Copyright © 2013 Jirka Kosek and John Lumley, published by the

This specification was published by the

This proposal provides an API for XPath 2.0 to handle binary data. It defines extension functions to process data from binary files, including extracting subparts, searching, basic binary operations and conversion between binary and structured forms. It has been designed to be compatible with XQuery 1.0 and XSLT 2.0, as well as any other XPath 2.0 usage.

revisiondesc

This document is in an interim draft stage. Comments are welcomed at

The module defined by this document defines several functions, all contained in the
namespace `http://expath.org/ns/binary`

. In this document, the
`bin`

prefix, when used, is bound to this namespace URI.

Error codes are defined in the same namespace
(`http://expath.org/ns/binary`

), and in this document are displayed
with the same prefix, `bin`

.

Binary file I/O uses facilities defined in the `http://expath.org/ns/file`

. In this document,
the `file`

prefix, when used, is bound to this namespace URI.

Error conditions are identified by a code (a `QName`

.) When such an error
condition is reached in the evaluation of an expression, a dynamic error is thrown,
with the corresponding error code (as if the standard XPath function
`error()`

had been called.)

The principal binary type within this module is `xs:base64Binary`

.

Conversion to and from `xs:hexBinary`

can be performed by casting with

As these types are normally implemented as wrappers around byte array structures containing the data, and differ only when being serialized to or parsed from text, such casting in-process should not involve data copying.

Development of this specification was driven by requirements which some XML developers regularly encounter in examining or generating data which is presented in binary, or other non-textual forms. Some typical use cases include:

Getting the dimensions of an image file.

Extracting image metadata.

Processing images embedded as base64 encodings within a SOAP message.

Processing legacy text files which use different encodings in separate sections.

Generating PDF files from SVG graphical data.

As an example, the following code reads the binary form of a JPEG image file, searches for the 'Start of Frame/DCT' segment, and unpacks the relevant binary sections to integers of height and width:

(The `'most-significant-first'()`

argument ensures the numeric conversion
is 'big-endian', which is the format in JPEG.)

Type - 1 octet - in this case the value `0x02`

Length – >=1 octet – the number of octets in the integer value. The length field itself can be variable in length - to accomodate VERY large integers (requiring more than 127 octets to represent, e.g. 2048-bit crypto keys.)

Payload – >=0 octets – the octets of the integer value in most-significant-first order.

To generate such a representation for an integer from XSLT/XPath, the following code might be used:

The function `bin:int-octets()`

returns a sequence of all the
'significant' octets of the integer (i.e. eliminating leading 'zeroes') in
most-significant order. Examples of the encoding are:

The first example requires no octets to encode zero, hence its octets are
`2,0`

. Both the second and third examples can be represented in less
than 128 octets (2 and 15 respectively), so length is encoded as a single octet. The
first three octets of the result for the last example, which encodes a 900-digit
integer, are: `2,130,1`

indicating that the data is represented by
(130-128) * 256 + 1 = 513 octets and the length required two octets to encode.

Decoding is a matter of compound use of the integer decoding function:

(all numbers in ASN are 'big-endian') and the examples from above reverse:

This module defines no specific functions for reading and writing binary data from
files. The EXPath File Module

There may be an argument for a positioned file:read-binary($file as

Users of the package may need to define binary 'constants' within their code or examine the basic octets. The following functions support these:

Returns the binary form of the set of octets written as a sequence of (ASCII) hex digits ([0-9A-Fa-f]).

`$in`

will be effectively zero-padded from the left to generate an integral
number of octets, i.e. an even number of hexadecimal digits. If `$in`

is an
empty string, then the result will be a `xs:base64Binary`

with no embedded
data.

Byte order in the result follows (per-octet) character order in the string.

If the value of `$in`

is the empty sequence, the function returns an empty
sequence.

`$in`

cannot be parsed as a
hexadecimal number.

When the input string has an even number of characters, this function behaves similarly
to the double cast `xs:base64Binary(xs:hexBinary(`

.*$string*))

Returns the binary form of the set of octets written as a sequence of (8-wise) (ASCII) binary digits ([01]).

`$in`

will be effectively zero-padded from the left to generate an integral
number of octets. If `$in`

is an empty string, then the result will be a
`xs:base64Binary`

with no embedded data.

