Message structure

A message can be of either map or list type. In a map each field has a name, in a list the members do not have names, but the order should be preserved.

The field types are:

||Name||ID||Description ||Map||1||Sub message of type map ||S64||2||Signed 64bit integer ||Str||3||UTF-8 encoded string ||Bin||4||Binary blob ||List||5||Sub message of type list

All in all the message structure is quite similar to JSON but most notably; no boolean nor null type exist and HTSMSG supports binary objects.

HTSMSG binary format

The binary format is designed to for back-to-back transmission of messages over a network (TCP) connection.

The root message must always be of type map.

Root body

||Length||4 byte integer||Total length of message (not including this length field itself)
 ||Body||HTSMSG-Field * N||Fields in the root body


||Type||1 byte integer||Type of field (see field type IDs above)
 ||Namelength||1 byte integer||Length of name of field. If a field is part of a list message this must be 0
 ||Datalength||4 byte integer||Length of field data
 ||Name||N bytes||Field name, length as specified by Namelength
 ||Data||N bytes||Field payload, for details see below

Field encoding for type: map and list

The data is repeated HTSMSG-Fields exactly as the root body. Note the subtle difference in that for the root-message the length includes the 4 bytes of length field itself whereas in the field encoding the length written includes just the actual payload.

Field encoding for type: s64

Integers are encoded using a very simple variable length encoding. All leading bytes that are 0 is discarded. So to encode the value 100, datalength should be 1 and the data itself should be just one byte [0x64]. To encode 1337; datalength=2, data=[0x39 0x05].

Note that there is no sign extension in this encoding scheme so if you need to encode -1 you need to set datalength=8 and data = [0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff]. This can certainly be thought of as a bug, but it is the way it is.

Field encoding for type: str

Datalength should be the length of the string (NOT including the null terminating character). Thus the null terminator should not be present in data either.

Field encoding for type: bin

Datalength should be the length of the binary object. Data is the binary object itself.

Updated by Andreas Smas over 11 years ago · 5 revisions