SCM language

From GTAMods Wiki
Revision as of 17:42, 27 September 2020 by Squ1dd13 (talk | contribs) ("Command ID" --> "Opcode" in line with most references. Reworded more formally.)
Jump to navigation Jump to search
Announcement: This article is currently being reworked. Check out the progresses by visiting this link from time to time. Contact Wesser for any suggestion.
This section deals with the native SCM syntax of GTA 3 series, nothing other than III, VC, SA, LCS and VCS.
It may contain non-standard SCM definitions as R* hasn't published enough documentation about it yet.

On the occasion of the GTAIII's Tenth Anniversary, after a long period of darkness where we fell about the real SCM syntax, R* finally treated us by attaching part of its own original source code into the GTAIII Anniversary game, available for iOS and Android devices. As far back as 2001, a snip of some debugging scripts has been already provided with main.sc and debug.sc files. However, many secrets are unrevealed yet, thus some things cannot be documented fully and so they can be only guessed. The SCM format abbreviation is one of countless proofs of this inconvenience, which may stand for Script Multifile. Other doubts come with source files, whose SC extension appears to be very close to Mission SCript. Although we have enough information to suppose the currently unknown mysteries of the used language, we still have no safe clue about which was its original denomination. Furthermore, it is a matter of fact that R* developers have been left untouched the miss2 executable name of the GTA 3 series compiler since the chapter 2. In this connection, we could imagine the new language is a variant or an evolution of the GTA2script. The ancient documentation by DMA (at present Rockstar North), mentions GTA2script as a successor to GBHscript - a language used in GTA 1 (GBH was a planned name for GTA). Therefore, the language used in GTA 3 series should've been called GTA3script. It was influenced by both BASIC and C programming languages.


Preliminary remarks

This article makes use of formatted codes to improve the reading comprehension. Note that:

  • Square brackets mean everything inside may be omitted;
  • Curly brackets denote the presence of useful codes but not necessarily needed;
  • Vertical bars divide what can be chosen alternatively.

Fundamentals

Comments

Comments are notes left in the code to aid readability or to explain a part of code that may be confusing otherwise. Comments are ignored by the compiler. R*'s compiler supports C++ style comments – single-line comments and multiline comments.

A single line comment only affects a single line. They are prefixed with // (two slashes). Anything between these slashes and the end of the line is a comment. Single line comments may appear at the end of a line of code as well as on their own line. 

[...] // This comment is at the end of a line.

// This comment is on its own line.

Multiline comments appear between /* and */. They do not actually have to span multiple lines, and anything outside of the opening and closing tokens is not ignored (i.e. these comments may appear with code on either side on the same line). 

[...]
/*
  This comment spans
  multiple lines.
*/
[...]

[...] /* This comment is within a line */ [...] /* As is this one */ [...]

R*'s compiler allows multiline comments to be nested.

Highlighters

Highlighters behaviour sounds trivial, that's to say they simply highlight one or more arguments per command within round brackets, individually or together. In GTA III, they appear to be used only for SETUP_ZONE_PED_INFO (in a various order) and GXT keys: 

SETUP_ZONE_PED_INFO FISHFAC DAY (0) 0 0 0 (0 0 0 0) 0
PRINT_BIG (T4X4_1) 5000 2
Limits
Opening and closing round brakets are treated as blank spaces;
An optional comma can be used as well to distinguish each argument, processed as a space.

Scopes

Scopes are delimited by curly brackets (or multiline brackets) which act like a local variable scope. Essentially, they enclose the code where local variables are used, including timers. They can be opened and closed many times in a script: 

{
    [...]
}
Limits
Scopes cannot be nested;
Opening and closing curly brakets are real commands.

Labels

A label is a sequence of characters which identify the reference of a location of the source code useful for gotoes. It can be accessed by any part of the source code. To define a label just append : (colon) to its name: 

[...]

{lblname}:
[...]

At the compiling time, they are automatically converted into an offset.

Variables

A variable is a memory address that is given a name. Variables are used to reference values that are stored in the game's memory and that can change (hence the name "variable"). Instructions may read the variable's memory or write to it. They act in the same way as pointers in other programming languages.

There are several words related to variables that are important to know:

  • The value of the variable is the data that lies at the variable's memory address.
  • The type is what kind of value the variable stores.
  • The scope is the region of code in which the variable may be used (and in which it is declared). There are two types:
    • The global scope is shared between all game scripts. Global variables have the same value everywhere. They are declared with the VAR prefix.
    • The local scope is a region of code within a script. Locals are only visible to other code inside the same pair of curly braces. They are declared with the LVAR prefix.
  • A timer is a unique local variable whose value rises automatically. They start counting when the script starts executing, and continue counting indefinitely. There are two such timers, TIMERA and TIMERB which are always defined, so do not need to be declared.

Data types

Among the available data types, some are equivalent to those of the most known programming languages. Their length is up to 4, 8 and 16 bytes. Each type is appended as a suffix in the variable declaration.

For an in-depth description of different data types, see the SCM instruction page.

LABEL

The LABEL type handles variable-length strings. It can refer to either a label name or a file name.

Notes
While inside a script file, R* compiler treats it unambiguously as a label;
R* compiler allocates 32 bytes per label.
Limit
LABEL variables aren't available.

INT

The INT type handles 32-bit signed integers. It is also used to store values with less bytes, such as a bool, a char and a short int.

FLOAT

The FLOAT type handles 32-bit floating-points. As it normally does, decimal precision of a float is usually stuck to 6-7 digits beyond which it may get lost.

Notes
R* compiler also accepts f and F suffixes for the immediate value.

TEXT_LABEL

The TEXT_LABEL type handles 8-byte strings. Generally, a string is an array of 1-byte characters. It requires 7 characters plus the null-terminator (a blank byte meaning the end of the string). It is used to hold GXT keys (those of town zones, interiors, help textes or dialogue subtitles) script names or any short string. Literal only TEXT_LABEL* strings are probably marked by single quotation marks to distinguish them from variable and constant identifiers: 

PRINT_BIG 'GXT_KEY'
Limit
TEXT_LABEL variables are supported in San Andreas and Vice City Stories.

