fabric-carpet/docs/scarpet/Full.md

Fundamental components of scarpet programming language (towards version 1.7).

Scarpet (a.k.a. Carpet Script, or Script for Carpet) is a programming language designed to provide the ability to write custom programs to run within Minecraft and interact with the world.

The project was initially built based on the EvalEx project, however it now diverged so far away from the original implementation, it would be hard to tell it without this mention. EvalEx is a handy expression evaluator for Java, that allows to evaluate simple mathematical and boolean expressions. For more information, see: EvalEx GitHub repository

This specification is divided into two sections: this one is agnostic to any Minecraft related features and could function on its own, and CarpetExpression for Minecraft specific routines and world manipulation functions.

Synopsis

script run print('Hello World!')

or an OVERLY complex example:

/script run
    block_check(x1, y1, z1, x2, y2, z2, block_to_check) ->
    (
        l(minx, maxx) = sort(l(x1, x2));
        l(miny, maxy) = sort(l(y1, y2));
        l(minz, maxz) = sort(l(z1, z2));
        'Need to compute the size of the area of course';
        'Cause this language doesn\'t support comments in the command mode';
        xsize = maxx - minx + 1;
        ysize = maxy - miny + 1;
        zsize = maxz - minz + 1;
        total_count = 0;
        loop(xsize,
            xx = minx + _ ;
            loop(ysize,
                yy = miny + _ ;
                loop(zsize,
                    zz = minz + _ ;
                    if ( block(xx,yy,zz) == block_to_check,
                        total_count += ceil(rand(1))
                    )
                )
            )
        );
        total_count
    );
    check_area_around_closest_player_for_block(block_to_check) ->
    (
        closest_player = player();
        l(posx, posy, posz) = query(closest_player, 'pos');
        total_count = block_check( posx-8,1,posz-8, posx+8,17,posz+8, block_to_check);
        print('There is '+total_count+' of '+block_to_check+' around you')
    )
/script invoke check_area_around_closest_player_for_block 'diamond_ore'

or simply

/script run print('There is '+for(rect(x,9,z,8,8,8), _ == 'diamond_ore')+' diamond ore around you')

It definitely pays to check what higher level scarpet functions have to offer.

Programs

You can think of an program like a mathematical expression, like "2.4*sin(45)/(2-4)" or "sin(y)>0 & max(z, 3)>3". Writing a program, is like writing a 2+3, just a bit longer.

Basic language components

Programs consist of constants, like 2, 3.14, pi, or 'foo', operators like +, /, ->, variables which you can define, like foo or special ones that will be defined for you, like _x, or _ , which I specific to a each built in function, and functions with name, and arguments in the form of f(a,b,c), where f is the function name, and a, b, c are the arguments which can be any other expression. And that's all the parts of the language, so all in all - sounds quite simple.

Code flow

Like any other proper programming language, scarpet needs brackets, basically to identify where stuff begins and where it ends. In the languages that uses much more complicated constructs, like Java, they tend to use all sort of them, round ones to indicate function calls, curly to indicate section of code, square to access lists, pointy for generic types etc... I mean - there is no etc, cause they have exhausted all the bracket options...

Scarpet is different, since it runs everything based on functions (although its not per se a functional language like lisp) only needs the round brackets for everything, and it is up to the programmer to organize its code so its readable, as adding more brackets does not have any effect on the performance of the programs as they are compiled before they are executed. Look at the following example usage of if() function:

if(x<y+6,set(x,8+y,z,'air');plop(x,top('surface',x,z),z,'birch'),sin(query(player(),'yaw'))>0.5,plop(0,0,0,'boulder'),particle('fire',x,y,z))

Would you prefer to read

if(   x<y+6,
           set(x,8+y,z,'air');
           plop(x,top('surface',x,z),z,'birch'),
      sin(query(player(),'yaw'))>0.5,
           plop(0,0,0,'boulder'),
      particle('fire',x,y,z)
)

Or rather:

if
(   x<y+6,
    (
        set(x,8+y,z,'air');
        plop(x,top('surface',x,z),z,'birch')
    ),
    // else if
    sin(query(player(),'yaw'))>0.5,
    (
        plop(0,0,0,'boulder')
    ),
    // else
    particle('fire',x,y,z)
)

