Work in progress
The content of this page was not yet updated for Godot
4.1
and may be outdated. If you know how to improve this page or you can confirm
that it's up to date, feel free to open a pull request.
Binary serialization API¶
Introduction¶
Godot has a serialization API based on Variant. It's used for
converting data types to an array of bytes efficiently. This API is exposed
via the global bytes_to_var()
and var_to_bytes() functions,
but it is also used in the get_var
and store_var
methods of
FileAccess as well as the packet APIs for PacketPeer.
This format is not used for binary scenes and resources.
Full Objects vs Object instance IDs¶
If a variable is serialized with full_objects = true
, then any Objects
contained in the variable will be serialized and included in the result. This
is recursive.
If full_objects = false
, then only the instance IDs will be serialized for
any Objects contained in the variable.
Packet specification¶
The packet is designed to be always padded to 4 bytes. All values are little-endian-encoded. All packets have a 4-byte header representing an integer, specifying the type of data.
The lowest value two bytes are used to determine the type, while the highest value two bytes contain flags:
base_type = val & 0xFFFF;
flags = val >> 16;
Type |
Value |
---|---|
0 |
null |
1 |
bool |
2 |
integer |
3 |
float |
4 |
string |
5 |
vector2 |
6 |
rect2 |
7 |
vector3 |
8 |
transform2d |
9 |
plane |
10 |
quaternion |
11 |
aabb |
12 |
basis |
13 |
transform3d |
14 |
color |
15 |
node path |
16 |
rid |
17 |
object |
18 |
dictionary |
19 |
array |
20 |
raw array |
21 |
int32 array |
22 |
int64 array |
23 |
float32 array |
24 |
float64 array |
25 |
string array |
26 |
vector2 array |
27 |
vector3 array |
28 |
color array |
29 |
max |
Following this is the actual packet contents, which varies for each type of
packet. Note that this assumes Godot is compiled with single-precision floats,
which is the default. If Godot was compiled with double-precision floats, the
length of "Float" fields within data structures should be 8, and the offset
should be (offset - 4) * 2 + 4
. The "float" type itself always uses double
precision.
0: null¶
1: bool¶
Offset |
Len |
Type |
Description |
---|---|---|---|
4 |
4 |
Integer |
0 for False, 1 for True |
2: int¶
If no flags are set (flags == 0), the integer is sent as a 32 bit integer: