectors, it is the elementAt method. This caused a lot of inconsistency meaning programmers were using different methods to implement their own collections. Also, prior to Java 2, the classes lacked a central and unifying theme. There are many advantages of Java Collections framework compared to frameworks such as those in C++( standard template library).

The primary difference between STL and JCF is that JCF has a singular focus - It focuses mainly on containers rather than on combination of containers and algorithms ( as STL does).

The Quality of Implementation of the C++ compiler has a large impact on usability of STL. First, the error messages involving templates are very long and hard to interpret. Then, if used without proper care, STL templates can lead to code bloat. The Java Collections Framework make programming easy by providing many useful data structures and algorithms so that the programmer does not have to write them. The performance of a program is increased considerably because of the implementation of these data structures and algorithms because the various implementations of each interface can be interchanged. Collections have much better performance compared to Vectors and HashTable in the old version. The entire standard collections framework is designed around a set of standard interfaces and several standard implementations such a LinkedList, HashSet and TreeSet are provided in these interfaces so the programmers can use them or also create their own. The table below highlights the differences between the C++ Standard Template Library and the Java Collection Framework.

Difference	Standard Template Library	Java Collections Framework
Simplicity of Programming	Harder	Relatively Easy
Architecture of Collections	All are inherited from the base class	Maps and Collections are two separate interfaces
Error Messages	Long, Hard to interpret	Easier to interpret
Memory : Resource Management	Has to be explicitly done by programmer	Taken care of by Garbage Collector
Level of Abstraction	Less	Higher level of abstraction
Types of Collections	Vectors List Deque Map Set Multimap Multiset	Lists Sets Maps Supports 14 collection interfaces

Clearly from the table above, the advantages of Java Collections Framework compared to the Standard Template Library is seen in terms of various parameters.

Built-in Collections vs Class Library Collections

In this section, we will describe collections in a built in language (like Ruby), and compare it to collections that use class libraries (like STL and JCF).

Collections in Ruby : An overview

In Ruby, we have containers which are entities that contain other entities. The two native containers in Ruby are :

Arrays, which can hold a collection of object references
Hashes, which are an indexed collection of object references

We can also create our own list structure in Ruby. The example below illustrates the syntax of how Arrays can hold a collection of object references. Initialization is not a issue in Ruby as its dynamically typed.

arr = ["Onion", 1.414, "Tomato"]
a = ["Rob", arr, 9]

As shown above, arr is an array of different objects, and so is a. The section below gives a clearer view of this, with the outputs after each step.

arr = ["Onion", 1.414, "Tomato"]
=>["Onion", 1.414, "Tomato"]
arr.type 
=> Array
arr.length
=> 3
arr[1]
=> 1.414
a = ["Rob", arr, 9]
=> ["Rob", ["Onion", 1.414, "Tomato"], 9]

Hashing can also be done very easily in Ruby, without the need for any reference libraries.

food = {'Apple' => 'Fruit', 'Onion' => 'Vegetable', 'Pork' => 'Meat'}

So food['Onion'] gives us the value corresponding to key 'Onion', which is 'Vegetable'. In Java when one has a nested methods on collections, the code is easier to read than in Ruby when one has to nested operators on collections within a single line. The nested operator code becomes very difficult to read.

Comparison of Ruby Collections and Collections as Class Libraries

When comparing the runtime of Java and Ruby, Java is way faster than the latter. But since Ruby, a dynamically typed language, has built-in lists/ arrays , hashes, there is no need for the programmer to have to build all the infrastructure. This is a win situation for Ruby when compared to Java's collections. Thus, the built-in arrays and hashes are a huge advantage over JCF. Ruby has a mark-and-sweep garbage collection which allows programmers the ability to code without having to worry about the need to maintain reference counts in extension libraries. Also, if an operating system allows for it, Ruby can dynamically load extension libraries. (Reference : Ruby-lang.org)

The table below throws light on the difference between Ruby and those that use collections as class libraries like JCF and STL :

Difference	Ruby	Collections as Class Libraries
Language Design	Uses Interpreter	Uses Compiler
Typing	Dynamically Typed	Statically Typed
Lines of Code	Lesser - Collections included in Language	More - Have to include JCF
Programmer's Psychology	Programmer feels its a part of language, and starts using it	Programmer feels something external is being used; uses it only when its very much needed

The table above shows a comparison of languages having collections built-in to them (like Ruby), compared to those that use collections as class libraries. The table does show that though each one does have its own advantages, there are a lot of advantages of having it built into the language compared to using libraries.