src/java.base/share/classes/java/util/doc-files/coll-designfaq.html

<!DOCTYPE html>
<!--
 Copyright (c) 1998, 2019, Oracle and/or its affiliates. All rights reserved.
 DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.

 This code is free software; you can redistribute it and/or modify it
 under the terms of the GNU General Public License version 2 only, as
 published by the Free Software Foundation.  Oracle designates this
 particular file as subject to the "Classpath" exception as provided
 by Oracle in the LICENSE file that accompanied this code.

 This code is distributed in the hope that it will be useful, but WITHOUT
 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 version 2 for more details (a copy is included in the LICENSE file that
 accompanied this code).

 You should have received a copy of the GNU General Public License version
 2 along with this work; if not, write to the Free Software Foundation,
 Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.

 Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 or visit www.oracle.com if you need additional information or have any
 questions.
-->
<html lang="en-US">
<head>
<title>Java Collections API Design FAQ</title>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body>
<h1>Java Collections API Design FAQ</h1>
<!-- Body text begins here -->
<hr>
This document answers frequently asked questions concerning the
design of the Java collections framework. It is derived from the
large volume of traffic on the collections-comments alias. It
serves as a design rationale for the collections framework.
<h2>Core Interfaces - General Questions</h2>
<ol>
<li><a href="#a1"><b>Why don't you support immutability directly in
the core collection interfaces so that you can do away with
<em>optional operations</em> (and
UnsupportedOperationException)?</b></a></li>
<li><a href="#a2"><b>Won't programmers have to surround any code
that calls optional operations with a try-catch clause in case they
throw an UnsupportedOperationException?</b></a></li>
<li><a href="#a3"><b>Why isn't there a core interface for "bags"
(AKA multisets)?</b></a></li>
<li><a href="#a28"><b>Why didn't you use "Beans-style names" for
consistency?</b></a></li>
</ol>
<h2>Collection Interface</h2>
<ol>
<li><a href="#a5"><b>Why doesn't Collection extend Cloneable and
Serializable?</b></a></li>
<li><a href="#a6"><b>Why don't you provide an "apply" method in
Collection to apply a given method ("upcall") to all the elements
of the Collection?</b></a></li>
<li><a href="#a7"><b>Why didn't you provide a "Predicate" interface,
and related methods (e.g., a method to find the first element in
the Collection satisfying the predicate)?</b></a></li>
<li><a href="#a8"><b>Why don't you provide a form of the addAll
method that takes an Enumeration (or an Iterator)?</b></a></li>
<li><a href="#a9"><b>Why don't the concrete implementations in the
JDK have Enumeration (or Iterator) constructors?</b></a></li>
<li><a href="#a10"><b>Why don't you provide an Iterator.add
method?</b></a></li>
</ol>
<h2>List Interface</h2>
<ol>
<li><a href="#a11"><b>Why don't you rename the List interface to
Sequence; doesn't "list" generally suggest "linked list"? Also,
doesn't it conflict with java.awt.List?</b></a></li>
<li><a href="#a12"><b>Why don't you rename List's set method to
replace, to avoid confusion with Set.</b></a></li>
</ol>
<h2>Map Interface</h2>
<ol>
<li><a href="#a14"><b>Why doesn't Map extend
Collection?</b></a></li>
</ol>
<h2>Iterator Interface</h2>
<ol>
<li><a href="#a18"><b>Why doesn't Iterator extend
Enumeration?</b></a></li>
<li><a href="#a19"><b>Why don't you provide an Iterator.peek method
that allows you to look at the next element in an iteration without
advancing the iterator?</b></a></li>
</ol>
<h2>Miscellaneous</h2>
<ol>
<li><a href="#a23"><b>Why did you write a new collections framework
instead of adopting JGL (a preexisting collections package from
ObjectSpace, Inc.) into the JDK?</b></a></li>
<li><a href="#a26"><b>Why don't you eliminate all of the methods and
classes that return "views" (Collections backed by other
collection-like objects). This would greatly reduce
aliasing.</b></a></li>
<li><a href="#a27"><b>Why don't you provide for "observable"
collections that send out Events when they're
modified?</b></a></li>
</ol>
<hr>
<h2>Core Interfaces - General Questions</h2>
<ol>
<li><a id="a1"><b>Why don't you support immutability
directly in the core collection interfaces so that you can do away
with <em>optional operations</em> (and
UnsupportedOperationException)?</b></a>
<p>This is the most controversial design decision in the whole API.