TEXT_LABEL16

The TEXT_LABEL16 type handles 16-byte strings. Like the previous, this type holds 15 characters plus the null-terminator. It is used to store model and texture names of player clothes, animation names or any long string.

Limit
TEXT_LABEL16 variables and values are supported only in San Andreas.

TEXT_LABEL32

The TEXT_LABEL32 type handles 32-byte strings or larger, depending on how many continuous parameters of the same type there are, each of which occupies 32 bytes. It can hold up to 127 characters plus the null-terminator, after which another TEXT_LABEL32 argument may begin. Strings of such type must be put within double quotation marks: 

SAVE_STRING_TO_DEBUG_FILE "32B-128B TEXT"
Limits
TEXT_LABEL32 values are supported since Vice City;
TEXT_LABEL32 variables aren't available.

TEXT

The TEXT type handles N-byte strings. It holds N characters plus the null-terminator. Strings of this type mustn't exceed 255 characters (including the null-byte).

Limits
TEXT values are supported since San Andreas;
TEXT variables aren't available.

CONST (pseudo)

The CONST type handles 32-bit signed integers. It is used only to assign and compare constants to INT variables regarding model identifiers, task statuses, ped or audio events and such. It is a pseudo type of INT.

Limit
The assignment and comparison of CONST values are supported since Vice City.

MULTI (pseudo)

The MULTI type handles a group of few data types acceptable per argument. It is used only for commands featuring optional arguments, those whose type is unpredictable before the compilation. It can be a pseudo type of INT, FLOAT and TEXT_LABEL.

Notes
INT and FLOAT can be used interchangeably in GTA III, Vice City, San Andreas, Liberty City Stories and Vice City Stories;
Vice City Stories also admits TEXT_LABEL arguments.

Declaration

Defining a variable means assigning a token string to a memory cell at the compiling time. Variables must be declared in the following manner: 

VAR_* {varname0}[,] [... {varnameN}]
LVAR_* {varname0}[,] [... {varnameN}]

As mentioned in the sections above, local variables have to be put within curly brackets: 

{
    VAR_* {varname0}[,] [... {varnameN}]
    LVAR_* {varname0}[,] [... {varnameN}]

    [...]
}

Inline variable declaration is allowed, you just have to separate them by spaces or tabulations. Adding a preceding comma before these characters is optional.

Limits
Whereas the variable buffer is limited, you can declare a certain amount of globals and locals. INT and FLOAT types take 1 variable, while TEXT_LABEL and TEXT_LABEL16 types occupy respectively 2 and 4 variables to store their data (have a look here for further details);
Vice City Stories isn't affected by what said above;
Global and local variable names must not collide.

Arrays

A array is a collection of variables having the same type which can be accessed by an index, an INT lesser than or equal to the size specified, enclosed by square brackets: 

{
    VAR_* {varname0}{[arrsize0]}[,] [... {varnameN}{[arrsizeN]}]
    LVAR_* {varname0}{[arrsize0]}[,] [... {varnameN}{[arrsizeN]}]

    [...]
}
Limits
The usage of arrays is allowed since Vice City;
Variable indices are quite buggy in Vice City and therefore unrecommended, but they are fully supported since San Andreas;
The aforesaid indices are one-based, possibly zero-based since San Andreas;
Multidimensional arrays are not supported.

Handles

A handle is an univocal identifier assigned to a game entity. It is given by the following statement: 

short nHandle = (iEntityIndexInPool << 8) | ucEntityFlag;
Note
R* compiler won't let you assign different entity types to the same variable or using a variable which hasn't been passed to any command designated to the creation of an entity.

Operators

In general, an operator is a token string that represents a math calculation or an operation of any other kind, in order to make the code understanding clearer at a glance.

Arithmetic

Arithmetic operators compute some of the most common algebric calculations between either a variable and a value or two variables. As well as in some programming language happens, CONST, TEXT_LABEL and TEXT_LABEL16 types are free from these operators, except for the basic assignment (see also Operators composition):

Operator Name Syntax Description
= Assignment expr0 = expr1 Store expr1 to expr0
+ Addition expr0 + expr1 Add expr1 to expr0
- Subtraction expr0 - expr1 Subtract expr1 from expr0
* Multiplication expr0 * expr1 Multiply expr0 by expr1
/ Division expr0 / expr1 Divide expr0 by expr1
+@ Timed addition expr0 +@ expr1 Multiply expr2 by delta time and add the result to expr1
-@ Timed subtraction expr0 -@ expr1 Multiply expr2 by delta time and subtract the result from expr1
++ Increment Pre[*] ++ expr0 Increment expr0 by 1 and store the result to expr0
Post expr0 ++
-- Decrement Pre[*] -- expr0 Decrement expr0 by 1 and store the result to expr0
Post expr0 --
Note
^ Pre and post increments have no difference unlike what you would expect.

Yet, you can put the assignment and algebric operators together inline as follows:

Operators Name Syntax Description
= + Addition and assignment expr0 = expr1[*] + expr2 Add expr2 to expr1 and store the result to expr0
= - Subtraction and assignment expr0 = expr1[*] - expr2 Subtract expr2 from expr1 and store the result to expr0
= * Multiplication and assignment expr0 = expr1[*] * expr2 Multiply expr1 by expr2 and store the result to expr0
= / Division and assignment expr0 = expr1[*] / expr2 Divide expr1 by expr2 and store the result to expr0
= +@ Timed addition and assignment expr0 = expr1[*] +@ expr2 Multiply expr2 by delta time, add the result to expr1 and store everything to expr0
= -@ Timed subtraction and assignment expr0 = expr1[*] -@ expr2 Multiply expr2 by delta time, subtract the result from expr1 and store everything to expr0
Note
^ expr1 can represent expr0 too.
Limit
Multiple algebric operators per line are not allowed.

