[manual index][section index]


ubfa - universal binary format for data transport


UBF(A) is the data transport encoding for Armstrong's Universal Binary Format. It provides four primitive types: atoms (symbolic constants), integers, strings, and binary data. There are two compound types: fixed-length tuples and variable-length lists. Ubfa(2) provides basic support in Limbo for reading and writing streams of UBF(A)-encoded data.

The input syntax is defined by the following rules:

input	::=	item* '$'
item	::=	integer | atom | string | binary | tuple | list | store | push | comment | tag

integer	::=	'-'?[0-9]+
atom	::=	"'" ([^\'] | '\\' | "\'")* "'"
string	::=	'"' ([^\"] | '\\' | '\"')* '"'
binary	::=	'~' byte* '~'      # preceded by integer byte count

tuple	::=	'{' item* '}'
list	::=	'#' (item '&')*

store	::=	'>' reg
push	::=	reg
reg	::=	[^-%"~'`{}#& \n\r\t,0-9]
comment	::=	'%' ([^\%] | '\\' | '\%')* '%'
tag	::=	'`' ([^\`] | '\\' | '\`')* '`'

White space is any sequence of blank, tab, newline or carriage-return characters, and can appear before or after any instance of item in the grammar.

The input data is interpreted by a simple virtual machine. The machine contains a stack of values of primitive and compound types, and a set of registers also containing values of those types. White space and comments are ignored. Primitive integer, atom and string values are pushed onto the stack as they are recognised. Certain input bytes outside any value act as operators:

Note the current stack depth.
Pop stack values to restore the most recently noted stack depth. Push a single value representing a tuple of those items; the left-most value in the tuple is the last one popped (the first in the original input stream).
Pop an integer value n from the stack. Read n bytes from the input stream and push a value onto the stack that represents them. The next byte must be the character ~, which is discarded.
Push a value representing an empty list onto the stack.
Pop a value v. Pop another value l, which must represent a list. Push a value that represents the list v::l. (Note that the items in a list therefore appear in reverse order in the input stream.)
Pop the top value from the stack and store it in a register labelled by the byte reg.
Push the value of register reg (which must be non-null) onto the stack.
Associate the tag string with the value on top of the stack. The ubfa(2) implementation does so by replacing it by a special Tag tuple.
End-of-input: there must be exactly one value on the stack, which is the result.

Applications using UBF(A) typically take turns to exchange input values on a communication channel.


ubfa(2), json(6), sexprs(6)
J L Armstrong, ``Getting Erlang to talk to the outside world'', ACM SIGPLAN Erlang workshop 2002 , Pittsburg, PA USA
UBF web page, http://www.sics.se/~joe/ubf/

UBFA(6 ) Rev:  Thu Feb 15 14:43:48 GMT 2007