Clearly, static (compile time) type checking is highly desirable,
and is the norm in Java. We would have supported it if we believed
it were feasible. Unfortunately, attempts to achieve this goal
cause an explosion in the size of the interface hierarchy, and do
not succeed in eliminating the need for runtime exceptions (though
they reduce it substantially).</p>
<p>Doug Lea, who wrote a popular Java collections package that did
reflect mutability distinctions in its interface hierarchy, no
longer believes it is a viable approach, based on user experience
with his collections package. In his words (from personal
correspondence) "Much as it pains me to say it, strong static
typing does not work for collection interfaces in Java."</p>
<p>To illustrate the problem in gory detail, suppose you want to
add the notion of modifiability to the Hierarchy. You need four new
interfaces: ModifiableCollection, ModifiableSet, ModifiableList,
and ModifiableMap. What was previously a simple hierarchy is now a
messy heterarchy. Also, you need a new Iterator interface for use
with unmodifiable Collections, that does not contain the remove
operation. Now can you do away with UnsupportedOperationException?
Unfortunately not.</p>
<p>Consider arrays. They implement most of the List operations, but
not remove and add. They are "fixed-size" Lists. If you want to
capture this notion in the hierarchy, you have to add two new
interfaces: VariableSizeList and VariableSizeMap. You don't have to
add VariableSizeCollection and VariableSizeSet, because they'd be
identical to ModifiableCollection and ModifiableSet, but you might
choose to add them anyway for consistency's sake. Also, you need a
new variety of ListIterator that doesn't support the add and remove
operations, to go along with unmodifiable List. Now we're up to ten
or twelve interfaces, plus two new Iterator interfaces, instead of
our original four. Are we done? No.</p>
<p>Consider logs (such as error logs, audit logs and journals for
recoverable data objects). They are natural append-only sequences,
that support all of the List operations except for remove and set
(replace). They require a new core interface, and a new
iterator.</p>
<p>And what about immutable Collections, as opposed to unmodifiable
ones? (i.e., Collections that cannot be changed by the client AND
will never change for any other reason). Many argue that this is
the most important distinction of all, because it allows multiple
threads to access a collection concurrently without the need for
synchronization. Adding this support to the type hierarchy requires
four more interfaces.</p>
<p>Now we're up to twenty or so interfaces and five iterators, and
it's almost certain that there are still collections arising in
practice that don't fit cleanly into any of the interfaces. For
example, the <em>collection-views</em> returned by Map are natural
delete-only collections. Also, there are collections that will
reject certain elements on the basis of their value, so we still
haven't done away with runtime exceptions.</p>
<p>When all was said and done, we felt that it was a sound
engineering compromise to sidestep the whole issue by providing a
very small set of core interfaces that can throw a runtime
exception.</p>
</li>
<li><a id="a2"><b>Won't programmers have to surround any
code that calls optional operations with a try-catch clause in case
they throw an UnsupportedOperationException?</b></a>
<p>It was never our intention that programs should catch these
exceptions: that's why they're unchecked (runtime) exceptions. They
should only arise as a result of programming errors, in which case,
your program will halt due to the uncaught exception.</p>
</li>
<li><a id="a3"><b>Why isn't there a core interface for
"bags" (AKA multisets)?</b></a>
<p>The Collection interface provides this functionality. We are not
providing any public implementations of this interface, as we think
that it wouldn't be used frequently enough to "pull its weight." We
occasionally return such Collections, which are implemented easily
atop AbstractCollection (for example, the Collection returned by
Map.values).</p>
</li>
<li><a id="a28"><b>Why didn't you use "Beans-style
names" for consistency?</b></a>
<p>While the names of the new collections methods do not adhere to
the "Beans naming conventions", we believe that they are
reasonable, consistent and appropriate to their purpose. It should
be remembered that the Beans naming conventions do not apply to the
JDK as a whole; the AWT did adopt these conventions, but that
decision was somewhat controversial. We suspect that the
collections APIs will be used quite pervasively, often with
multiple method calls on a single line of code, so it is important
that the names be short. Consider, for example, the Iterator