Compound assignment

Compound assignment operators store values or variable content to other variables having a particular type afterwards the computation of an arithmetic operation, to squeeze the code and clear it up from granted repetitions:

Operator Name Syntax Description
+= Addition assignment expr0 += expr1 Add expr1 to expr0 and store the result to expr0
-= Subtraction assignment expr0 -= expr1 Subtract expr1 from expr0 and store the result to expr0
*= Multiplication assignment expr0 *= expr1 Multiply expr0 by expr1 and store the result to expr0
/= Division assignment expr0 /= expr1 Divide expr0 by expr1 and store the result to expr0
+=@ Timed addition assignment expr0 +=@ expr1 Multiply expr1 by delta time, add the result to expr0 and store everything to expr0
-=@ Timed subtraction assignment expr0 -=@ expr1 Multiply expr1 by delta time, subtract the result from expr0 and store everything to expr0

Uncompounded assignment

Uncompounded assignment operators are those on their own, or rather they are neither derivable nor decomposable similarly as those compounds:

Operator Name Syntax Description
=# Cast assignment expr0 =# expr1 Cast expr1 to any other type and store the result to expr0
Limit
Supported conversions are FLOAT to INT and INT to FLOAT.

Logical

Logical operators influence the way conditions are evalueted and enable to test more of them at a time. More than anything, they are built-in statements:

Operator Name Syntax Description
NOT Logical negation IF NOT condition0 Test if condition0 is false
AND Logical conjunction IF condition0
AND condition8 		Test if both condition0 and conditionN are true
OR Logical disjunction IF condition0
OR condition8 		Test if either condition0 or conditionN is true

Comparison

Comparison operators test the truth or falsity of the relation between either a variable and a value, a value and a variable or two variables:

Operator Name Syntax Description
= Equal to IF expr0 = expr1 Test if expr0 and expr1 are equal
> Greater than IF expr0 > expr1 Test if expr0 is greater than expr1
<[*] Lesser than IF expr0 < expr1 Test if expr0 is lesser than expr1
>= Greater than or equal to IF expr0 >= expr1 Test if expr0 is greater than or equal to expr1
<=[*] Lesser than or equal to IF expr0 <= expr1 Test if expr0 is lesser than or equal to expr1
Note
^ As a result of a critical bug, R* compiler mistakenly applies the operator inversion.

Commands

A command is a symbolic name associated to an identifier which executes a portion of code that specifies the operation to be performed by passing zero or more arguments. An argument is in turn some data given as input to a command. Normally, commands have a defined amount of arguments and those not, such as START_NEW_SCRIPT, can pass as many arguments as the available local variables are, except timers. Being a procedure, a command does not return values that can be assigned to a variable, even though the boolean flag is kept whenever it is used as a condition. It follows the common programming syntax adopted for function calls: 

{commandname} [{anyvalue0|varname0} ... {anyvalueN|varnameN}]

Overloads

An overload is an implementation of an operation that works with specific types. They are used almost exclusively for mathematical operations, for which they are necessary. An operation that is implemented multiple times in this way is sometimes referred to as an alternator.

Overloads are required because there are often multiple different combinations of argument types that one might wish to use for an operation. While 1 + 2 and 1 + 2.0 are equivalent to a human, they are not to a computer, so they must be implemented differently.

Each overload has the same arity as the other overloads of the same operation. They differ only the in combination and order of their parameters' types.

Overloaded operations

  • GTA III Vice City San Andreas Liberty City Stories Vice City Stories
    • SET (=)
    • CSET (=#)
    • ADD_THING_TO_THING (+=)
    • SUB_THING_FROM_THING (-=)
    • MULT_THING_BY_THING (*=)
    • DIV_THING_BY_THING (/=)
    • IS_THING_EQUAL_TO_THING (=)
    • IS_THING_NOT_EQUAL_TO_THING (NOT =)
    • IS_THING_GREATER_THAN_THING (>, <=)
    • IS_THING_GREATER_OR_EQUAL_TO_THING (>=, <)
    • ADD_THING_TO_THING_TIMED (+=@)
    • SUB_THING_FROM_THING_TIMED (-=@)
    • ABS
  • San Andreas
    • IS_EMPTY
    • IS_BIT_SET
    • SET_BIT
    • CLEAR_BIT
    • STRING_CAT

This section is incomplete. You can help by fixing and expanding it.

Hardcoded

Hardcoded commands are those which have unique characteristics and are handled internally:

List of hardcoded commands

  • GTA III Vice City San Andreas Liberty City Stories Vice City Stories:
    • GOTO
    • GOTO_IF_FALSE
    • TERMINATE_THIS_SCRIPT
    • START_NEW_SCRIPT
    • VAR_INT
    • VAR_FLOAT
    • LVAR_INT
    • LVAR_FLOAT
    • {
    • }
    • REPEAT
    • ENDREPEAT
    • IF
    • IFNOT
    • ELSE
    • ENDIF
    • WHILE
    • WHILENOT
    • ENDWHILE
    • ANDOR
    • LAUNCH_MISSION
    • SAVE_VAR_INT
    • SAVE_VAR_FLOAT
    • START_CUTSCENE[*]
    • PLAYER_MADE_PROGRESS
    • SET_PROGRESS_TOTAL[*]
    • REGISTER_MISSION_GIVEN[*]
    • REGISTER_MISSION_PASSED
    • SCRIPT_NAME
    • LOAD_AND_LAUNCH_MISSION
    • LOAD_AND_LAUNCH_MISSION_INTERNAL
    • SET_TOTAL_NUMBER_OF_MISSIONS[*]
    • VAR_TEXT_LABEL
    • LVAR_TEXT_LABEL
  • Vice City San Andreas Liberty City Stories Vice City Stories:
    • REGISTER_ODDJOB_MISSION_PASSED
  • GTA III Liberty City Stories Vice City Stories:
    • GOTO_IF_TRUE
    • GOSUB_FILE
  • GTA III Vice City:
    • CREATE_COLLECTABLE1
    • SET_COLLECTABLE1_TOTAL[*]
  • Vice City San Andreas:
    • LOAD_AND_LAUNCH_MISSION_EXCLUSIVE
  • Liberty City Stories Vice City Stories:
    • CALL
    • CALLNOT
  • San Andreas:
    • VAR_TEXT_LABEL16
    • LVAR_TEXT_LABEL16
    • SWITCH
    • ENDSWITCH
    • CASE
    • DEFAULT
    • BREAK
    • SWITCH_START
    • SWITCH_CONTINUED
  • Vice City Stories:
    • SAVE_VAR_TEXT_LABEL[*]
    • SET_COLLECTABLE2_TOTAL
Notes
^ The argument of these commands must be set respectively according to:
  • The sum of PLAYER_MADE_PROGRESS values;
  • The amount of REGISTER_MISSION_PASSED (those that don't have an immediate value are excluded) and REGISTER_ODDJOB_MISSION_PASSED;
  • The amount of CREATE_COLLECTABLE1.
If the argument of the listed commands differs from what expected, a 0-value must be passed;
^ This command was intended to be counted originally but its counter got deprecated.

