path: root/system/doc/extensions/bit_syntax.xml

<?xml version="1.0" encoding="latin1" ?>
<!DOCTYPE chapter SYSTEM "chapter.dtd">

<chapter>
  <header>
    <copyright>
      <year>2000</year><year>2009</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>The Bit Syntax</title>
    <prepared>Bj&ouml;rn Gustavsson</prepared>
    <responsible>Bjarne D&auml;cker</responsible>
    <docno>1</docno>
    <approved>Bjarne D&auml;Ker</approved>
    <checked></checked>
    <date>00-06-21</date>
    <rev>PA1</rev>
    <file>bit_syntax.sgml</file>
  </header>
  <p>This section describes the "bit syntax" which was added to
    the Erlang language in release 5.0 (R7).
    Compared to the original bit syntax prototype
    by Claes Wikstr&ouml;m and Tony Rogvall (presented on the
    Erlang User's Conference 1999), this implementation differs
    primarily in the following respects,
    </p>
  <list type="ordered">
    <item>
      <p>the character pairs '&lt;&lt;' and '&gt;&gt;' are used to delimit
        a binary patterns and constructor (not '&lt;' and '&gt;' as in
        the prototype),
        </p>
    </item>
    <item>
      <p>the tail syntax ('|Variable') has been eliminated,
        </p>
    </item>
    <item>
      <p>all size expressions must be bound,
        </p>
    </item>
    <item>
      <p>a type <c>unit:U</c> has been added,
        </p>
    </item>
    <item>
      <p>lists and tuples cannot be generated
        </p>
    </item>
    <item>
      <p>there are no paddings whatsoever.
        </p>
    </item>
  </list>

  <section>
    <title>Introduction</title>
    <p>In Erlang a Bin is used for constructing binaries and 
      matching binary patterns. A Bin is written with the
      following syntax:</p>
    <code type="none"><![CDATA[
      <<E1, E2, ... En>>
    ]]></code>
    <p>A Bin is a low-level sequence of bytes. The purpose of a
      Bin is to be able to, from a high level,
      <em>construct</em> a binary,
      </p>
    <code type="none"><![CDATA[
      Bin = <<E1, E2, ... En>>
    ]]></code>
    <p>in which case all elements must be bound, or to
      <em>match</em> a binary, 
      </p>
    <code type="none"><![CDATA[
      <<E1, E2, ... En>> = Bin 
    ]]></code>
    <p>where <c>Bin</c> is bound, and where the elements are bound or unbound,
      as in any match.
      </p>
    <p>Each element specifies a certain <em>segment</em> of the binary.
      A segment is is a set of contiguous bits of the binary (not
      necessarily on a byte boundary). The first element specifies
      the initial segment, the second element specifies the following
      segment etc.
      </p>
    <p>The following examples illustrate how binaries are constructed
      or matched, and how elements and tails are specified.
      </p>

    <section>
      <title>Examples</title>
      <p><em>Example 1: </em>A binary can be constructed from a set of
        constants or a string literal:</p>
      <code type="none"><![CDATA[
        Bin11 = <<1, 17, 42>>,
        Bin12 = <<"abc">>
      ]]></code>
      <p>yields binaries of size 3; <c>binary_to_list(Bin11)</c>
        evaluates to <c>[1, 17, 42]</c>, and
        <c>binary_to_list(Bin12)</c> evaluates to <c>[97, 98, 99]</c>.
        </p>
      <p><em>Example 2: </em>Similarly, a binary can be constructed
        from a set of bound variables:</p>
      <code type="none"><![CDATA[
        A = 1, B = 17, C = 42,
        Bin2 = <<A, B, C:16>>
      ]]></code>
      <p>yields a binary of size 4, and <c>binary_to_list(Bin2)</c>
        evaluates to <c>[1, 17, 00, 42]</c> too. Here we used a
        <em>size expression</em> for the variable <c>C</c> in order to
        specify a 16-bits segment of <c>Bin2</c>.
        </p>
      <p><em>Example 3: </em>A Bin can also be used for matching: if
        <c>D</c>, <c>E</c>, and <c>F</c> are unbound variables, and
        <c>Bin2</c> is bound as in the former example,</p>
      <code type="none"><![CDATA[
        <<D:16, E, F/binary>> = Bin2
      ]]></code>
      <p>yields <c>D = 273</c>, <c>E = 00</c>, and F binds to a binary
        of size 1: <c>binary_to_list(F) = [42]</c>.
        </p>
      <p><em>Example 4: </em>The following is a more elaborate example
        of matching, where <c>Dgram</c> is bound to the consecutive
        bytes of an IP datagram of IP protocol version 4, and where we
        want to extract the header and the data of the datagram:</p>
      <code type="none"><![CDATA[
        -define(IP_VERSION, 4).
        -define(IP_MIN_HDR_LEN, 5).

