This repository is intended to house packet and system message information for TERA. It is intended to be platform agnostic, with details on the file formats explained below.
The currently known open source parsers are:
- JavaScript (Node): tera-data-parser
The map
directory contains files which map unique identifiers (names) to their
numeric values. Currently, this includes:
protocol.map
, which links opcodes to "packet" namessysmsg.map
, which links system message IDs to readable names
These generally come from the client binary and should not be built or modified by hand unless you know what you're doing.
Methods and instructions on dumping opcodes and sysmsgs can be found from:
TERA's network data follows a custom protocol. It is convenient to describe the
order and meaning of each element in a "packet", which is done through a .def
file under the protocol
directory, and named after the opcode it belongs to.
Each line in the .def
must consist of the following, in order:
- An optional series of
-
for array and object definitions. These may be separated by spaces. To nest arrays or objects, just add another-
to the front. - A field type. Valid types listed below.
- At least one space.
- The name of the field.
A #
and anything after it on the line are treated as comments and will be
ignored when parsing.
The formal grammar for this syntax is in the appendix at the end of this document.
The following simple field types are supported:
bool
: A single byte that equalstrue
for any non-zero value (and optionally warns for any value above 1).byte
float
: A four-byte floating-point number.int16
int32
int64
uint16
uint32
uint64
There is one type that is not directly represented by the raw data and instead serves organizational purposes:
object
: Any fields under this one should be collected into some sort of encapsulating object. For instance, x/y/z vectors can be under anobject
called "location".
There are also a few variable-length fields:
array
: Almost likeobject
, except there can be 0 or more that should be collected into an array.bytes
: A series ofbyte
data.string
: String data, encoded as null-terminated UTF-16LE (in other words, a series ofuint16
where the final value is 0).
Each of the variable-length fields has accompanying metadata. These fields are determined implicitly (see the section below for the generation rules), but in some cases, particularly with legacy definitions, the following metatypes can be used. These types are deprecated, and use of them will disable implicit handling of these types:
count
: Acts asuint16
. Dictates the length of anarray
orbytes
field of the same name.offset
: Acts asuint16
. Indicates the byte offset from the beginning of the message for anarray
,bytes
, orstring
field of the same name.
More details on the original message format are below, while details on your language's or library's implementation of these types should be described in your library's documentation.
TERA's networking encodes all data in little-endian.
There are a few fields which are implied because they are never omitted. Every packet begins with two fields:
uint16 length
, which describes the byte length of the message, including this header.uint16 opcode
, which describes which kind of message this is. By looking up which name has this number in the mapping, you will know what the message is called.
Following these two fields are the metatypes for all variable length fields. To
produce the list, iterate through all variable length fields in order, and
depending on the type, add a count
and/or offset
field for it:
array
: Addcount
thenoffset
.bytes
: Addoffset
thencount
.string
: Addoffset
only.
Additionally, all array elements begin with two fields:
offset here
, which can be used to verify correctness. If this is the first element, theoffset
for the array should match this; otherwise, it should match thenext
for the previous element.offset next
, which points to the byte offset of the next element in the array, or zero if this is the final element.
In general, you will find count
and offset
fields at the beginning of a
message or array definition, and their corresponding fields at the end.
Given the definition:
int32 number
array list
- int16 value
A message will be parsed as if it were:
uint16 (length)
uint16 (opcode)
count list
offset list
int32 number
array list
- offset (here)
- offset (next)
- int16 value
Protocol definitions contain version information in the filename:
<NAME>.<VERSION>.def
where <NAME>
is an opcode name and <VERSION>
is an
integer starting from 1 and incrementing with each change.
When submitting changes, contributors must leave older versions untouched unless they are trivially backwards compatible. Instead, submit the changed definition as a new file with the version number incremented.
Whenever TERA receives a major patch, a tag will be added to the repository, and then the mappings will be updated and all outdated definition files will be deleted.
Feel free to submit pull requests! Please read the above notice in bold.
# Assume comments are removed (find the first "#" token and remove everything until the end of the line).
# Assume all lines have leading and trailing whitespace removed.
definition = { line };
line = [field] ["\r"] "\n";
whitespace = " " | "\t";
letter = "a".."z" | "A".."Z" | "_";
digit = "0".."9";
identifier = letter { letter | digit };
simple_type = "bool"
| "byte"
| "float"
| ["u"] "int" ("16" | "32" | "64");
complex_type = "array"
| "bytes"
| "object"
| "string";
meta_type = "count"
| "offset";
type = simple_type | complex_type | meta_type;
depth = { "-" { whitespace } };
field = depth type whitespace { whitespace } identifier;