This section is incomplete. You can help by fixing and expanding it.

WAIT

WAIT stops the execution of a script according to some milliseconds after which it will resume again. Indeed, it is absolutely necessary into infinite loops or those that may or may not break after more than one frame, such as the WHILE control flow. In this case, a INT equal to 0 is passed.

GOTO

GOTO jumps to the label of any location of the source code but conceptually it should never point off the current context. It is also used internally to build the control flows offered by the scripting language: 

// File: any.sc

goto_ref0:
GOTO goto_refN

// File: any.sc

goto_refN:
GOTO goto_ref0

ANDOR

ANDOR sets out the way the comparison among more conditions have to occur (see also Compare flag).

GOTO_IF_TRUE

GOTO_IF_TRUE operates in conjunction with ANDOR and jumps to a label if the returned boolean flag is true.

GOTO_IF_FALSE

Unlike GOTO_IF_TRUE, GOTO_IF_FALSE jumps to the desired label only if the comparison returns false.

SCRIPT_NAME

SCRIPT_NAME simply associates an unique name to the current working script.

Note
R* compiler doesn't enable you to associate a name previously used for another script.

SAVE_STRING_TO_DEBUG_FILE

SAVE_STRING_TO_DEBUG_FILE accepts an argument which can admit up to 127 characters plus the null-terminator. In the compiling process, the argument is skipped but its string is copied to a predefined 128-bytes buffer, compiled afterwards. Since Vice City, these are the seemingly predetermined bytes of a random empty string block which are actually the result of uninitialized data: 

00 00 41 00 09 2E 00 00 00 00 00 00 00 00 00 00 
09 2E 00 00 00 00 00 00 1C FB 12 00 D8 A8 41 00
00 00 41 00 09 2E 00 00 00 00 00 00 01 00 00 00 
09 2E 00 00 00 00 00 00 1C FB 12 00 D8 A8 41 00
00 00 41 00 09 2E 00 00 00 00 00 00 02 00 00 00 
09 2E 00 00 00 00 00 00 1C FB 12 00 D8 A8 41 00
00 00 41 00 09 2E 00 00 00 00 00 00 03 00 00 00 
09 2E 00 00 00 00 00 00 1C FB 12 00 D8 A8 41 00

The split of such bytes into 4 blocks of 32-bytes each is quite noticeable.

Constants

A constant is a symbolic name associated to a specific value. When compiling, their caption is converted to the assigned value. Since Vice City, names and identifiers of objects within OBJS and TOBJ blocks are loaded from every IDE file defined into gta_vc.dat, then those of vehicles and pedestrians within PEDS and CARS blocks are retrieved from default.ide. In San Andreas, they are listed into TXT files, whose name follows the Pascal Case (eg. AudioEvents.txt). These files respect the syntax below: 

{constname0} {constvalue0}

{constnameN} {constvalueN}

Constant names and values are divided by as many spaces or tabulations as you want. Constant lines are distinguished by two \n (new line) characters. The model names which aren't assigned to a constant are still valid (see also Identifiers). Keep in mind arguments of some commands having the CONST type accept only constant values of a single namespace.

Notes
Constants don't collide even though they belong to different namespaces;
In GTA III and Vice City, they are hardcoded as everything inside R* compiler;
In San Andreas, the subdivision of constant namespaces in files might be just a listing of hardcoded constants useful for developers. The same would apply to Liberty City Stories and Vice City Stories.

Formatting

Everything is case-insensitive, that means the uppercase and lowercase letters have no dissimilarities when taken. Usually, the source code is conform to the same formatting according to:

Compiling

Structure

The source code is split up into several SC files which comprehend main file, foreign gosubs, subscripts, mission scripts, and streamed scripts. These files can be included more times because they are actually processed once.

Main file

The main file is the most significant part of the whole source. It can include many script files and/or embedded gosubs, scripts or functions. Originally, it is characterized by the absence of the local scope. It must be put outside the directory, having the same name as the main script file, where all other foreign scripts must be: 

<directory>
| main
|  | gosub
|  |  |- gosub1.sc
|  |  \- gosubN.sc
|  | subscript
|  |  |- subscript1.sc
|  |  \- subscriptN.sc
|  | mission_guy
|  |  |- mission_guy1.sc
|  |  \- mission_guyN.sc
|  |- gosub.sc
|  |- subscript.sc
|  \- mission.sc
\- main.sc
Note
R* compiler will scan subfolders too.

Foreign gosubs

Foreign gosubs (also called subroutines) are main extension files. They are called using the GOSUB_FILE command which jumps to a specific label and executes some code that returns back to the place where it has been called with RETURN. You are able to specify the gosub label to start jumping at as well: 

// File: main.sc

GOSUB_FILE gosub_ref foreign_gosub.sc

// File: foreign_gosub.sc

gosub_ref:
{
    [...]
}
RETURN
Limit
Foreign gosubs were introduced since GTA III. They were unused in Vice City and got removed in San Andreas, but then they were reimplemented in Liberty City Stories and Vice City Stories.

Gosubs

As mentioned, gosubs are also embedded in any script file. They follow almost the same rules, except they are called by GOSUB and can actually inehrit the local scope of the parent script: 

// File: any.sc

GOSUB gosub_ref

// File: any.sc

gosub_ref:
{
    [...]
}
RETURN
Note
R* compiler doesn't take care if the code within a scope jumps to a gosub inside which another scope is declared. It is strongly recommended to pay attention at this issue or you will fall down into an irreparable local variable mismatch.