Whichever style you prefer it doesn't matter. It typically depends on the situation and the complexity of the subcomponents. No matter how many whitespaces and extra brackets you add - the code will evaluate to exactly the same expression, and will run exactly the same, so make sure your programs are nice and clean so others don't have problems with them

Functions and scoping

Users can define functions in the form fun(args....) -> expression and they are compiled and saved for further execution in this, but also subsequent calls of /script command, added to events, etc. Functions can also be assigned to variables, passed as arguments, called with call('fun', args...) function, but in most cases you would want to call them directly by name, in the form of fun(args...). This means that once defined functions are saved with the world for further use. For variables, there are two types of them, global - which are shared anywhere in the code, and those are all which name starts with 'global_', and local variables which is everything else and those are only visible inside each function. This also means that all the parameters in functions are passed 'by value', not 'by reference'.

Outer variables

Functions can still 'borrow' variables from the outer scope, by adding them to the function signature wrapped around built-in function outer. It adds the specified value to the function call stack so they behave exactly like capturing lambdas in Java, but unlike java captured variables don't need to be final. Scarpet will just attach their new values at the time of the function definition, even if they change later. Most value will be copied, but mutable values, like maps or lists, allow to keep the 'state' with the function, allowing them to have memory and act like objects so to speak. . Check outer(var) for details.

Code delivery, line indicators

Note that this should only apply to pasting your code to execute with commandblock. Scarpet recommends placing your code in apps (files with .sc extension that can be placed inside "/scripts" folder in the world files and loaded as a scarpet app with command /script load [app_name]. Scarpet apps loaded from disk should only contain code, no need to start with "/script run" prefix

The following is the code that could be provided in a foo.sc app file located in world /scripts folder

run_program() -> (
  loop( 10,
    // looping 10 times
    // comments are allowed in scripts located in world files
    // since we can tell where that line ends
    foo = floor(rand(10));
    check_not_zero(foo);
    print(_+' - foo: '+foo);
    print('  reciprocal: '+  _/foo )
  )
);
check_not_zero(foo) -> (
  if (foo==0, foo = 1)
)

Which we then call in-game with:

/script load foo
/script in foo invoke run_program

However the following code can also be input as a command, or in a command block.

Since the maximum command that can be input to the chat is limited in length, you will be probably inserting your programs by pasting them to command blocks or reading from world files, however pasting to command blocks will remove some whitespaces and squish your newlines making the code not readable. If you are pasting a program that is perfect and will never cause an error, I salute you, but for the most part it is quite likely that your program might break, either at compile time, when its initially analyzed, or at execute time, when you suddenly attempt to divide something by zero. In these cases you would want to get a meaningful error message, but for that you would need to indicate for the compiler where did you put these new lines, since command block would squish them. For that, place at the beginning of the line to let the compiler know where are you. This makes so that $ is the only character that is illegal in programs, since it will be replaced with new lines. As far as I know, $ is not used anywhere inside Minecraft identifiers, so this shouldn't hinder the abilities of your programs.

Consider the following program executed as command block command:

/script run
run_program() -> (
  loop( 10,
    foo = floor(rand(_));
    check_not_zero(foo);
    print(_+' - foo: '+foo);
    print('  reciprocal: '+  _/foo )
  )
);
check_not_zero(foo) -> (
   if (foo==0, foo = 1)
)

Lets say that the intention was to check if the bar is zero and prevent division by zero in print, but because the foo is passed as a variable, it never changes the original foo value. Because of the inevitable division by zero, we get the following message:

Your math is wrong, Incorrect number format for NaN at pos 98
run_program() -> ( loop( 10, foo = floor(rand(_)); check_not_zero(foo); print(_+' - foo: '+foo);
HERE>> print(' reciprocal: '+ _/foo ) ));check_not_zero(foo) -> ( if (foo==0, foo = 1))

As we can see, we got our problem where the result of the mathematical operation was not a number (infinity, so not a number), however by pasting our program into the command made it squish the newlines so while it is clear where the error happened and we still can track the error down, the position of the error (98) is not very helpful and wouldn't be useful if the program gets significantly longer. To combat this issue we can precede every line of the script with dollar signs $:

/script run
$run_program() -> (
$  loop( 10,
$    foo = floor(rand(_));
$    check_not_zero(foo);
$    print(_+' - foo: '+foo);
$    print('  reciprocal: '+  _/foo )
$  )
$);
$check_not_zero(foo) -> (
$   if (foo==0, foo = 1)
$)

Then we get the following error message

Your math is wrong, Incorrect number format for NaN at line 7, pos 2
  print(_+' - foo: '+foo);
   HERE>> print(' reciprocal: '+ _/foo )
  )

As we can note not only we get much more concise snippet, but also information about the line number and position, so means its way easier to locate the potential problems problem

Obviously that's not the way we intended this program to work. To get it foo modified via a function call, we would either return it as a result and assign it to the new variable:

foo = check_not_zero(foo);
...
check_not_zero(foo) -> if(foo == 0, 1, foo)

.. or convert it to a global variable, which in this case passing as an argument is not required

global_foo = floor(rand(10));
check_foo_not_zero();
...
check_foo_not_zero() -> if(global_foo == 0, global_foo = 1)

# Variables and Constants

scarpet provides a number of constants that can be used literally in scripts

• null: nothing, zilch, not even false
• true: pure true, or just 1 (one)
• false: false truth, or true falsth, 0 (zero) actually
• pi: for the fans of perimeters, its a perimeter of an apple pi of diameter 1. About 3.14
• euler: clever guy. Derivative of its exponent is goto 1. About 2.72

Apart from that, there is a bunch of system variables, that start with _ that are set by scarpet built-ins, like _, typically each consecutive value in loops, _i indicating iteration, or _a like an accumulator for reduce function. Certain calls to Minecraft specific calls would also set _x, _y, _z, indicating block positions. All variables starting with _ are read-only, and cannot be declared and modified in client code.

Literals

scarpet accepts numeric and string liters constants. Numbers look like 1, 2.5, -3e-7, 0xff, and are internally represented as Java's double but scarpet will try to trim trailing zeros as much as possible so if you need to use them as intergers, you can. Strings using single quoting, for multiple reasons, but primarily to allow for easier use of strings inside doubly quoted command arguments (when passing a script as a parameter of /script fill for example, or when typing in jsons inside scarpet to feed back into a /data merge command for example. Strings also use backslashes \ for quoting special characters, in both plain strings and regular expressions

'foo'
print('This doesn\'t work')
nbt ~ '\\.foo'   // matching '.' as a '.', not 'any character match'

Operators

There is a number of operators you can use inside the expressions. Those could be considered generic type operators that apply to most data types. They also follow standard operator precedence, i.e. 2+2*2 is understood as 2+(2*2), not (2+2)*2, otherwise they are applied from left to right, i.e. 2+4-3 is interpreted as (2+4)-3, which in case of numbers doesn't matter, but since scarpet allows for mixing all value types the associativity would matter, and may lead to unintended effects:

Important operator is function definition -> operator. It will be covered in User Defined Functions and Program Control Flow

'123'+4-2 => ('123'+4)-2 => '1234'-2 => '134'
'123'+(4-2) => '123'+2 => '1232'
3*'foo' => 'foofoofoo'
1357-5 => 1352
1357-'5' => 137
3*'foo'-'o' => 'fff'
l(1,3,5)+7 => l(8,10,12)

As you can see, values can behave differently when mixed with other types in the same expression. In case values are of the same types, the result tends to be obvious, but Scarpet tries to make sense of whatever it has to deal with

Operator Precedence

Here is the complete list of operators in scarpet including control flow operators. Note, that commas and brackets are not technically operators, but part of the language, even if they look like them:

• Match, Get ~ :
• Unary + - !
• Exponent ^
• Multiplication * / %
• Addition + -
• Comparison == != > >= <= <
• Logical And&&
• Logical Or ||
• Assignment = += <>
• Definition ->
• Next statement;
• Comma ,
• Bracket ( )

Get, Accessor Operator :

Operator version of the get(...) function to access elements of lists, maps, and potentially other containers (i.e. NBTs). It is important to distinguish from ~ operator, which is a matching operator, which is expected to perform some extra computations to retrieve the result, while : should be straightforward and immediate, and the source object should behave like a container and support full container API, meaning get(...), put(...), delete(...), and has(...) functions

For certain operators and functions (get, put, delete, has, =, +=) objects can use : annotated fields as l-values, meaning construct like foo:0 = 5, would act like put(foo, 0, 5), rather than get(foo, 0) = 5, which would result in an error.

