YARP  2.3.70.1
Yet Another Robot Platform
Standard data representation format

The policy for representing data externally in YARP is chosen carefully to encourage both efficiency and the ability to read/write network data by hand.

We define the following:

  • A standard set of data-types
  • A standard binary format
  • A standard text format
  • A standard mapping between the two formats

The NetType data-types

We limit our basic data-types to ones which can be easily read, written, and distinguished from each other in text form. The types, and their arbitrary names, are:

  • NetInt: an integer
  • NetFloat: a floating point number
  • NetString: a string
  • NetBlob: An arbitrary sequence of bytes
  • NetVocab: A short identifier that can be interpreted either as an integer or a string
  • NetList: An arbitrary sequence of NetTypes (they do not need to all be the same type, and can include NetLists)

Some of the types have limitations on them because of how they are represented in binary format. The important point is that they can all be represented easily in both text and binary form.

Binary format

  • NetInt: Represented as a 32-bit little-endian signed integer.
  • NetFloat: Represented as a 64-bit C "double" floating-point number (IEEE 754 double-precision).
  • NetString: Represented as a null-terminated sequence of bytes.
  • NetList: The NetList representation is discussed separately.
  • NetVocab: Represented as a 32-bit little-endian signed integer. If interpreted as a string, with the lowest order 8-bits as the first character, and the highest order non-zero 8-bits as the last character.
  • NetBlob: Represented as a NetInt giving the number N of bytes, and then the sequence of N bytes.

The NetList representation is as follows. Every NetType is assigned a code:

  • NetInt.code is 1
  • NetFloat.code is 10
  • NetString.code is 4
  • NetBlob.code is 12
  • NetVocab.code is 9
  • NetList.code is 256 if it may contain elements of diverse types, or 256+T.code if it contains elements of a single type T.

Suppose a NetList is the list of NetTypes (e1 e2 e3 ... eN). If the elements are all of the same type T (other than NetList itself) then the list is represented as:

  • A NetInt containing the number NetList.code+T.code, followed by
  • A NetInt containing the number of elements, N, followed by
  • The representations of the N individual elements.

If the elements are not all of the same type, or are all NetLists in turn, then the list is represented as:

  • A NetInt containing the number NetList.code, followed by
  • A NetInt containing the number of elements, N, followed by
  • The representations of the N individual elements, where each element is preceded by a NetInt containing their type code.

Text format

  • NetInt: Represented in the text forms supported by C99 strtol, e.g. -15, 9991, 0xfa, ...
  • NetFloat: Represented in text forms supported by ANSI C strtod, except that a period must be present in order distinguish this type from NetInts, e.g. 10.57, .0
  • NetString: Represented as characters within double-quotes, with special characters escaped C-style. The quotes may be omitted for strings that contain only alphanumeric characters.
  • NetList: The NetList representation is a white-space separated list of elements. Nested lists are placed within parentheses.
  • NetVocab: Represented as a short string between square brackets, e.g. [get]. There should be at most 4 characters between the brackets.
  • NetBlob: Represented as a series of numbers between curly braces, e.g. {1 10 255 6 3}. Each number represents a byte.

Examples

The list of primes less than 20 would be represented as:

text: 2 3 5 7 11 13 17 19
binary: [4:257] [4:8] [4:2] [4:3] [4:5] [4:7] [4:11] [4:13] [4:17] [4:19]

where [b:n] represents the number n in b bytes.

A list of numbers followed by a list of strings might be:

text: (91 92 93) (this is a "good list")
binary: [4:256] [4:2] [4:257] [4:3] [4:91] [4:92] [4:93] [4:260] [4:4]
[4:5] [5:"this"] [4:3] [3:"is"] [4:2] [2:"a"] [4:10] [10:"good list"]

In practice

The yarp::os::Bottle class in YARP is a useful helper class for reading/writing the standard data format in binary and text mode.

We also use mappings from common non-network external data such as command line-options and configuration files.

Command-line mapping

Command-line options of the form:

--opt1 arga argb --opt2 argc --opt3

are interpreted to correspond to a list with the following elements in unspecified order (expressed here as text)

(opt1 arga argb) (opt2 argc) (opt3)

This mapping is implemented by yarp::os::Property::fromCommand().

int argc = 5;
char *argv[] = { "fake_program", "--size", "10", "20", "--name", "mr frog" };
p.fromCommand(argc, argv[]);
ConstString result = p.toString();
// result will be a permutation of (size 10 20) (name "mr frog")

Configuration-file mapping

Configuration files of the form:

[SECTION1]
opt1 arga argb
opt2 argc
[SECTION2]
joints 5
mins 0 0 0 0 10
maxs 100 100 50 100 20

are interpreted to correspond to a list with the following elements (expressed here as text)

(SECTION1 (opt1 arga argb) (opt2 argc))
(SECTION2 (joints 5) (mins 0 0 0 0 10) (maxs 100 100 50 100 20))

This mapping is implemented by yarp::os::Property::fromConfigFile().