Subscripts

Subscripts are code blocks which take part of a queue of other scripts and are allocated over the memory by LAUNCH_MISSION. They are denoted by the presence of MISSION_START at the very top of the mission file. As long as they aren't ended with MISSION_END, their execution never expires till the end of the game process. Each one works independently, even though they are able to share global variables: 

// File: main.sc

LAUNCH_MISSION subscript.sc

// File: subscript.sc

MISSION_START

[VAR_* {varname0}[,] [... {varnameN}]]

SCRIPT_NAME main

subscript_loop:
{
    [LVAR_* {varname0}[,] [... {varnameN}]]

    [...]
}
//GOTO subscript_loop
MISSION_END
Notes
MISSION_START is a special and fake directive that isn't assigned to any command. R* compiler will notify an error if it isn't placed at the first line of a subscript or a mission script;
MISSION_END is an alias of TERMINATE_THIS_SCRIPT.

Scripts

As for gosubs, scripts can be embedded everywhere in a script file. They are started by START_NEW_SCRIPT which has an undefined amount of arguments, whose type must match with those of each local variable of the starting script in order to be passed, else the compilation will interrupt. Unlike subscripts, they get terminated by TERMINATE_THIS_SCRIPT or TERMINATE_ALL_SCRIPTS_WITH_THIS_NAME (elsewhere in another script): 

// File: any.sc

START_NEW_SCRIPT script [{anyvalue0|varname0} ... {anyvalueN|varnameN}]

// File: any.sc

script:
{
    SCRIPT_NAME script

script_loop:

    [LVAR_* {varname0}[,] [... {varnameN}]]

    [...]

    //GOTO script_loop
    TERMINATE_THIS_SCRIPT
}
Notes
Scripts must have a local scope;
Script commands must be inserted within or after the local scope;
Since Vice City, the opening curly bracket must be put before the script label when more arguments are passed.

Functions

This section is incomplete. You can help by fixing and expanding it.

Mission scripts

Mission scripts are those subscripts which are responsible for the presence of a storyline in the game. When they are launched with LOAD_AND_LAUNCH_MISSION, the mission is loaded in the mission block, allocated over the memory and the script pointer is moved to the corresponding mission offset. Do not forget to begin a mission script with MISSION_START and end it with MISSION_END: 

// File: main.sc

LOAD_AND_LAUNCH_MISSION mission.sc

// File: mission.sc

MISSION_START

GOSUB mission_start

IF HAS_DEATHARREST_BEEN_EXECUTED
    GOSUB mission_failed
ENDIF

GOSUB mission_cleanup

MISSION_END

[VAR_* {varname0}[,] [... {varnameN}]]

mission_start:

REGISTER_MISSION_GIVEN
SCRIPT_NAME mission

// Variables initialization

{
    [LVAR_* {varname0}[,] [... {varnameN}]]

    [...]
}
GOTO mission_passed

mission_failed:
[...]
RETURN

mission_passed:
REGISTER_MISSION_PASSED mission
//PLAYER_MADE_PROGRESS 1
[...]
RETURN

// Mark everything as no longer needed

mission_cleanup:
//MISSION_HAS_FINISHED
[...]
RETURN

Some missions doesn't need to be executed multiple times because they may just initialize some global variables defined in the main script or launch the intro mission. For this matter, here comes the usage of LOAD_AND_LAUNCH_MISSION_EXCLUSIVE: 

// File: main.sc

LOAD_AND_LAUNCH_MISSION_EXCLUSIVE initial.sc
LOAD_AND_LAUNCH_MISSION_EXCLUSIVE intro.sc
Note
Exclusive missions are never launched in the source code. It's likely, it was an idea not came to the end successfully or rather they were useful for debugging purposes.
Limits
LOAD_AND_LAUNCH_MISSION_EXCLUSIVE is available only in Vice City and San Andreas;
Only 2 exclusive missions in Vice City and 3 in San Andreas are handled, plus they must be launched before any of the counterpart.

Streamed scripts

This section is incomplete. You can help by fixing and expanding it.

Control flow

As usual, the evolution of something implies its development over the years. Alongside, the control flows implementation has been distributed equally into every chapter. Their definitions are similar to those used in pseudocodes resulting in a raw source code. However, you are still able to build your own control flows: 

ANDOR {value}
    [NOT] {condition0}
    [[NOT] {condition8}]
GOTO_IF_FALSE ELSE
    {consequence}
    [GOTO ENDIF]
ELSE:
    [{alternative}
ENDIF:]
Note
It's likely, user-made control flows weren't intended to be usable because R* compiler cannot recognize an equal to rather than an assignment operator.

If

IF is the most common conditional statement. It executes one block of code if a boolean condition is true, and (optionally) another if the condition is false. The condition may be made of multiple conditions combined using the logical operators AND or OR, or it may just be a single condition. AND makes the final condition true only if all of the input conditions are true, whereas OR only requires one input condition to be true. The condition may be inverted with NOT, another logical operator. Syntax: 

IF [NOT] {condition0}
[AND|OR [NOT] {condition8}]
    {consequence}
[ELSE
    {alternative}]
ENDIF
Limitations
  • There is a maximum of 8 conditions.

If-not

IFNOT works in the same way as IF except that the condition is always inverted. 

IFNOT [NOT] {condition0}
[AND|OR [NOT] {condition8}]
    {consequence}
[ELSE
    {alternative}]
ENDIF
Limitations
  • IFNOT is only supported in games that have GOTO_IF_TRUE: GTA III, Liberty City Stories and Vice City Stories.

While

WHILE is similar to IF, but it continues executing the body until the condition is false. 

WHILE [NOT] {condition0}
[AND|OR [NOT] {condition8}]
    {consequence}
ENDWHILE

While-not

WHILENOT is to WHILE as IFNOT is to IF: it loops until the condition is true. 

WHILENOT [NOT] {condition0}
[AND|OR [NOT] {condition8}]
    {consequence}
ENDWHILE
Limitations
  • Like IFNOT, WHILENOT is only supported in games that have GOTO_IF_TRUE: GTA III, Liberty City Stories and Vice City Stories.

Repeat

REPEAT acts as a range-based loop (generally a for-loop in modern programming languages). It executes the body and increments a certain variable until it reaches a target value, at which point the loop stops. 

REPEAT {target} {varname}
    {consequence}
ENDREPEAT
Limitations
  • REPEAT is supported in games from Vice City;
  • The target value must be positive;
  • The body will be read at least once in any case.

