BinaryWML
From Wesnoth
Contents |
Binary WML
When sending WML over the network, it is compressed. This is a description of the used compression scheme. (Note: this comes mainly from a comment in binary_wml.cpp, the rest is reverse engineered from the source code and in some small parts from watching the network data - so it may be slightly incorrect)
Assumptions
- most space is taken up by element names and attribute names
- there are relatively few element names and attribute names that are repeated many times
Therefore, data are never compressed, just the element and attribute names.
How it works
Binary WML is a byte code. Each byte really is a command:
- 0 Open element
- 1 Close element
- 2 Add new code
- 3 Attribute
- 4..255 Code
The format of the commands is:
- 0 3 <name> Open element "name"
- 0 <C> Open element code[C]
- 1 Close the current element
- 2 <name> Dynamically add a new code for "name"
- 3 <name> <data> Attribute "name" with value "data"
- <C> <data> Attribute code[C] with value "data"
where:
- <name> and <data> are any byte sequences ended by a 0 byte.
- <C> is a single byte with a value from 4 to 255
Binary data
To send binary data (e.g. the campaign server when sending files), the <data> must escape any 0 bytes as a 0 byte marks the end of the data. It is done in this way:
- 1 1 -> 0
- 1 2 -> 1
Example
Assume we have this WML to send:
[message]text=blah[/message] [message]text=bleh[/message]
Then we would construct the following byte stream:
<0> <2> m e s s a g e <0> <4> <2> t e x t <0> <5> b l a h <0> <1> <0> <4> <5> b l e h <0> <1>
Or another variant without compression:
<0> <3> m e s s a g e <0> <3> t e x t <0> b l a h <0> <1> <0> <3> m e s s a g e <0> <3> t e x t <0> b l e h <0> <1>