        DgramSize = byte_size(Dgram),
        case Dgram of 
        <<?IP_VERSION:4, HLen:4, SrvcType:8, TotLen:16, 
        ID:16, Flgs:3, FragOff:13,
        TTL:8, Proto:8, HdrChkSum:16,
        SrcIP:32,
        DestIP:32, RestDgram/binary>> when HLen >= 5, 4*HLen =< DgramSize ->
        OptsLen = 4*(HLen - ?IP_MIN_HDR_LEN),
        <<Opts:OptsLen/binary,Data/binary>> = RestDgram,
        ...
        end.
      ]]></code>
      <p>Here the segment corresponding to the <c>Opts</c> variable
        has a <em>type modifier</em> specifying that <c>Opts</c> should
        bind to a binary. All other variables have the default type
        equal to unsigned integer.
        </p>
      <p>An IP datagram header is of variable length, and its length -
        measured in the number of 32-bit words - is given in the segment
        corresponding to <c>HLen</c>, the minimum value of which is
        5. It is the segment corresponding to <c>Opts</c> that is
        variable: if <c>HLen</c> is equal to 5, <c>Opts</c> will be an
        empty binary.
        </p>
      <p>The tail variables <c>RestDgram</c> and <c>Data</c> bind to
        binaries, as all tail variables do. Both may bind to empty
        binaries.
        </p>
      <p>If the first 4-bits segment of <c>Dgram</c> is not equal to
        4, or if <c>HLen</c> is less than 5, or if the size of
        <c>Dgram</c> is less than <c>4*HLen</c>, the match of
        <c>Dgram</c> fails.
        </p>
    </section>
  </section>

  <section>
    <title>A Lexical Note</title>
    <p>Note that "<c><![CDATA[B=<<1>>]]></c>" will be interpreted as
      "<c><![CDATA[B =< ;<1>>]]></c>", which is a syntax error.
      The correct way to write the expression is "<c><![CDATA[B = <<1>>]]></c>".</p>
  </section>