Switch

A SWITCH statement is functionally equivalent to multiple nested IF...ELSE statements. It takes a single value and executes a block of code that is specific to that value (a CASE) until it finds a BREAK, which moves execution to the end of the construct. If no matching case is found, the DEFAULT clause is executed. 

SWITCH {varname}
    CASE {value0}
        {consequence}
        BREAK
    [CASE {valueN}
        {consequence}
        BREAK]
    [DEFAULT
        {alternative}
        BREAK]
ENDSWITCH
Limitations
  • It is supported from San Andreas;
  • CASE allows the use of INT and CONST values only;
  • In San Andreas, the CASE values should be sorted (R*'s compiler should do it implicitly);
  • Every CASE, including DEFAULT, must end with a BREAK.

Decompiling

Structure

For further information about the SCM file format, read this article. Take into account the compiling order of each SC file is main file » foreign gosubs » subscripts » mission scripts apart from the reading order of the commands used to include them. Streamed scripts are compiled individually into the script.img file. On the other hand, functions are compiled like gosubs.

Identifiers

Undefined constants of model identifiers, whose name refers to a DFF which is presumably archived into any of the IMGs, loaded by the game, are overwritten by a decrementing value in the order they get compiled. These model names are then put into the second segment of the SCM header. Those of mission scripts and streamed scripts respect the same rule except the fact they are turned into a zero-based growing identifier, while exclusive mission scripts are launched by a negative identifier resulting from the bits inversion (bitwise complement).

Offsets

An offset is a 32-bit signed integer which points to a location of the script file. Those within the main file, foreign gosubs and subscripts are absolute offsets that start from the beginning of the main script, while the ones inside mission scripts and streamed scripts are relative and negative offsets starting from their beginning. The offset is related to global variables as well, whose interval goes from 8 and ends to 65532 (0xFFFC), each one is aligned to the nearest 4 bytes.

Variables range

The following table shows the variables range of the local scope for each game version:

Context GTA III Vice City San Andreas Liberty City Stories Vice City Stories
Foreign gosub/Gosub 0-15 0-15 n/a 0-95 0-95
Subscript/Script 0-15 0-15 0-31 0-95 0-95
Mission script 0-15 0-15 0-1023 0-95 0-95
Streamed script n/a n/a 0-31 n/a n/a
Function n/a n/a n/a 0-95 0-95
Timer 16-17 16-17 32-33 t0-t1 t0-t1

Operators composition

As far as you wouldn't know, SCM's operators always take two operands to compute an operation. Their composition is listed below:

Operator/s Name Syntax Composition
++ Increment Pre ++ expr0 expr0 += 1
Post expr0 ++
-- Decrement Pre -- expr0 expr0 -= 1
Post expr0 --
= + Addition and assignment expr0 = expr1 + expr2 expr0 = expr1
expr0 += expr2
= - Subtraction and assignment expr0 = expr1 - expr2 expr0 = expr1
expr0 -= expr2
= * Multiplication and assignment expr0 = expr1 * expr2 expr0 = expr1
expr0 *= expr2
= / Division and assignment expr0 = expr1 / expr2 expr0 = expr1
expr0 /= expr2
= +@ Timed addition and assignment expr0 = expr1 +@ expr2 expr0 = expr1
expr0 +=@ expr2
= -@ Timed subtraction and assignment expr0 = expr1 -@ expr2 expr0 = expr1
expr0 -=@ expr2

Opcode

A command's opcode is a 16-bit signed integer identifier which the game uses to find the implementation for the command. The maximum number of commands possible is 32,767 (0x7FFF), because the most significant bit (0x8000) is the sign bit, which is only set when the command's return value is inverted (e.g. not x()).

There are no commands with negative opcodes; the game takes the absolute value (|opcode|) before finding the implementation, after noting whether it is inverted.

Command arguments

The limitation of the amount of arguments a variadic command can pass is game specific:

  • 16 for GTA III and Vice City;
  • 32 for San Andreas;
  • 96 for Liberty City Stories and Vice City Stories.

Managed commands

Here is the list of all managed commands and their relative specifications:

Legend:

  • Suffix:
    • O, OPTIONAL.

List:

Be aware, the argument data types of the commands below are just informative:

Notes
^ A special mission directive which is never compiled;
^ It is used to build the various control flows internally;
^ It is used to declare one or more variables;
^ It embeds a variable scope;
^ It has an undefined amount of arguments;
^ It is a likely definition of the standard command;
^ It should exist but its position is purely guessed;
^ Arguments amount varies when compiling.

This section is incomplete. You can help by fixing and expanding it.

Uncommon values

Arguments of some commands keep uncommon values which look familiar after encoding:

Command Arg.
ID 		Value Encoded
Command Arg.
ID		 Value Type
GTA III Vice City San Andreas Liberty City Stories Vice City Stories
GOTO 1 Generic label GOTO 1 Positive/Negative offset INT
GOTO_IF_FALSE 1 Generic label GOTO_IF_FALSE 1 Positive/Negative offset INT
GOSUB 1 Gosub label GOSUB 1 Positive/Negative offset INT
GOSUB_FILE 1 Gosub label GOSUB_FILE 1 Zero-based offset INT
2 Foreign gosub file 2
1 Gosub label 1 Positive/Negative offset
2 Foreign gosub label 2
START_NEW_SCRIPT 1 Script label START_NEW_SCRIPT 1 Positive/Negative offset INT
l Passed locals l Passed locals ANY_MULTI
LAUNCH_MISSION 1 Subscript file LAUNCH_MISSION 1 Zero-based offset INT
Subscript label Positive/Negative offset
LOAD_AND_LAUNCH_MISSION 1 Mission script file LOAD_AND_LAUNCH_MISSION_INTERNAL 1 Mission identifier INT
Mission script label LOAD_AND_LAUNCH_MISSION Positive/Negative offset
GTA III Liberty City Stories Vice City Stories
GOTO_IF_TRUE 1 Generic label GOTO_IF_TRUE 1 Positive/Negative offset INT
Vice City San Andreas
LOAD_AND_LAUNCH_MISSION_EXCLUSIVE 1 Mission script file LOAD_AND_LAUNCH_MISSION_INTERNAL 1 Negative mission identifier INT
Mission script label LOAD_AND_LAUNCH_MISSION_EXCLUSIVE Positive/Negative offset
Liberty City Stories Vice City Stories
CALL
CALLNOT 		1 Function label CALL
CALLNOT 		1 # of input arguments INT
2 # of output arguments
3 # of script locals
4 Positive/Negative offset
i Input arguments i Input arguments ANY_MULTI
o Output arguments o Output arguments HOLD_MULTI

This section is incomplete. You can help by fixing and expanding it.

