20002009 Ericsson AB. All Rights Reserved. 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. The Bit Syntax Björn Gustavsson Bjarne Däcker 1 Bjarne DäKer 00-06-21 PA1 bit_syntax.sgml

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 Wikström and Tony Rogvall (presented on the Erlang User's Conference 1999), this implementation differs primarily in the following respects,

the character pairs '<<' and '>>' are used to delimit a binary patterns and constructor (not '<' and '>' as in the prototype),

the tail syntax ('|Variable') has been eliminated,

all size expressions must be bound,

a type unit:U has been added,

lists and tuples cannot be generated

there are no paddings whatsoever.

Introduction

In Erlang a Bin is used for constructing binaries and matching binary patterns. A Bin is written with the following syntax:

> ]]>

A Bin is a low-level sequence of bytes. The purpose of a Bin is to be able to, from a high level, construct a binary,

> ]]>

in which case all elements must be bound, or to match a binary,

> = Bin ]]>

where Bin is bound, and where the elements are bound or unbound, as in any match.

Each element specifies a certain segment 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.

The following examples illustrate how binaries are constructed or matched, and how elements and tails are specified.

Examples

Example 1: A binary can be constructed from a set of constants or a string literal:

>, Bin12 = <<"abc">> ]]>

yields binaries of size 3; binary_to_list(Bin11) evaluates to [1, 17, 42], and binary_to_list(Bin12) evaluates to [97, 98, 99].

Example 2: Similarly, a binary can be constructed from a set of bound variables:

> ]]>

yields a binary of size 4, and binary_to_list(Bin2) evaluates to [1, 17, 00, 42] too. Here we used a size expression for the variable C in order to specify a 16-bits segment of Bin2.

Example 3: A Bin can also be used for matching: if D, E, and F are unbound variables, and Bin2 is bound as in the former example,

> = Bin2 ]]>

yields D = 273, E = 00, and F binds to a binary of size 1: binary_to_list(F) = [42].

Example 4: The following is a more elaborate example of matching, where Dgram 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:

> when HLen >= 5, 4*HLen =< DgramSize -> OptsLen = 4*(HLen - ?IP_MIN_HDR_LEN), <> = RestDgram, ... end. ]]>

Here the segment corresponding to the Opts variable has a type modifier specifying that Opts should bind to a binary. All other variables have the default type equal to unsigned integer.

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 HLen, the minimum value of which is 5. It is the segment corresponding to Opts that is variable: if HLen is equal to 5, Opts will be an empty binary.

The tail variables RestDgram and Data bind to binaries, as all tail variables do. Both may bind to empty binaries.

If the first 4-bits segment of Dgram is not equal to 4, or if HLen is less than 5, or if the size of Dgram is less than 4*HLen, the match of Dgram fails.

A Lexical Note

Note that ">]]>" will be interpreted as ">]]>", which is a syntax error. The correct way to write the expression is ">]]>".

Segments

Each segment has the following general syntax:

Value:Size/TypeSpecifierList

Both the Size and the TypeSpecifier or both may be omitted; thus the following variations are allowed:

Value

Value:Size

Value/TypeSpecifierList

Default values will be used for missing specifications. The default values are described in the section "Defaults" below.

Used in binary construction, the Value part is any expression. Used in binary matching, the Value part must be a literal or variable. You can read more about the Value part in the sections about constructing binaries and matching binaries.

The Size part of the segment multiplied by the unit in the TypeSpecifierList (described below) gives the number of bits for the segment. In construction, Size is any expression that evaluates to an integer. In matching, Size must be a constant expression or a variable.

The TypeSpecifierList is a list of type specifiers separated by hyphens.

Type The type can be integer, float, or binary. Signedness The signedness specification can be either signed or unsigned. Note that signedness only matters for matching. Endianness The endianness specification can be either big, little, or native. 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. Unit The unit size is given as unit:IntegerLiteral. The allowed range is 1-256. It will be multiplied by the Size specifier to give the effective size of the segment.

Example:

X:4/little-signed-integer-unit:8

This element has a total size of 4*8 = 32 bits, and it contains a signed integer in little-endian order.

Defaults

The default type for a segment is integer. The default type does not depend on the value, even if the value is a literal. For instance, the default type in '>]]>' is integer, not float.

The default Size depends on the type. For integer it is 8. For float it is 64. For binary 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.

The default unit depends on the the type. For integer and float it is 1. For binary it is 8.

The default signedness is unsigned.

The default endianness is big.

Constructing Binaries

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 badarg exception.

There can be zero or more segments in a binary to be constructed. The expression '>]]>' constructs a zero length binary.

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 badarg exception will be thrown if the resulting binary is not byte-aligned. Example:

> ]]>

The total number of bits is 7, which is not evenly divisible by 8; thus, there will be badarg exception (and a compiler warning as well). The following example

> ]]>

will successfully construct a binary of 8 bits, or one byte. (Provided that all of X, Y and Z are integers.)

As noted earlier, segments have the following general syntax:

Value:Size/TypeSpecifierList

When constructing binaries, Value and Size can be any Erlang expression. However, for syntactical reasons, both Value and Size 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:

> ]]>

This expression must be rewritten to

> ]]>

in order to be accepted by the compiler.

Including Literal Strings

As syntactic sugar, an literal string may be written instead of a element.

> ]]>

which is syntactic sugar for

> ]]>
Matching Binaries

This section describes the rules for matching binaries using the bit syntax.

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.

The pattern '>]]>' matches a zero length binary.

Each segment in a binary can consist of zero or more bits.

A segment of type binary must have a size evenly divisible by 8.

This means that the following head will never match:

>) -> ]]>

As noted earlier, segments have the following general syntax:

Value:Size/TypeSpecifierList

When matching Value value must be either a variable or an integer or floating point literal. Expressions are not allowed.

Size must be an integer literal, or a previously bound variable. Note that the following is not allowed:

>) -> {X,T}. ]]>

The two occurrences of N are not related. The compiler will complain that the N in the size field is unbound.

The correct way to write this example is like this:

<> = Bin, {X,T}. ]]>
Getting the Rest of the Binary

To match out the rest of binary, specify a binary field without size:

>) -> ]]>

As always, the size of the tail must be evenly divisible by 8.

Traps and Pitfalls

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:

triples_to_bin(T, <<>>). triples_to_bin([{X,Y,Z} | T], Acc) -> triples_to_bin(T, <>); % inefficient triples_to_bin([], Acc) -> Acc. ]]>

The reason for the inefficiency of this function is that for each triple, the binary constructed so far (Acc) is copied. (Note: The original bit syntax prototype avoided the copy operation by using segmented binaries, which are not implemented in R7.)

The efficient way to write this function in R7 is:

triples_to_bin(T, []). triples_to_bin([{X,Y,Z} | T], Acc) -> triples_to_bin(T, [<> | Acc]); triples_to_bin([], Acc) -> list_to_binary(lists:reverse(Acc)). ]]>

Note that list_to_binary/1 handles deep lists of binaries and small integers. (This fact was previously undocumented.)