  <section>
    <title>Segments</title>
    <p>Each segment has the following general syntax:</p>
    <p><c>Value:Size/TypeSpecifierList</c></p>
    <p>Both the <c>Size</c> and the <c>TypeSpecifier</c> or both may be
      omitted; thus the following variations are allowed:
      </p>
    <p><c>Value</c></p>
    <p><c>Value:Size</c></p>
    <p><c>Value/TypeSpecifierList</c></p>
    <p>Default values will be used for missing specifications. The default
      values are described in the section "Defaults" below.
      </p>
    <p>Used in binary construction, the <c>Value</c> part is any expression.
      Used in binary matching, the <c>Value</c> part must be a literal or
      variable. You can read more about the <c>Value</c> part in the
      sections about constructing binaries and matching binaries.
      </p>
    <p>The <c>Size</c> part of the segment multiplied by the unit in the
      <c>TypeSpecifierList</c> (described below) gives the number of bits
      for the segment. In construction, <c>Size</c> is any expression that
      evaluates to an integer. In matching, <c>Size</c> must be a constant
      expression or a variable.
      </p>
    <p>The <c>TypeSpecifierList</c> is a list of type specifiers separated by
      hyphens.
      </p>
    <taglist>
      <tag>Type</tag>
      <item>The type can be <c>integer</c>, <c>float</c>, or <c>binary</c>.</item>
      <tag>Signedness</tag>
      <item>The signedness specification can be either <c>signed</c>
       or <c>unsigned</c>. Note that signedness only matters for matching.</item>
      <tag>Endianness</tag>
      <item>The endianness specification can be either <c>big</c>,
      <c>little</c>, or <c>native</c>. Native-endian means that
       the endian will be resolved at load time to be either
       big-endian or little-endian, depending on what is "native"
       for the CPU that the Erlang machine is run on.</item>
      <tag>Unit</tag>
      <item>The unit size is given as <c>unit:IntegerLiteral</c>.
       The allowed range is 1-256. It will be multiplied by the <c>Size</c>
       specifier to give the effective size of the segment.</item>
    </taglist>
    <p>Example:
      </p>
    <code type="none">
X:4/little-signed-integer-unit:8
    </code>
    <p>This element has a total size of 4*8 = 32 bits, and it contains a
      signed integer in little-endian order.</p>
  </section>

  <section>
    <title>Defaults</title>
    <p>The default type for a segment is <c>integer</c>. The default type
      does <em>not</em> depend on the value, even if the value is a literal.
      For instance, the default type in '<c><![CDATA[<<3.14>>]]></c>' is <c>integer</c>,
      not <c>float</c>.
      </p>
    <p>The default <c>Size</c> depends on the type.
      For <c>integer</c> it is 8. For <c>float</c> it is 64.
      For <c>binary</c> it is all of the binary. In matching, this default
      value is only valid for the very last element. All other binary elements
      in matching must have a size specification.
      </p>
    <p>The default unit depends on the the type. 
      For <c>integer</c> and <c>float</c> it is 1.
      For <c>binary</c> it is 8.
      </p>
    <p>The default signedness is <c>unsigned</c>.
      </p>
    <p>The default endianness is <c>big</c>.</p>
  </section>

  <section>
    <title>Constructing Binaries</title>
    <p>This section describes the rules for constructing binaries using
      the bit syntax. Unlike when constructing lists or tuples, the construction
      of a binary can fail with a <c>badarg</c> exception.
      </p>
    <p>There can be zero or more segments in a binary to be constructed.
      The expression '<c><![CDATA[<<>>]]></c>' constructs a zero length binary.
      </p>
    <p>Each segment in a binary can consist of zero or more bits.
      There are no alignment rules for individual segments, but the total
      number of bits in all segments must be evenly divisible by 8,
      or in other words, the resulting binary must consist of a whole number
      of bytes. An <c>badarg</c> exception will be thrown if the resulting
      binary is not byte-aligned. Example:
      </p>
    <code type="none"><![CDATA[
<<X:1,Y:6>>
    ]]></code>
    <p>The total number of bits is 7, which is not evenly divisible by 8;
      thus, there will be <c>badarg</c> exception (and a compiler warning
      as well). The following example
      </p>
    <code type="none"><![CDATA[
<<X:1,Y:6,Z:1>>
    ]]></code>
    <p>will successfully construct a binary of 8 bits, or one byte. (Provided
      that all of X, Y and Z are integers.)
      </p>
    <p>As noted earlier, segments have the following general syntax:
      </p>
    <p><c>Value:Size/TypeSpecifierList</c></p>
    <p>When constructing binaries, <c>Value</c> and <c>Size</c> can be
      any Erlang expression. However, for syntactical reasons,
      both <c>Value</c> and <c>Size</c> must be enclosed in parenthesis
      if the expression consists of anything more than a single literal
      or variable. The following gives a compiler syntax error:
      </p>
    <code type="none"><![CDATA[
<<X+1:8>>
    ]]></code>
    <p>This expression must be rewritten to
      </p>
    <code type="none"><![CDATA[
<<(X+1):8>>
    ]]></code>
    <p>in order to be accepted by the compiler.
      </p>