Compare flag

The compare flag is an internal script-dependent flag which makes conditional GOTOs such as GOTO_IF_TRUE (unavailable in GTA III Vice City San Andreas) and GOTO_IF_FALSE deciding whether to jump otherwise. It can handle up to 8 checks per conditional statement and indicates you are verifying a single condition (0, see also Optimization) or multiple conditions with either AND (1 to 8) or OR (21 to 28) logical operators (see also ANDOR).

Control flows analysis

As an overview of the compiled source, control flows are literally nested meaning that the code is unoptimized. Furthermore, the jump of an embedded construct doesn't get merged with that of the construct itself, which consists of a benefit for the code parsing.

IF

As regards the IF control flow, if the whole check is true the consequence is performed and the code jumps to the end of the construct, otherwise it skips to the alternative (see also Compare flag):

Decompiled GTA III Vice City San Andreas Liberty City Stories Vice City Stories Compiled 
IF [NOT] {condition0}
[AND|OR [NOT] {condition8}]
    {consequence}
[ELSE
    {alternative}]
ENDIF

{..214}
{.....}
{.....}
{...77}
{.....}
{....2}
 ----- 
{.....}
 ----- 

{..219}
{.....}
{.....}
{...77}
{.....}
{....2}
 ----- 
{.....}
 ----- 

{..120}
{.....}
{.....}
{...34}
{.....}
{....2}
 ----- 
{.....}
 ----- 

ANDOR {value}
    [NOT] {condition0}
    [[NOT] {condition8}]
GOTO_IF_FALSE ELSE
    {consequence}
    [GOTO ENDIF]
ELSE:
    [{alternative}
ENDIF:]

IFNOT

Not that much to say more than the preceding construct, the IFNOT control flow is built nearly in the same way. In fact, the ELSE clause points to the alternative, whereas the GOTO jumps to its end. The substantial difference consists in the substitution of GOTO_IF_FALSE with GOTO_IF_TRUE:

Decompiled GTA III Liberty City Stories Vice City Stories Compiled 
IFNOT [NOT] {condition0}
[AND|OR [NOT] {condition8}]
    {consequence}
[ELSE
    {alternative}]
ENDIF

{..214}
{.....}
{.....}
{...76}
{.....}
{....2}
 ----- 
{.....}
 ----- 

{..219}
{.....}
{.....}
{...76}
{.....}
{....2}
 ----- 
{.....}
 ----- 

{..120}
{.....}
{.....}
{...33}
{.....}
{....2}
 ----- 
{.....}
 ----- 

ANDOR {value}
    [NOT] {condition0}
    [[NOT] {condition8}]
GOTO_IF_TRUE ELSE
    {consequence}
    [GOTO ENDIF]
ELSE:
    [{alternative}
ENDIF:]

WHILE

The WHILE control flow is built pretty much similarly to the previous, even though when the consequence is read the code is moved to the beginning of the construct:

Decompiled GTA III Vice City San Andreas Liberty City Stories Vice City Stories Compiled 
WHILE [NOT] {condition0}
[AND|OR [NOT] {condition8}]
    {consequence}
ENDWHILE

 ----- 
{..214}
{.....}
{.....}
{...77}
{.....}
{....2}
 ----- 

 ----- 
{..219}
{.....}
{.....}
{...77}
{.....}
{....2}
 ----- 

 ----- 
{..120}
{.....}
{.....}
{...34}
{.....}
{....2}
 ----- 

WHILE:
ANDOR {value}
    [NOT] {condition0}
    [[NOT] {condition8}]
GOTO_IF_FALSE ENDWHILE
    {consequence}
    GOTO WHILE
ENDWHILE:

WHILENOT

To say the least, WHILENOT control flow follows the same constitution of both WHILE and IFNOT constructs, by exchanging GOTO_IF_FALSE with GOTO_IF_TRUE:

Decompiled GTA III Liberty City Stories Vice City Stories Compiled 
WHILENOT [NOT] {condition0}
[AND|OR [NOT] {condition8}]
    {consequence}
ENDWHILE

 ----- 
{..214}
{.....}
{.....}
{...76}
{.....}
{....2}
 ----- 

 ----- 
{..219}
{.....}
{.....}
{...76}
{.....}
{....2}
 ----- 

 ----- 
{..120}
{.....}
{.....}
{...33}
{.....}
{....2}
 ----- 

WHILENOT:
ANDOR {value}
    [NOT] {condition0}
    [[NOT] {condition8}]
GOTO_IF_TRUE ENDWHILE
    {consequence}
    GOTO WHILENOT
ENDWHILE:

REPEAT

Seemingly, the REPEAT control flow is the first construct ever optimized as a result of a possible R* compiler fault. Moreover, it sounds ambiguous as it loops at least once. This was probably the intention of R* programmers, that is iterating at least once else the construct is useless. However, there are few chance they decide to use such structure to avoid some conflict with some other constructs:

Decompiled G/L Vice City San Andreas Liberty City Stories G/L Vice City Stories Compiled 
REPEAT {times} {varname}
    {consequence}
ENDREPEAT

{....4}
 ----- 
{.....}
{....8}
{...40}
{...77}

{....5}
 ----- 
{.....}
{....9}
{...41}
{...77}

{....4}
 ----- 
{.....}
{....7}
{...21}
{...34}

{varname} = {value0}
LOOP:
{consequence}
++ {varname}
    {varname} >= {valueN}
GOTO_IF_FALSE LOOP

SWITCH

In San Andreas, the SWITCH control flow is more complex and efficient because the game uses internally a binary search algorithm to jump at the label that matches with the value of a particular case. This method requires a known amount of cases which is up to 75. When a case is true, a consequence is executed and the code jumps to the end of the construct, otherwise the alternative may be performed. As the code is unoptimized, the GOTO of the last case is still compiled even though its label points right after the command mentioned earlier:

Decompiled San Andreas Compiled 
SWITCH {varname}
    CASE {value0}
        {consequence}
        BREAK
    [CASE {valueN}
        {consequence}
        BREAK]
    [DEFAULT
        {alternative}
        BREAK]
ENDSWITCH

{.2161}
{.2162}
 ----- 
{.....}
{....2}
 ----- 
{.....}
{....2}
 ----- 
{.....}
{....2}
 ----- 

SWITCH_START {varname} {numcases} {isdefault} DEFAULT {value0} CASE0 ...
[SWITCH_CONTINUED ... {valueN} CASEN [-1 ENDSWITCH]]
CASE0:
    {consequence}
    GOTO ENDSWITCH
[CASEN:
    {consequence}
    GOTO ENDSWITCH]
[DEFAULT:
    {alternative}
    GOTO ENDSWITCH]
ENDSWITCH:

In Liberty City Stories and Vice City Stories, such control flow is a set of nested IF constructs which causes a very slight loss of performance by considering that ANDOR isn't compiled:

Decompiled G/L Liberty City Stories G/L Vice City Stories Compiled 
SWITCH {varname}
    CASE {value0}
        {consequence}
        BREAK]
    [CASE {valueN}
        {consequence}
        BREAK]
    [DEFAULT
        {alternative}
        BREAK]
ENDSWITCH

 ----- 
{...56}
{...77}
{.....}
{....2}
 ----- 
{...56}
{...77}
{.....}
{....2}
 ----- 
{.....}
 ----- 
 ----- 

 ----- 
{...57}
{...77}
{.....}
{....2}
 ----- 
{...57}
{...77}
{.....}
{....2}
 ----- 
{.....}
 ----- 
 ----- 

 ----- 
{...27}
{...34}
{.....}
{....2}
 ----- 
{...27}
{...34}
{.....}
{....2}
 ----- 
{.....}
 ----- 
 ----- 

CASE0:
    {varname} = {value0}
GOTO_IF_FALSE CASEN
    {consequence}
    GOTO ENDSWITCH0
CASEN:
        [{varname} = {valueN}
    GOTO_IF_FALSE DEFAULT
        {consequence}
        GOTO ENDSWITCHN
    DEFAULT:
        [{alternative}
    ENDSWITCHN:]]
ENDSWITCH0:

Optimization

In Liberty City Stories and Vice City Stories, whenever a single condition is checked ANDOR doesn't get compiled cause no logical operator (AND, OR) is used and so they become really useless. Its lack increases the script efficiency a lot. However, the jump of the ELSE clause of an IF statement which points to the end of the construct is still compiled after a GOTO. Furthermore, Stories Games come with an improved data type managing which causes a considerable decrease of the compiled file size.

Tools

See also

External links

v · d · e
San Andreas Grand Theft Auto: San Andreas
File Formats.b • .col • .cfg • .cut • .dff • .dat • .fxp • .gxt • .ide • .ifp • .img • .ipl • nodes*.dat • .ped • .rep • .rrr • .scm • .set • .txd
DocumentationAudio • Cryptography • Cutscenes • Game memory • Handling.cfg • Map Listing • Mission Packs • Instructions • Paths • Replays • Saves • Scripts • Sound Effects • Statistics • Vehicles • Wanted levels
ToolsCLEO • Collision File Editor II • ENBSeries • G-Tools • IMG Tool • Limit Adjuster • Map Editor • Mod Loader • San Andreas Audio Toolkit • Sanny Builder • TXD Workshop • Magic.TXD
TutorialsSan Andreas v2.0 Modding • How to create a mission • How to create a script • How to use Map Editor • Vehicle Mod Installation
ModificationsDesign Your Own Mission • Gostown Paradise • GTA: United • Myriad Islands
MultiplayergtaTournament • Multi Theft Auto • San Andreas Multiplayer • (more...)
Useful linksCommunity Portal • Discussion Forums • Modding Forums • Mods on GTAGarage.com • Mobile Modding • Opcodes Database
v · d · e
Vice City Grand Theft Auto: Vice City
File Formats.adf • .b • .col • .cfg • .dff • .dat • .gxt • .ide • .ifp • .img/.dir • .ipl • .raw/.sdt • .rep • .set • .scm • .txd
DocumentationAudio • Handling • Map Listing • Memory Addresses • Skin List • Opcodes • RenderWare • Script Paths • Saves • Sound Effects • Statistics • Vehicles • Text • Wanted levels • Weapons
ToolsCLEO • Collision File Editor II • G-Tools • GXT Editor • IMG Tool • Limit Adjuster • KEd (map editor) • Sanny Builder • TXD Workshop • Magic.TXD
TutorialsHow to create a mission • How to create a script • Vehicle Mod Installation
MultiplayerGTA:LC Multiplayer • gtaTournament • Multi Theft Auto • State Of Liberty Online • Vice City Multiplayer • Vice City Online
Useful linksCommunity Portal • Discussion Forums • Modding Forums • Mods on GTAGarage.com • Mobile Modding • Opcodes Database
v · d · e
GTA III Grand Theft Auto III
File Formats.b • .col • .cfg • .dff • .dat • .gxt • .ide • .ifp • .img • .ipl • .raw/.sdt • .set • .scm • .txd
DocumentationAudio • Handling.cfg • Map Listing • Opcodes • Paths • RenderWare • Saves • Sound Effects • Statistics • Vehicles • Wanted levels • Weapons
ToolsCLEO • Collision File Editor II • G-Tools • IMG Tool • Map Editor • Sanny Builder • TXD Workshop • Magic.TXD • Water Editor
TutorialsHow to create a mission • How to create a script • How to use Map Editor • Vehicle Mod Installation
MultiplayerLiberty Unleashed • Multi Theft Auto • More...
Useful linksCommunity Portal • Discussion Forums • Modding Forums • Mods on GTAGarage.com • Mobile Modding • Opcodes Database
Retrieved from "https://gtamods.com/mediawiki/index.php?title=SCM_language&oldid=18622"