Skyrim Mod:Compiled Script File Format

Skyrim supports version 3.0 to 3.1 compiled script files.

The format appears to be stored in big-endian regardless of platform (Windows, PS3, and 360 observed as big-endian).

Consult the list of Conventions for a list of primitive types and how to parse them.

Header

Name	Type/Size	Info
magic	uint32	0xFA57C0DE (FASTCODE?)
majorVersion	uint8	3
minorVersion	uint8	1 (Dawnguard, Hearthfire and Dragonborn scripts are 2)
gameId	uint16	1 = Skyrim?
compilationTime	uint64	the compilation time
sourceFileName	wstring	Name of the source file this file was compiled from (.psc extension).
username	wstring	Username used to compile the script
machinename	wstring	Machine name used to compile the script
stringTable	String Table	StringTable to look up member names and other stuff from
debugInfo	Debug Info
userFlagCount	uint16
userFlags	User Flag[userFlagCount]
objectCount	uint16
objects	Object[objectCount]

Sections

String Table

Name	Type/Size	Info
count	uint16
strings	wstring[count]

Debug Info

Name	Type/Size	Info
hasDebugInfo	uint8	Flag, if zero then no debug info is present and the rest of the record is skipped
modificationTime	uint64	time_t
functionCount	uint16
functions	Debug Function[functionCount]

Debug Function

Name	Type/Size	Info
objectNameIndex	uint16	Index(base 0) into string table.
stateNameIndex	uint16	Index(base 0) into string table.
functionNameIndex	uint16	Index(base 0) into string table.
functionType	uint8	valid values 0-3
instructionCount	uint16
lineNumbers	uint16[instructionCount]	Maps instructions to their original lines in the source.

User Flag

Name	Type/Size	Info
nameIndex	uint16	Index(base 0) into string table.
flagIndex	uint8	Bit index

Object

Name	Type/Size	Info
nameIndex	uint16	Index(base 0) into string table.
size	uint32
data	Object Data [size-4]	size includes itself for some reason, hence size-4

Object Data

Name	Type/Size	Info
parentClassName	uint16	Index(base 0) into string table.
docstring	uint16	Index(base 0) into string table.
userFlags	uint32
autoStateName	uint16	Index(base 0) into string table.
numVariables	uint16
variables	Variable[numVariables]
numProperties	uint16
properties	Property[numProperties]
numStates	uint16
states	State[numStates]

Variable

Name	Type/Size	Info
name	uint16	Index(base 0) into string table.
typeName	uint16	Index(base 0) into string table.
userFlags	uint32
data	Variable Data	Default value

Variable Data

Name	Type/Size	Info
type	uint8	0 = null, 1 = identifier, 2 = string, 3 = integer, 4 = float, 5 = bool
data	uint16	Index(base 0) into string table, present for identifier and string types only
data	int32	present for integer types only
data	float32	present for float types only
data	uint8	present for bool types only

Property

Name	Type/Size	Info
name	uint16	Index(base 0) into string table
type	uint16	Index(base 0) into string table
docstring	uint16	Index(base 0) into string table
userFlags	uint32
flags	uint8	bitfield: 1(bit 1) = read, 2(bit 2) = write, 4(bit 3) = autovar. For example, Property in a source script contains only get() or is defined AutoReadOnly then the flags is 0x1, contains get() and set() then the flags is 0x3.
autoVarName	uint16	Index(base 0) into string table, present if (flags & 4) != 0
readHandler	Function	present if (flags & 5) == 1
writeHandler	Function	present if (flags & 6) == 2

State

Name	Type/Size	Info
name	uint16	Index(base 0) into string table, empty string for default state
numFunctions	uint16
functions	Named Function[numFunctions]

Named Function

Name	Type/Size	Info
functionName	uint16	Index(base 0) into string table
function	Function

Function

Function is nameless. Its name will be defined in Named Function at State.

Name	Type/Size	Info
returnType	uint16	Index(base 0) into string table
docstring	uint16	Index(base 0) into string table
userFlags	uint32
flags	uint8	bit 0 = global function bit 1 = native function (i.e., no code)
numParams	uint16
params	Variable Type[numParams]
numLocals	uint16
locals	Variable Type[numLocals]
numInstructions	uint16
instructions	Instruction[numInstructions]

Variable Type

Name	Type/Size	Info
name	uint16	Index(base 0) into string table
type	uint16	Index(base 0) into string table

Instruction

Name	Type/Size	Info
op	uint8	see Opcodes
arguments	Variable Data[changes depending on opcode]	Length is dependent on opcode, also varargs

Opcodes

Opcode	Name	Arguments	Description
00	nop	none	do nothing
01	iadd	SII	add two integers
02	fadd	SFF	add two floats
03	isub	SII	subtract two integers
04	fsub	SFF	subtract two floats
05	imul	SII	multiply two integers
06	fmul	SFF	multiply two floats
07	idiv	SII	divide two integers
08	fdiv	SFF	divide two floats
09	imod	SII	remainder of two integers
0A	not	SA	flip a bool, type conversion may occur?
0B	ineg	SI	negate an integer
0C	fneg	SF	negate a float
0D	assign	SA	store a variable
0E	cast	SA	type conversion?
0F	cmp_eq	SAA	set a bool to true if a == b
10	cmp_lt	SAA	set a bool to true if a < b
11	cmp_le	SAA	set a bool to true if a <= b
12	cmp_gt	SAA	set a bool to true if a > b
13	cmp_ge	SAA	set a bool to true if a >= b
14	jmp	L	relative unconditional branch
15	jmpt	AL	relative conditional branch if a bool is true
16	jmpf	AL	relative conditional branch if a bool is false
17	callmethod	NSS*
18	callparent	NS*
19	callstatic	NNS*
1A	return	A
1B	strcat	SQQ	concatenate two strings
1C	propget	NSS	retrieve an instance property
1D	propset	NSA	set an instance property
1E	array_create	Su	create an array of the specified size
1F	array_length	SS	get an array's length
20	array_getelement	SSI	get an element from an array
21	array_setelement	SIA	set an element to an array
22	array_findelement	SSII	find an element in an array. The 4th arg is the startIndex, default = 0
23	array_rfindelement	SSII	find an element in an array, starting from the end. The 4th arg is the startIndex, default = -1

Each character in the "arguments" string specifies an argument type. They're all read as Variable Data.

• I = integer, i.e. a Variable Data with type = 0x03 and data consisting of a signed int32.
• F = float, i.e. a Variable Data with type = 0x04 and data consisting of a float32.
• A = boolean, i.e. a Variable Data with type = 0x05 and data consisting of a uint8.
• S = string, i.e. a Variable Data with type = 0x01 or 0x02 and data consisting of a uint16 index to the String Table.
• N = identifier, i.e. a Variable Data with type = 0x01 and data consisting of a uint16 index to the String Table.
• * = variable-length arguments. These consist of a Variable Data integer which specifies the number of additional arguments, followed by that number of Variable Data groups.
• u = unsigned integer? Presumably a 0x03 Variable Data object with the data interpreted as a uint32.
• Q = S? It's not clear why Q is used here.
• L = a code label of some sort, probably a program counter absolute value or offset?

A tool for dumping compiled scripts w/ source may be found at [1].