methods. Currently, a loop over a collection looks like this:</p>
<pre>
    for (Iterator i = c.iterator(); i.hasNext(); )
        System.out.println(i.next());
</pre>
Everything fits neatly on one line, even if the Collection name is
a long expression. If we named the methods "getIterator",
"hasNextElement" and "getNextElement", this would no longer be the
case. Thus, we adopted the "traditional" JDK style rather than the
Beans style.</li>
</ol>
<hr>
<h2>Collection Interface</h2>
<ol>
<li><a id="a5"><b>Why doesn't Collection extend Cloneable
and Serializable?</b></a>
<p>Many Collection implementations (including all of the ones
provided by the JDK) will have a public clone method, but it would
be mistake to require it of all Collections. For example, what does
it mean to clone a Collection that's backed by a terabyte SQL
database? Should the method call cause the company to requisition a
new disk farm? Similar arguments hold for serializable.</p>
<p>If the client doesn't know the actual type of a Collection, it's
much more flexible and less error prone to have the client decide
what type of Collection is desired, create an empty Collection of
this type, and use the addAll method to copy the elements of the
original collection into the new one.</p>
</li>
<li><a id="a6"><b>Why don't you provide an "apply" method
in Collection to apply a given method ("upcall") to all the
elements of the Collection?</b></a>
<p>This is what is referred to as an "Internal Iterator" in the
"Design Patterns" book (Gamma et al.). We considered providing it,
but decided not to as it seems somewhat redundant to support
internal and external iterators, and Java already has a precedent
for external iterators (with Enumerations). The "throw weight" of
this functionality is increased by the fact that it requires a
public interface to describe upcalls.</p>
</li>
<li><a id="a7"><b>Why didn't you provide a "Predicate"
interface, and related methods (e.g., a method to find the first
element in the Collection satisfying the predicate)?</b></a>
<p>It's easy to implement this functionality atop Iterators, and
the resulting code may actually look cleaner as the user can inline
the predicate. Thus, it's not clear whether this facility pulls its
weight. It could be added to the Collections class at a later date
(implemented atop Iterator), if it's deemed useful.</p>
</li>
<li><a id="a8"><b>Why don't you provide a form of the
addAll method that takes an Enumeration (or an Iterator)?</b></a>
<p>Because we don't believe in using Enumerations (or Iterators) as
"poor man's collections." This was occasionally done in prior
releases, but now that we have the Collection interface, it is the
preferred way to pass around abstract collections of objects.</p>
</li>
<li><a id="a9"><b>Why don't the concrete implementations
in the JDK have Enumeration (or Iterator) constructors?</b></a>
<p>Again, this is an instance of an Enumeration serving as a "poor
man's collection" and we're trying to discourage that. Note
however, that we strongly suggest that all concrete implementations
should have constructors that take a Collection (and create a new
Collection with the same elements).</p>
</li>
<li><a id="a10"><b>Why don't you provide an Iterator.add
method?</b></a>
<p>The semantics are unclear, given that the contract for Iterator
makes no guarantees about the order of iteration. Note, however,
that ListIterator does provide an add operation, as it does
guarantee the order of the iteration.</p>
</li>
</ol>
<hr>
<h2>List Interface</h2>
<ol>
<li><a id="a11"><b>Why don't you rename the List
interface to Sequence; doesn't "list" generally suggest "linked
list"? Also, doesn't it conflict with java.awt.List?</b></a>
<p>People were evenly divided as to whether List suggests linked
lists. Given the implementation naming convention,
&lt;<em>Implementation</em>&gt;&lt;<em>Interface</em>&gt;, there
was a strong desire to keep the core interface names short. Also,
several existing names (AbstractSequentialList, LinkedList) would
have been decidedly worse if we changed List to Sequence. The
naming conflict can be dealt with by the following incantation:</p>
<pre>
    import java.util.*;
    import java.awt.*;
    import java.util.List;   // Dictates interpretation of "List"
</pre></li>
<li><a id="a12"><b>Why don't you rename List's set
method to replace, to avoid confusion with Set.</b></a>
<p>It was decided that the "set/get" naming convention was strongly
enough enshrined in the language that we'd stick with it.</p>
</li>
</ol>
<hr>
<h2>Map Interface</h2>
<ol>
<li><a id="a14"><b>Why doesn't Map extend
Collection?</b></a>
<p>This was by design. We feel that mappings are not collections
and collections are not mappings. Thus, it makes little sense for
