Property List/Binary

The binary property list format is one of several formats that have been use for plist files, the standard means for applications under operating systems descended from NeXTSTEP (including current Apple OSs) to store configuration data. The binary format was introduced more recently than the other formats, but is now the default for Apple systems.

= Format description =

Available types
The following types are available in a binary plist:
 * 1) singleton (boolean, null, fill)
 * 2) integer
 * 3) float
 * 4) date
 * 5) binary data
 * 6) single byte string
 * 7) double byte string
 * 8) UID
 * 9) array
 * 10) dictionary

File Sections
A binary plist file has four sections:
 * 1) The first 8 bytes are an identifier, equal to 'bplistXX' where XX is two digits. 'bplist00' and 'bplist01' are known, I don't know the differences.
 * 2) Second is all of the elements in the plist, encoded and concatenated.
 * 3) Third is the concatenation of the offsets of all of the elements in the plist, each offset given as an unsigned integer in a fixed number of bytes. An object in the plist has a reference number that is based on the 0-based indexing of this table, e.g. object number 0 is the object at the offset given first in this table.
 * 4) The final 32 bytes, the "trailer" section.

Trailer
The final 32 bytes of a binary plist have the following format:
 * 1) 6 bytes of \x00 padding
 * 2) a 1 byte integer which is the number of bytes for an offset value. Valid values are 1, 2, 3, or 4. Offset values are encoded as unsigned, big endian integers. Must be wide enough to encode the offset of the offset table, not just the highest object offset.
 * 3) a 1 byte integer which is the number of bytes for an object reference number. Valid values are 1 or 2. Reference numbers are encoded as unsigned, big endian integers.
 * 4) 4 bytes of \x00 padding
 * 5) a 4 byte integer which is the number of objects in the plist
 * 6) 4 bytes of \x00 padding
 * 7) a 4 byte integer which is the reference number of the root object in the plist. This is usually zero.
 * 8) 4 bytes of \x00 padding
 * 9) a 4 byte integer which is the offset in the file of the start of the offset table, named above as the third element in a binary plist

Object encoding
The encoding of the available types are as follows:

The first four bits are an id number of the object type, according to the following mapping:
 * 0x0 singleton
 * 0x1 integer
 * 0x2 float
 * 0x3 date
 * 0x4 binary data
 * 0x5 single byte string
 * 0x6 double byte string
 * 0x8 UID
 * 0xa array
 * 0xd dictionary

The second four bits are the size of the object. If the value given is 15, this means that the true object size is greater than can be expressed in four bits. In this case, the next byte is the start of the true size, encoded like the integer objects, except the value is unsigned. This is not the case for 0x0 type objects however, see below for details. I will refer to this value as the object length.

The encoding for the remainder of the object varies by type. The "byte length" is the number of bytes used in encoding the object, not counting the encoding of the type id number and object length already discussed.

Singleton
Type number 0x0 groups together boolean values, a null value, and something called 'fill' by plutil.pl. I don't know what fill is for. The object length is actually the value of the object, and the byte length is always zero. A value of 0 means null, 8 means False, 9 means True, and 15 means fill. Any other value is invalid.

Integer
The byte length is equal to 2 to the power of the object length. Valid object lengths are 0, 1, 2, and 3 for 1, 2, 4, and 8 byte integers, respectively. The encoding is as a big-endian, signed integer in the appropriate number of bytes.

Float
The object length to byte length conversion is the same as for integers. The object length is 2 or 3, corresponding to a byte length of 4 or 8. The encoding is as a big-endian, single-precision or a double-precision float, accordingly.

Date
Dates are stored as a float with a value of seconds since the epoch of 1 January 2001, 0:00:00 GMT. Encoding is the same as the encoding for floats, except that the object length is always 3, for a byte length of 8.

Binary Data
The byte length is the object length, and any value is valid. The bytes are not interpreted.

Single Byte String
The byte length is the object length, and any value is valid. Encoding is ASCII.

Double Byte String
The byte length is twice the object length, and any value is valid. The encoding is utf-16 (big endian).

UID
Encoding is as integers, except values are unsigned. These are used extensively in files written using NSKeyedArchiver, a serializer for Objective-C objects. The UID value seems to be the index of a value within the $objects list inside such a plist.

Array
The byte length is the object length times the number of bytes per object reference for this plist file, i.e. either one or two times the object length. Any object length is valid. The encoding is the concatenation of object reference numbers as unsigned, big-endian integers each encoded in the number of bytes per object reference for this plist file.

Dictionary
The byte length is twice the object length times the number of bytes per object reference. The encoding is as the concatenation of two encoded arrays, the first of keys and the second of values. The first value in the list of keys corresponds to the first value in the list of values, and so on.

Final Notes
When writing a binary plist file, any values that repeat within the file should be encoded only once and that single object referenced where ever that value repeats.

Software

 * plist Editor for Windows: edits both XML and binary versions
 * Apple have documented the binary plist format
 * Python module for processing plists
 * binplist module
 * plutil.pl (Perl)
 * Convert .iChat files to RTF/TXT
 * Converting Binary Plists

Other links

 * Page giving documentation on this format... which links to this File Formats Wiki page as a reference, so I guess things are circular now.
 * Understanding Apple’s binary property list format