Byte order in the result follows (per-octet) character order in the string.

If the value of `$in`

is the empty sequence, the function returns an empty
sequence.

`$in`

cannot be parsed as a
binary number.

Returns the binary form of the set of octets written as a sequence of (ASCII) octal digits ([0-7]).

`$in`

will be effectively zero-padded from the left to generate an integral
number of octets. If `$in`

is an empty string, then the result will be a
`xs:base64Binary`

with no embedded data.

Byte order in the result follows (per-octet) character order in the string.

If the value of `$in`

is the empty sequence, the function returns an empty
sequence.

`$in`

cannot be parsed as an
octal number.

Returns binary data as a sequence of octets.

If `$in`

is a zero length binary data then the empty sequence is
returned.

Octets are returned as integers from 0 to 255.

Converts a sequence of octets into binary data.

Octets are integers from 0 to 255.

If the value of `$in`

is the empty sequence, the function returns zero-sized
binary data.

The `bin:length`

function returns the size of binary data in octets.

Returns the size of binary data in octets.

The `bin:part`

function returns a specified part of binary data.

Returns a section of binary data starting at the `$offset`

octet. If
`$size`

is defined, the size of the returned binary data is
`$size`

octets. If `$size`

is absent, all remaining data from
`$offset`

is returned.

The `$offset`

is zero based.

The values of `$offset`

and `$size`

It is a dynamic error if `$offset`

+ `$size`

is larger than the
size of the binary data in `$in`

.

If the value of `$in`

is the empty sequence, the function returns an empty
sequence.

`$offset`

is negative.

`$size`

is negative.

`$offset + $size`

extends
beyond the binary data of `$in`

.

Note that `fn:subsequence()`

and `fn:substring()`

`xs:double`

for offset and size – this is a
legacy from XPath 1.0.

Testing whether `$data`

variable starts with binary content consistent
with a PDF file:

`25504446`

is the magic number for PDF files: it is the US-ASCII encoded
hexadecimal value for `%PDF`

.

Returns the binary data created by concatenating the binary data items in a sequence.

The function returns an `xs:base64Binary`

created by concatenating the items
in the sequence `$in`

, in order.

If the value of `$in`

is the empty sequence, the function returns a binary
item containing no data bytes.

The `bin:insert-before`

function inserts additional binary data at a given
point in other binary data.

Returns binary data consisting sequentially of the data from `$in`

upto and
including the `$offset - 1`

octet, followed by all the data from
`$extra`

, and then the remaining data from `$in`

.

The `$offset`

is zero based.

The value of `$offset`

It is a dynamic error if `$offset`

is larger than the size of the binary data
in `$in`

.

If the value of `$in`

is the empty sequence, the function returns an empty
sequence.

If the value of `$extra`

is the empty sequence, the function returns
`$in`

.

If `$offset eq 0`

the result is the binary concatenation of `$in`

and `$extra`

.

`$offset`

is negative.

`$offset`

extends beyond
the binary data of `$in`

.

Note that under non-error conditions and when `$offset gt 0`

the function is
equivalent to:

Returns the binary data created by padding `$in`

with `$size`

octets from the left. The padding octet values are `$octet`

or zero if
omitted.

The function returns an `xs:base64Binary`

created by padding the input with
`$size`

octets *in front of* the input. If `$octet`

is
specified, the padding octets each have that value, otherwise they are blank.

`$size`

If the value of `$in`

is the empty sequence, the function returns an empty
sequence.

`$size`

is negative.

`$octet`

lies outside the range
0 - 255.

Padding with a non-zero octet value can also be accomplished by the XPath expressions:

Returns the binary data created by padding `$in`

with `$size`

blank octets from the right. The padding octet values are `$octet`

or zero if
omitted.

The function returns an `xs:base64Binary`

created by padding the input with
`$size`

blank octets *after* the input. If `$octet`

is
specified, the padding octets each have that value, otherwise they are blank.

`$size`

If the value of `$in`

is the empty sequence, the function returns an empty
sequence.

`$size`

is negative.

`$octet`

lies outside the range
0 - 255.

Padding with a non-zero octet value can also be accomplished by the XPath expressions:

Returns the first location in `$in`

of `$search`

, starting at the
`$offset`

octet.