TODO: add more information about l-value behaviour.

Matching Operator ~

This operator should be understood as 'matches', 'contains', 'is_in', or 'find me some stuff about something else. For strings it matches the right operand as a regular expression to the left, returning the first match. This can be used to extract information from unparsed nbt's in a more convoluted way (use get(...) for more appropriate way of doing it). For lists it checks if an element is in the list, and returns the index of that element, or null if no such element was found, especially that the use of first(...) function will not return the index. Currently it doesn't have any special behaviour for numbers - it checks for existence of characters in string representation of the left operand with respect of the regular expression on the right hand side.

In Minecraft API portion entity ~ feature is a shortcode for query(entity,feature) for queries that do not take any extra arguments.

l(1,2,3) ~ 2  => 1
l(1,2,3) ~ 4  => null
'foobar' ~ '.b'  => 'ob'
player('*') ~ 'gnembon'  // null unless player gnembon is logged in (better to use player('gnembon') instead
p ~ 'sneaking' // if p is an entity returns whether p is sneaking

Or a longer example of an ineffective way to searching for a squid

entities = entities_area('all',x,y,z,100,10,100);
sid = entities ~ 'Squid';
if(sid != null, run('execute as '+query(get(entities,sid),'id')+' run say I am here '+query(get(entities,sid),'pos') ) )

Or an example to find if a player has specific enchantment on a held axe (either hand) and get its level (not using proper NBTs query support via get(...)):

global_get_enchantment(p, ench) -> (
$   for(l('main','offhand'),
$      holds = query(p, 'holds', _);
$      if( holds,
$         l(what, count, nbt) = holds;
$         if( what ~ '_axe' && nbt ~ ench,
$            lvl = max(lvl, number(nbt ~ '(?<=lvl:)\\d') )
$         )
$      )
$   );
$   lvl
$);
/script run global_get_enchantment(players(), 'sharpness')

Basic Arithmetic Operators + - * /

Allows to add the results of two expressions. If the operands resolve to numbers, the result is arithmetic operation. In case of strings, adding or subtracting from a string results in string concatenation and removal of substrings from that string. Multiplication of strings and numbers results in repeating the string N times and division results in taking the first k'th part of the string, so that str*n/n ~ str In case first operand is a list, either it results in a new list with all elements modified one by one with the other operand, or if the operand is a list with the same number of items - element-wise addition/subtraction

Examples:

2+3 => 5
'foo'+3+2 => 'foo32'
'foo'+(3+2) => 'foo5'
3+2+'bar' => '5bar'
'foo'*3 => 'foofoofoo'
'foofoofoo' / 3 => 'foo'
'foofoofoo'-'o' => 'fff'
l(1,2,3)+1  => l(2,3,4)
b = l(100,63,100); b+l(10,0,10)  => l(110,63,110)

Just Operators % ^

The modulo and exponent (power) operators work only if both operands are numbers

pi^pi%euler  => 1.124....
-9 % 4  => -1
9 % -4  => 0 ¯\_(ツ)_/¯ Java
-3 ^ 2  => 9
-3 ^ pi => // Error

Comparison Operators == != < > <= >=

Allows to compare the results of two expressions. For numbers it is considers arithmetic order of numbers, for strings - lexicographical, nulls are always 'less' than everything else, and lists check their elements - if the sizes are different, the size matters, otherwise, pairwise comparisons for each elements are performed. The same order rules than with all these operators are used with the default sortographical order as used by sort function. All of these are true:

null == null
null != false
0 == false
1 == true
null < 0
null < -1000
1000 < 'a'
'bar' < 'foo'
3 == 3.0

Logical Operators && ||

These operator compute respective boolean operation on the operands. What it important is that if calculating of the second operand is not necessary, it won't be evaluated, which means one can use them as conditional statements. In case of success returns first positive operand (||) or last one (&&).

