bee - BitTorrent Serialization Format Encoder/Decoder
The bee package provides de- and encoder commands for data in bencoding (speak 'bee'), the serialization format for data and messages used by the BitTorrent application.
The package provides one encoder command for each of the basic forms, and two commands per container, one taking a proper tcl data structure to encode in the container, the other taking the same information as several arguments.
Returns the bee-encoding of the string.
Returns the bee-encoding of the integer number.
Takes zero or more bee-encoded values and returns the bee-encoding of their list.
Takes a list of bee-encoded values and returns the bee-encoding of the list.
Takes zero or more pairs of keys and values and returns the bee-encoding of the dictionary they form. The values are expected to be already bee-encoded, but the keys must not be. Their encoding will be done by the command itself.
Takes a dictionary list of string keys and bee-encoded values and returns the bee-encoding of the list. Note that the keys in the input must not be bee-encoded already. This will be done by the command itself.
The package provides two main decoder commands, one for decoding a string expected to contain a complete data structure, the other for the incremental decoding of bee-values arriving on a channel. The latter command is asynchronous and provides the completed decoded values to the user through a command callback.
Takes the bee-encoding in the string and returns one decoded value. In the case of this being a container all contained values are decoded recursively as well and the result is a properly nested tcl list and/or dictionary.
If the optional endvar is set then it is the name of a variable to store the index of the first character after the decoded value into. In other words, if the string contains more than one value then endvar can be used to obtain the position of the bee-value after the bee-value currently decoded. together with start, see below, it is possible to iterate over the string to extract all contained values.
The optional start index defaults to 0, i.e. the beginning of the string. It is the index of the first character of the bee-encoded value to extract.
Takes the same arguments as ::bee::decode and returns the same information in endvar. The result however is different. Instead of the tcl value contained in the string it returns a list describing the value with respect to type and location (indices for the first and last character of the bee-value). In case of a container the structure also contains the same information for all the embedded values.
Formally the results for the various types of bee-values are:
A list containing three elements:
The constant string string, denoting the type of the value.
An integer number greater than or equal to zero. This is the index of the first character of the bee-value in the input string.
An integer number greater than or equal to zero. This is the index of the last character of the bee-value in the input string.
Note that this information is present in the results for all four types of bee-values, with only the first element changing according to the type of the value.
The result is like for strings, except that the type element contains the constant string integer.
The result is like before, with two exceptions: One, the type element contains the constant string list. And two, the result actually contains four elements. The last element is new, and contains the index data as described here for all elements of the bee-list.
The result is like for strings, except that the type element contains the constant string dict. A fourth element is present as well, with a slightly different structure than for lists. The element is a dictionary mapping from the strings keys of the bee-dictionary to a list containing two elements. The first of them is the index information for the key, and the second element is the index information for the value the key maps to. This structure is the only which contains not only index data, but actual values from the bee-string. While the index information of the keys is unique enough, i.e. serviceable as keys, they are not easy to navigate when trying to find particular element. Using the actual keys makes this much easier.
The command creates a decoder for a series of bee-values arriving on the channel chan and returns its handle. This handle can be used to remove the decoder again. Setting up another bee decoder on chan while a bee decoder is still active will fail with an error message.
The command prefix cmdprefix specified by the required option -command is used to report extracted values and exceptional situations (error, and EOF on the channel). The callback will be executed at the global level of the interpreter, with two or three arguments. The exact call signatures are
The decoder has reached eof on the channel chan. No further invocations of the callback will be made after this. The channel has already been closed at the time of the call, and the token is not valid anymore as well.
The decoder encountered an error, which is not eof. For example a malformed bee-value. The message provides details about the error. The decoder token is in the same state as for eof, i.e. invalid. The channel however is kept open.
The decoder received and successfully decoded a bee-value. The format of the equivalent tcl value is the same as returned by ::bee::decode. The channel is still open and the decoder token is valid. This means that the callback is able to remove the decoder.
By default the decoder assumes that the remainder of the data in the channel consists only of bee-values, and reads as much as possible per event, without regard for boundaries between bee-values. This means that if the the input contains non-bee data after a series of bee-value the beginning of that data may be lost because it was already read by the decoder, but not processed.
The -exact was made for this situation. When specified the decoder will take care to not read any characters behind the currently processed bee-value, so that any non-bee data is kept in the channel for further processing after removal of the decoder.
If this option is specified its value is assumed to be the beginning of the bee-value and used to initialize the internal decoder buffer. This feature is required if the creator of the decoder used data from the channel to determine if it should create the decoder or not. Without the option this data would be lost to the decoding.
This command cancels the decoder set up by ::bee::decodeChannel and represented by the handle token.
This command appends the string to the internal decoder buffer. It is the runtime equivalent of the option -prefix of ::bee::decodeChannel. Use it to push data back into the decoder when the value callback used data from the channel to determine if it should decode another bee-value or not.
Data in the bee serialization format is constructed from two basic forms, and two container forms. The basic forms are strings and integer numbers, and the containers are lists and dictionaries.
A string S of length L is encoded by the string "L:S", where the length is written out in textual form.
An integer number N is encoded by the string "iNe".
A list of the values v1 to vn is encoded by the string "lBV1...BVne" where "BVi" is the bee-encoding of the value "vi".
A dictionary mapping the string key ki to the value vi, for i in 1 ... n is encoded by the string "dBKiBVi...e" for i in 1 ... n, where "BKi" is the bee-encoding of the key string "ki". and "BVi" is the bee-encoding of the value "vi".
Note: The bee-encoding does not retain the order of the keys in the input, but stores in a sorted order. The sorting is done for the "raw strings".
Note that the type of each encoded item can be determined immediately from the first character of its representation:
By wrapping an integer number into i...e the format makes sure that they are different from strings, which all begin with a digit.
This document, and the package it describes, will undoubtedly contain bugs and other problems. Please report such in the category bee of the Tcllib Trackers. Please also report any ideas for enhancements you may have for either package and/or documentation.
Copyright © 2004 Andreas Kupries <firstname.lastname@example.org>