The function returns the first location of the binary search sequence in the input, or if not found, the empty sequence.

The search sequence cannot be 'empty'.

The value of `$offset`

The `$offset`

is zero based.

The returned location is zero based.

If the value of `$in`

is the empty sequence, the function returns an empty
sequence.

`$offset`

is negative.

`$offset`

extends beyond
the binary data of `$input`

.

`$search`

is empty binary
data.

Finding all the matches can be accomplished with simple recursive application:

Decodes binary data as a string in a given encoding.

If `$offset`

and `$size`

are provided, the `$size`

octets from `$offset`

are decoded. If `$offset`

alone is provided,
octets from `$offset`

to the end are decoded, otherwise the entire octet
sequence is used.

The `$encoding`

argument is the name of an encoding. The values for this
attribute follow the same rules as for the `encoding`

attribute in an XML
declaration. The only values which every implementation is `utf-8`

and `utf-16`

.

The values of `$offset`

and `$size`

If the value of `$in`

is the empty sequence, the function returns an empty
sequence.

`$offset`

is zero based.

`$offset`

is negative.

`$size`

is negative.

`$offset`

+
`$size`

- 1 extends beyond the binary data of `$in`

.

`$encoding`

is invalid or
not supported by the implementation.

Testing whether `$data`

variable starts with binary content consistent
with a PDF file:

The first four characters of a PDF file are `'%PDF'`

.

Encodes a string into binary data using a given encoding.

The `$encoding`

argument is the name of an encoding. The values for this
attribute follow the same rules as for the `encoding`

attribute in an XML
declaration. The only values which every implementation is `utf-8`

and `utf-16`

.

If the value of `$in`

is the empty sequence, the function returns an empty
sequence.

`$encoding`

is invalid or not
supported by the implementation.

Packing and unpacking numeric values can be performed in 'most-significant-first'
('big-endian') or 'least-significant-first' ('little-endian') octet order. The
default is `$octet-order`

whose string value controls the order.
Least-significant-first order is indicated by any of the values
`least-significant-first`

, `little-endian`

or
`LE`

. Most-significant-first order is indicated by any of the values
`most-significant-first`

, `big-endian`

or
`BE`

.

Integers within binary data are represented, or assumed to be represented, as an
integral number of octets. Integers where `$length`

is greater than 8
octets (and thus not representable as a `long`

) might be expected in some
situations, e.g. encryption. Whether the range of integers is limited to
`±2^63`

may be implementation-dependant.

Care should be taken with the packing and unpacking of floating point numbers
(`xs:float`

and `xs:double`

). The binary representations are
expected to correspond with those of the IEEE single/double-precision 32/64-bit
floating point types

Positive and negative infinities are supported. `INF`

maps to ```
0x7f80
0000
```

(float), `0x7ff0 0000 0000 0000`

(double).
`-INF`

maps to `0xff80 0000`

(float), ```
0xfff0 0000 0000
0000
```

(double).

Negative zero (`0x8000 0000 0000 0000`

double, `0x8000 0000`

float) encountered during unpacking will yield negative zero forms (e.g.
`-xs:double(0.0)`

) and negative zeros will be written as a result of
packing.

```
0x7fc0
0000
```

(float), `0x7ff8 0000 0000 0000`

(double). These are the
bit forms that will be packed. 'Signalling' NaN values (`0x7f80 0001`

-> `0x7fbf ffff`

or `0xff80 0001`

-> ```
0xffbf
ffff
```

, `0x7ff0 0000 0000 0001`

-> ```
0x7ff7 ffff ffff
ffff
```