    <section>
      <title>Including Literal Strings</title>
      <p>As syntactic sugar, an literal string may be written instead of a
        element.</p>
      <code type="none"><![CDATA[
<<"hello">>      ]]></code>
      <p>which is syntactic sugar for</p>
      <code type="none"><![CDATA[
<<$h,$e,$l,$l,$o>>      ]]></code>
    </section>
  </section>

  <section>
    <title>Matching Binaries</title>
    <p>This section describes the rules for matching binaries using the
      bit syntax.
      </p>
    <p>There can be zero or more segments in a binary binary pattern.
      A binary pattern can occur in every place patterns are allowed, also
      inside other patterns. Binary patterns cannot be nested.
      </p>
    <p>The pattern '<c><![CDATA[<<>>]]></c>' matches a zero length binary.
      </p>
    <p>Each segment in a binary can consist of zero or more bits.
      </p>
    <p>A segment of type <c>binary</c> must have a size evenly divisible by 8.
      </p>
    <p>This means that the following head will never match:</p>
    <code type="none"><![CDATA[
foo(<<X:7/binary,Y:1/binary>>) ->    ]]></code>
    <p>As noted earlier, segments have the following general syntax:
      </p>
    <p><c>Value:Size/TypeSpecifierList</c></p>
    <p>When matching <c>Value</c> value must be either a variable or an integer
      or floating point literal. Expressions are not allowed.
      </p>
    <p><c>Size</c> must be an integer literal, or a previously bound variable.
      Note that the following is not allowed:</p>
    <code type="none"><![CDATA[
foo(N, <<X:N,T/binary>>) ->
   {X,T}.    ]]></code>
    <p>The two occurrences of <c>N</c> are not related. The compiler
      will complain that the <c>N</c> in the size field is unbound.
      </p>
    <p>The correct way to write this example is like this:</p>
    <code type="none"><![CDATA[
foo(N, Bin) ->
   <<X:N,T/binary>> = Bin,
   {X,T}.    ]]></code>

    <section>
      <title>Getting the Rest of the Binary</title>
      <p>To match out the rest of binary, specify a binary field without size:</p>
      <code type="none"><![CDATA[
foo(<<A:8,Rest/binary>>) ->      ]]></code>
      <p>As always, the size of the tail must be evenly divisible by 8.
        </p>
    </section>
  </section>

  <section>
    <title>Traps and Pitfalls</title>
    <p>Assume that we need a function that creates a binary out of a
      list of triples of integers. A first (inefficient) version of such
      a function could look like this:</p>
    <code type="none"><![CDATA[
triples_to_bin(T) ->
    triples_to_bin(T, <<>>).

triples_to_bin([{X,Y,Z} | T], Acc) ->
    triples_to_bin(T, <<Acc/binary, X:32, Y:32, Z:32>>);   % inefficient
triples_to_bin([], Acc) -> 
    Acc.    ]]></code>
    <p>The reason for the inefficiency of this function is that for
      each triple, the binary constructed so far (<c>Acc</c>) is copied.
      (Note: The original bit syntax prototype avoided the copy operation
      by using segmented binaries, which are not implemented in R7.)
      </p>
    <p>The efficient way to write this function in R7 is:</p>
    <code type="none"><![CDATA[
triples_to_bin(T) ->
    triples_to_bin(T, []).

triples_to_bin([{X,Y,Z} | T], Acc) ->
    triples_to_bin(T, [<<X:32, Y:32, Z:32>> | Acc]);
triples_to_bin([], Acc) -> 
    list_to_binary(lists:reverse(Acc)).    ]]></code>
    <p>Note that <c>list_to_binary/1</c> handles deep lists of binaries
      and small integers. (This fact was previously undocumented.)
      </p>
  </section>
</chapter>