Map to extend the Collection interface (or vice versa).</p>
<p>If a Map is a Collection, what are the elements? The only
reasonable answer is "Key-value pairs", but this provides a very
limited (and not particularly useful) Map abstraction. You can't
ask what value a given key maps to, nor can you delete the entry
for a given key without knowing what value it maps to.</p>
<p>Collection could be made to extend Map, but this raises the
question: what are the keys? There's no really satisfactory answer,
and forcing one leads to an unnatural interface.</p>
<p>Maps can be <em>viewed</em> as Collections (of keys, values, or
pairs), and this fact is reflected in the three "Collection view
operations" on Maps (keySet, entrySet, and values). While it is, in
principle, possible to view a List as a Map mapping indices to
elements, this has the nasty property that deleting an element from
the List changes the Key associated with every element before the
deleted element. That's why we don't have a map view operation on
Lists.</p>
</li>
</ol>
<hr>
<h2>Iterator Interface</h2>
<ol>
<li><a id="a18"><b>Why doesn't Iterator extend
Enumeration?</b></a>
<p>We view the method names for Enumeration as unfortunate. They're
very long, and very frequently used. Given that we were adding a
method and creating a whole new framework, we felt that it would be
foolish not to take advantage of the opportunity to improve the
names. Of course we could support the new and old names in
Iterator, but it doesn't seem worthwhile.</p>
</li>
<li><a id="a19"><b>Why don't you provide an
Iterator.peek method that allows you to look at the next element in
an iteration without advancing the iterator?</b></a>
<p>It can be implemented atop the current Iterators (a similar
pattern to java.io.PushbackInputStream). We believe that its use
would be rare enough that it isn't worth including in the interface
that everyone has to implement.</p>
</li>
</ol>
<hr>
<h2>Miscellaneous</h2>
<ol>
<li><a id="a23"><b>Why did you write a new collections
framework instead of adopting JGL (a preexisting collections
package from ObjectSpace, Inc.) into the JDK?</b></a>
<p>If you examine the goals for our Collections framework (in the
Overview), you'll see that we are not really "playing in the same
space" as JGL. Quoting from the "Design Goals" Section of the Java
Collections Overview: "Our main design goal was to produce an API
that was reasonably small, both in size, and (more importantly) in
'conceptual weight.'"</p>
<p>JGL consists of approximately 130 classes and interfaces; its
main goal was consistency with the C++ Standard Template Library
(STL). This was <em>not</em> one of our goals. Java has
traditionally stayed away from C++'s more complex features (e.g.,
multiple inheritance, operator overloading). Our entire framework,
including all infrastructure, contains approximately 25 classes and
interfaces.</p>
<p>While this may cause some discomfort for some C++ programmers,
we feel that it will be good for Java in the long run. As the Java
libraries mature, they inevitably grow, but we are trying as hard
as we can to keep them small and manageable, so that Java continues
to be an easy, fun language to learn and to use.</p>
</li>
<li><a id="a26"><b>Why don't you eliminate all of the
methods and classes that return "views" (Collections backed by
other collection-like objects). This would greatly reduce
aliasing.</b></a>
<p>Given that we provide core collection interfaces behind which
programmers can "hide" their own implementations, there will be
aliased collections whether the JDK provides them or not.
Eliminating all views from the JDK would greatly increase the cost
of common operations like making a Collection out of an array, and
would do away with many useful facilities (like synchronizing
wrappers). One view that we see as being particularly useful is
<a href=
"../List.html#subList(int,int)">List.subList</a>.
The existence of this method means that people who write methods
taking List on input do not have to write secondary forms taking an
offset and a length (as they do for arrays).</p>
</li>
<li><a id="a27"><b>Why don't you provide for
"observable" collections that send out Events when they're
modified?</b></a>
<p>Primarily, resource constraints. If we're going to commit to
such an API, it has to be something that works for everyone, that
we can live with for the long haul. We may provide such a facility
some day. In the meantime, it's not difficult to implement such a
facility on top of the public APIs.</p>
</li>
</ol>
<hr>
<p style="font-size:smaller">
Copyright &copy; 1998, 2017, Oracle and/or its affiliates. 500 Oracle Parkway<br>
    Redwood Shores, CA 94065 USA. All rights reserved.</p>
<!-- Body text ends here -->
</body>
</html>