The value of _array_data.data contains the array data
encapsulated in a STAR string.
The representation used is a variant on the
Multipurpose Internet Mail Extensions (MIME) specified
in RFC 2045-2049 by N. Freed et al. The boundary
delimiter used in writing an imgCIF or CBF is
'\n--CIF-BINARY-FORMAT-SECTION--' (including the
required initial '\n--').
The Content-Type may be any of the discrete types permitted
in RFC 2045; 'application/octet-stream' is recommended
for diffraction images in the ARRAY_DATA category.
Note: When appropriate in other categories, e.g. for
photographs of crystals, more precise types, such as
'image/jpeg', 'image/tiff', 'image/png', etc. should be used.
If an octet stream was compressed, the compression should
be specified by the parameter
or the parameter
or the parameter
or the parameter
If the parameter
is given it may be further modified with the parameters
If the '"uncorrelated_sections"' parameter is
given, each section will be compressed without using
the prior section for averaging.
If the '"flat"' parameter is given, each the
image will be treated as one long row.
Note that the X-CBF_CANONICAL and X-CBF_PACKED are
slower but more efficient compressions that the others.
The X-CBF_BYTE_OFFSET compression is a good compromise
between speed and efficiency for ordinary diffraction
images. The X-CBF_BACKGROUND_OFFSET_DELTA compression
is oriented towards sparse data, such as masks and
tables of replacement pixel values for images with
The Content-Transfer-Encoding may be 'BASE64',
'Quoted-Printable', 'X-BASE8', 'X-BASE10',
'X-BASE16' or 'X-BASE32K', for an imgCIF or 'BINARY'
for a CBF. The octal, decimal and hexadecimal transfer
encodings are provided for convenience in debugging and
are not recommended for archiving and data interchange.
In a CIF, one of the parameters 'charset=us-ascii',
'charset=utf-8' or 'charset=utf-16' may be used on the
Content-Transfer-Encoding to specify the character set
used for the external presentation of the encoded data.
If no charset parameter is given, the character set of
the enclosing CIF is assumed. In any case, if a BOM
flag is detected (FE FF for big-endian UTF-16, FF FE for
little-endian UTF-16 or EF BB BF for UTF-8) is detected,
the indicated charset will be assumed until the end of the
encoded data or the detection of a different BOM. The
charset of the Content-Transfer-Encoding is not the character
set of the encoded data, only the character set of the
presentation of the encoded data and should be respecified
for each distinct STAR string.
In an imgCIF file, the encoded binary data begins after
the empty line terminating the header. In an imgCIF file,
the encoded binary data ends with the terminating boundary
in the currently effective charset or with the '\n; '
that terminates the STAR string.
In a CBF, the raw binary data begins after an empty line
terminating the header and after the sequence:
Octet Hex Decimal Purpose
0 0C 12 (ctrl-L) Page break
1 1A 26 (ctrl-Z) Stop listings in MS-DOS
2 04 04 (Ctrl-D) Stop listings in UNIX
3 D5 213 Binary section begins
None of these octets are included in the calculation of
the message size or in the calculation of the
The X-Binary-Size header specifies the size of the
equivalent binary data in octets. If compression was
used, this size is the size after compression, including
any book-keeping fields. An adjustment is made for
the deprecated binary formats in which eight bytes of binary
header are used for the compression type. In this case,
the eight bytes used for the compression type are subtracted
from the size, so that the same size will be reported
if the compression type is supplied in the MIME header.
Use of the MIME header is the recommended way to
supply the compression type. In general, no portion of
the binary header is included in the calculation of the size.
The X-Binary-Element-Type header specifies the type of
binary data in the octets, using the same descriptive
phrases as in _array_structure.encoding_type. The default
value is 'unsigned 32-bit integer'.
An MD5 message digest may, optionally, be used. The 'RSA Data
Security, Inc. MD5 Message-Digest Algorithm' should be used.