or `0xfff0 0000 0000 0001`

-> ```
0xfff7 ffff ffff
ffff
```

) encountered during unpacking will be replaced by 'quiet' NaN. Any
low-order payload in a unpacked quiet NaN is also zeroed.

Returns the 8-octet binary representation of a

Most-significant-octet-first number representation is assumed unless the
`$octet-order`

parameter is specified. Acceptable values for
`$octet-order`

are described in

The binary representation will correspond with that of the IEEE double-precision 64-bit
floating point type

`$octet-order`

is
unrecognized.

Returns the 4-octet binary representation of a

Most-significant-octet-first number representation is assumed unless the
`$octet-order`

parameter is specified. Acceptable values for
`$octet-order`

are described in

The binary representation will correspond with that of the IEEE single-precision 32-bit
floating point type

`$octet-order`

is
unrecognized.

Returns the *twos-complement* binary representation of an integer value treated
as `$size`

octets long. Any 'excess' high-order bits are discarded.

Most-significant-octet-first number representation is assumed unless the
`$octet-order`

parameter is specified. Acceptable values for
`$octet-order`

are described in

Specifying a `$size`

of zero yields an empty binary data.

`$octet-order`

is
unrecognized.

`$size`

is negative.

If the integer being packed has a maximum precision of `$size`

octets, then
signed/unsigned versions are not necessary. If the data is considered unsigned, then the
most significant bit of the bottom `$size`

octets has a normal positive
(`2^(8 *$size - 1)`

) meaning. If it is considered to be a signed value,
then the MSB and all the higher order, discarded bits will be '1' for a negative value
and '0' for a positive or zero. If this function were to check the 'sizing' of the
supplied integer against the packing size, then any values of MSB and the discarded
higher order bits other than 'all 1' or 'all 0' would constitute an error. *This
function does not perfom such checking.*

Extract

`$octet-order`

parameter is specified. Acceptable values for
`$octet-order`

are described in

The value of `$offset`

The `$offset`

is zero based.

The binary representation is expected to correspond with that of the IEEE
double-precision 64-bit floating point type

`$offset`

is negative.

`$offset`

+ `7`

(octet-length of `$in`

.

`$octet-order`

is
unrecognized.

Extract

`$octet-order`

parameter is specified. Acceptable values for
`$octet-order`

are described in

The value of `$offset`

The `$offset`

is zero based.

The binary representation is expected to correspond with that of the IEEE
single-precision 32-bit floating point type

`$offset`

is negative.

`$offset`

+ `3`

(octet-length of `$in`

.

`$octet-order`

is
unrecognized.

Returns a signed integer value represented by the `$size`

octets starting
from `$offset`

in the input binary representation. Necessary sign extension
is performed (i.e. the result is negative if the high order bit is '1').

`$octet-order`

parameter is specified. Acceptable values for
`$octet-order`

are described in

The values of `$offset`

and `$size`

`$offset`

is zero based.

Specifying a `$size`

of zero yields the integer `0`

.

`$offset`

is negative.

`$size`

is negative.

`$offset`

+
`$size`

- 1 extends beyond the binary data of `$in`

.

`$octet-order`

is
unrecognized.

For discussion on integer range see

Returns an unsigned integer value represented by the `$size`

octets starting
from `$offset`

in the input binary representation.

`$octet-order`

parameter is specified. Acceptable values for
`$octet-order`

are described in

The values of `$offset`

and `$size`

The `$offset`

is zero based.

Specifying a `$size`

of zero yields the integer `0`

.

`$offset`

is negative.

`$size`

is negative.

`$offset`

+
`$size`

- 1 extends beyond the binary data of `$in`

.

`$octet-order`

is
unrecognized.

For discussion on integer range see

Returns the "bitwise or" of two binary arguments.

Returns "bitwise or" applied between `$a`

and `$b`

.

Returns the "bitwise xor" of two binary arguments.

Returns "bitwise exclusive or" applied between `$a`

and `$b`

.

Returns the "bitwise and" of two binary arguments.

Returns "bitwise and" applied between `$a`

and `$b`

.

Returns the "bitwise not" of a binary argument.

Returns "bitwise not" applied to `$in`

.

Shift bits in binary data.

If `$by`

is positive then bits are shifted `$by`

times to the
left.

If `$by`

is negative then bits are shifted `-$by`

times to the
right.

If `$by`

is zero, the result is identical to `$in`

.

If `|$by|`

is greater than the bit-length of `$in`

then an
all-zeros result, of the same length as `$in`

, is returned.

The result always has the same size as `$in`

.

The shifting is logical: zeros are placed into discarded bits.

If the value of `$in`

is the empty sequence, the function returns an empty
sequence.

Bit shifting across byte boundaries implies 'big-endian' treatment, i.e. the leftmost (high-order) bit when shifted left becomes the low-order bit of the preceding byte.