<?xml version="1.0" encoding="utf-8" ?>
<!DOCTYPE chapter SYSTEM "chapter.dtd">
<chapter>
<header>
<copyright>
<year>2001</year><year>2013</year>
<holder>Ericsson AB. All Rights Reserved.</holder>
</copyright>
<legalnotice>
The contents of this file are subject to the Erlang Public License,
Version 1.1, (the "License"); you may not use this file except in
compliance with the License. You should have received a copy of the
Erlang Public License along with this software. If not, it can be
retrieved online at http://www.erlang.org/.
Software distributed under the License is distributed on an "AS IS"
basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See
the License for the specific language governing rights and limitations
under the License.
</legalnotice>
<title>List Handling</title>
<prepared>Bjorn Gustavsson</prepared>
<docno></docno>
<date>2007-11-16</date>
<rev></rev>
<file>listHandling.xml</file>
</header>
<section>
<title>Creating a List</title>
<p>Lists can only be built starting from the end and attaching list
elements at the beginning. If you use the "<c>++</c>" operator as
follows, a new list is created that is a copy of the elements in
<c>List1</c>, followed by <c>List2</c>:</p>
<code type="erl">
List1 ++ List2</code>
<p>Looking at how <c>lists:append/1</c> or <c>++</c> would be
implemented in plain Erlang, clearly the first list is copied:</p>
<code type="erl">
append([H|T], Tail) ->
[H|append(T, Tail)];
append([], Tail) ->
Tail.</code>
<p>When recursing and building a list, it is important to ensure
that you attach the new elements to the beginning of the list. In
this way, you will build <em>one</em> list, not hundreds or thousands
of copies of the growing result list.</p>
<p>Let us first see how it is not to be done:</p>
<p><em>DO NOT</em></p>
<code type="erl"><![CDATA[
bad_fib(N) ->
bad_fib(N, 0, 1, []).
bad_fib(0, _Current, _Next, Fibs) ->
Fibs;
bad_fib(N, Current, Next, Fibs) ->
bad_fib(N - 1, Next, Current + Next, Fibs ++ [Current]).]]></code>
<p>Here more than one list is built. In each iteration step a new list
is created that is one element longer than the new previous list.</p>
<p>To avoid copying the result in each iteration, build the list in
reverse order and reverse the list when you are done:</p>
<p><em>DO</em></p>
<code type="erl"><![CDATA[
tail_recursive_fib(N) ->
tail_recursive_fib(N, 0, 1, []).
tail_recursive_fib(0, _Current, _Next, Fibs) ->
lists:reverse(Fibs);
tail_recursive_fib(N, Current, Next, Fibs) ->
tail_recursive_fib(N - 1, Next, Current + Next, [Current|Fibs]).]]></code>
</section>
<section>
<title>List Comprehensions</title>
<p>Lists comprehensions still have a reputation for being slow.
They used to be implemented using funs, which used to be slow.</p>
<p>In recent Erlang/OTP releases (including R12B), a list comprehension:</p>
<code type="erl"><![CDATA[
[Expr(E) || E <- List]]]></code>
<p>is basically translated to a local function:</p>
<code type="erl">
'lc^0'([E|Tail], Expr) ->
[Expr(E)|'lc^0'(Tail, Expr)];
'lc^0'([], _Expr) -> [].</code>
<p>In R12B, if the result of the list comprehension will <em>obviously</em>
not be used, a list will not be constructed. For example, in this code:</p>
<code type="erl"><![CDATA[
[io:put_chars(E) || E <- List],
ok.]]></code>
<p>or in this code:</p>
<code type="erl"><![CDATA[
...
case Var of
... ->
[io:put_chars(E) || E <- List];
... ->
end,
some_function(...),
...]]></code>
<p>the value is not assigned to a variable, not passed to another function,
and not returned. This means that there is no need to construct a list and
the compiler will simplify the code for the list comprehension to:</p>
<code type="erl">
'lc^0'([E|Tail], Expr) ->
Expr(E),
'lc^0'(Tail, Expr);
'lc^0'([], _Expr) -> [].</code>
</section>
<section>
<title>Deep and Flat Lists</title>
<p><seealso marker="stdlib:lists#flatten/1">lists:flatten/1</seealso>
builds an entirely new list. It is therefore expensive, and even
<em>more</em> expensive than the <c>++</c> operator (which copies its
left argument, but not its right argument).</p>
<p>In the following situations, you can easily avoid calling
<c>lists:flatten/1</c>:</p>
<list type="bulleted">
<item>When sending data to a port. Ports understand deep lists
so there is no reason to flatten the list before sending it to
the port.</item>
<item>When calling BIFs that accept deep lists, such as
<seealso marker="erts:erlang#list_to_binary/1">list_to_binary/1</seealso> or
<seealso marker="erts:erlang#iolist_to_binary/1">iolist_to_binary/1</seealso>.</item>
<item>When you know that your list is only one level deep, you can use
<seealso marker="stdlib:lists#append/1">lists:append/1</seealso>.</item>
</list>
<section>
<title>Port Example</title>
<p><em>DO</em></p>
<pre>
...
port_command(Port, DeepList)
...</pre>
<p><em>DO NOT</em></p>
<pre>
...
port_command(Port, lists:flatten(DeepList))
...</pre>
<p>A common way to send a zero-terminated string to a port is the following:</p>
<p><em>DO NOT</em></p>
<pre>
...
TerminatedStr = String ++ [0], % String="foo" => [$f, $o, $o, 0]
port_command(Port, TerminatedStr)
...</pre>
<p>Instead:</p>
<p><em>DO</em></p>
<pre>
...
TerminatedStr = [String, 0], % String="foo" => [[$f, $o, $o], 0]
port_command(Port, TerminatedStr)
...</pre>
</section>
<section>
<title>Append Example</title>
<p><em>DO</em></p>
<pre>
> lists:append([[1], [2], [3]]).
[1,2,3]
></pre>
<p><em>DO NOT</em></p>
<pre>
> lists:flatten([[1], [2], [3]]).
[1,2,3]
></pre>
</section>
</section>
<section>
<title>Recursive List Functions</title>
<p>In Section 7.2, the following myth was exposed:
<seealso marker="myths#tail_recursive">Tail-Recursive Functions
are Much Faster Than Recursive Functions</seealso>.</p>
<p>To summarize, in R12B there is usually not much difference between
a body-recursive list function and tail-recursive function that reverses
the list at the end. Therefore, concentrate on writing beautiful code
and forget about the performance of your list functions. In the
time-critical parts of your code (and only there), <em>measure</em>
before rewriting your code.</p>
<note><p>This section is about list functions that <em>construct</em>
lists. A tail-recursive function that does not construct a list runs
in constant space, while the corresponding body-recursive function
uses stack space proportional to the length of the list.</p></note>
<p>For example, a function that sums a list of integers, is
<em>not</em> to be written as follows:</p>
<p><em>DO NOT</em></p>
<code type="erl">
recursive_sum([H|T]) -> H+recursive_sum(T);
recursive_sum([]) -> 0.</code>
<p>Instead:</p>
<p><em>DO</em></p>
<code type="erl">
sum(L) -> sum(L, 0).
sum([H|T], Sum) -> sum(T, Sum + H);
sum([], Sum) -> Sum.</code>
</section>
</chapter>