No portion of the header is included in the calculation of the
If the Transfer Encoding is 'X-BASE8', 'X-BASE10' or
'X-BASE16', the data are presented as octal, decimal or
hexadecimal data organized into lines or words. Each word
is created by composing octets of data in fixed groups of
2, 3, 4, 6 or 8 octets, either in the order ...4321 ('big-
endian') or 1234... ('little-endian'). If there are fewer
than the specified number of octets to fill the last word,
then the missing octets are presented as '==' for each
missing octet. Exactly two equal signs are used for each
missing octet even for octal and decimal encoding.
The format of lines is:
rnd xxxxxx xxxxxx xxxxxx
where r is 'H', 'O' or 'D' for hexadecimal, octal or
decimal, n is the number of octets per word and d is '<'
or '>' for the '...4321' and '1234...' octet orderings,
respectively. The '==' padding for the last word should
be on the appropriate side to correspond to the missing
H4< FFFFFFFF FFFFFFFF 07FFFFFF ====0000
H3> FF0700 00====
For these hexadecimal, octal and decimal formats only,
comments beginning with '#' are permitted to improve
BASE64 encoding follows MIME conventions. Octets are
in groups of three: c1, c2, c3. The resulting 24 bits
are broken into four six-bit quantities, starting with
the high-order six bits (c1 >> 2) of the first octet, then
the low-order two bits of the first octet followed by the
high-order four bits of the second octet [(c1 & 3)<<4 | (c2>>4)],
then the bottom four bits of the second octet followed by the
high-order two bits of the last octet [(c2 & 15)<<2 | (c3>>6)],
then the bottom six bits of the last octet (c3 & 63). Each
of these four quantities is translated into an ASCII character
using the mapping:
1 2 3 4 5 6
| | | | | | |
With short groups of octets padded on the right with one '='
if c3 is missing, and with '==' if both c2 and c3 are missing.
X-BASE32K encoding is similar to BASE64 encoding, except that
sets of 15 octets are encoded as sets of 8 16-bit unicode
characters, by breaking the 120 bits into 8 15-bit quantities.
256 is added to each 15 bit quantity to bring it into a
printable uncode range. When encoding, zero padding is used
to fill out the last 15 bit quantity. If 8 or more bits of
padding are used, a single equals sign (hexadecimal 003D) is
appended. Embedded whitespace and newlines are introduced
to produce lines of no more than 80 characters each. On
decoding, all printable ascii characters and ascii whitespace
characters are ignored except for any trailing equals signs.
The number of trailing equals signs indicated the number of
trailing octets to be trimmed from the end of the decoded data.
(see Georgi Darakev, Vassil Litchev, Kostadin Z. Mitev, Herbert
J. Bernstein, 'Efficient Support of Binary Data in the XML
Implementation of the NeXus File Format',absract W0165,
ACA Summer Meeting, Honolulu, HI, July 2006).
QUOTED-PRINTABLE encoding also follows MIME conventions, copying
octets without translation if their ASCII values are 32...38,
42, 48...57, 59, 60, 62, 64...126 and the octet is not a ';'
in column 1. All other characters are translated to =nn, where
nn is the hexadecimal encoding of the octet. All lines are
'wrapped' with a terminating '=' (i.e. the MIME conventions
for an implicit line terminator are never used).
The "X-Binary-Element-Byte-Order" can specify either
'"BIG_ENDIAN"' or '"LITTLE_ENDIAN"' byte order of the imaage
data. Only LITTLE_ENDIAN is recommended. Processors
may treat BIG_ENDIAN as a warning of data that can
only be processed by special software.
The "X-Binary-Number-of-Elements" specifies the number of
elements (not the number of octets) in the decompressed, decoded
The optional "X-Binary-Size-Fastest-Dimension" specifies the
number of elements (not the number of octets) in one row of the
fastest changing dimension of the binary data array. This
information must be in the MIME header for proper operation of
some of the decompression algorithms.
The optional "X-Binary-Size-Second-Dimension" specifies the
number of elements (not the number of octets) in one column of
the second-fastest changing dimension of the binary data array.
This information must be in the MIME header for proper operation
of some of the decompression algorithms.
The optional "X-Binary-Size-Third-Dimension" specifies the
number of sections for the third-fastest changing dimension of
the binary data array.
The optional "X-Binary-Size-Padding" specifies the size in
octets of an optional padding after the binary array data and
before the closing flags for a binary section.