true || false  => 1
null || false => 0
null != false || run('kill gnembon')  => 1 // gnembon survives
null != false && run('kill gnembon')  => 0 // when cheats not allowed
null != false && run('kill gnembon')  => 1 // gnembon dies, cheats allowed

Assignment Operators = <> +=

A set of assignment operators. All require bounded variable on the LHS, <> requires bounded arguments on the right hand side as well (bounded, meaning being variables). Additionally they can also handle list constructors with all bounded variables, and work then as list assignment operators. When += is used on a list, it extends that list of that element, and returns the list (old == new). scarpet doesn't support currently removal of items. Removal of items can be obtaine via filter command, and reassigning it fo the same variable. Both operations would require rewriting of the array anyways.

a = 5  => a == 5
l(a,b,c) = l(3,4,5) => a==3, b==4, c==5
l(minx,maxx) = sort(xi,xj);  // minx assumes min(xi, xj) and maxx, max(xi, xj)
l(a,b,c,d,e,f) = l(range(6)); l(a,b,c) <> l(d,e,f); l(a,b,c,d,e,f)  => [3,4,5,0,1,2]
a = l(1,2,3); a += 4  => [1,2,3,4]
a = l(1,2,3,4); a = filter(a,_!=2)  => [1,3,4]

Unary Operators - +

Require a number, flips the sign. One way to assert it's a number is by crashing the script. gg.

-4  => -4
+4  => 4
+'4'  // Error message

Negation Operator !

flips boolean condition of the expression. Equivalent of bool(expr)==false

!true  => 0
!false  => 1
!null  => 1
!5  => 0
!l() => 1
!l(null) => 0

Arithmetic operations

Basic Arithmetic Functions

There is bunch of them - they require a number and spit out a number, doing what you would expect them to do.

fact(n)

Factorial of a number, a.k.a n!, just not in scarpet. Gets big... quick...

sqrt(n)

Square root. For other fancy roots, use ^, math and yo noggin. Imagine square roots on a tree...

abs(n)

Absolut value.

round(n)

Closest integer value. Did you know the earth is also round?

floor(n)

Highest integer that is still no larger then n. Insert a floor pun here.

ceil(n)

First lucky integer that is not smalller than n. As you would expect, ceiling is typically right above the floor.

ln(n)

Natural logarithm of n. Naturally.

ln1p(n)

Natural logarithm of n+1. Very optimistic.

log10(n)

Decimal logarithm of n. Its ceiling is the length of its floor.

log(n)

Binary logarithm of n. Finally, a proper one, not like the previous 11.

log1p(n)

Binary logarithm of n+1. Also always positive.

mandelbrot(a, b, limit)

Computes the value of the mandelbrot set, for set a and b spot. Spot the beetle. Why not.

min(arg, ...), min(list), max(arg, ...), max(list)

Compute minimum or maximum of supplied arguments assuming default sorthoraphical order. In case you are missing argmax, just use a ~ max(a), little less efficient, but still fun.

Interesting bit - min and max don't remove variable associations from arguments, which means can be used as LHS of assignments (obvious case), or argument spec in function definitions (far less obvious).

a = 1; b = 2; min(a,b) = 3; l(a,b)  => [3, 2]
a = 1; b = 2; fun(x, min(a,b)) -> l(a,b); fun(3,5)  => [5, 0]

Absolutely no idea, how the latter might be useful in practice. But since it compiles, can ship it.

relu(n)

Linear rectifier of n. 0 below 0, n above. Why not. max(0,n) with less moral repercussions.

Trigonometric / Geometric Functions

sin(x)

cos(x)

tan(x)

asin(x)

acos(x)

atan(x)

atan2(x,y)

sinh(x)

cosh(x)

tanh(x)

sec(x)

csc(x)

sech(x)

csch(x)

cot(x)

acot(x)

coth(x)

asinh(x)

acosh(x)

atanh(x)

rad(deg)

deg(rad)

Use as you wish

System functions

Type conversion functions

copy(expr)

Returns the deep copy of the expression. Can be used to copy mutable objects, like maps and lists

type(expr)

Returns the string value indicating type of the expression. Possible outcomes are null, number, string, list, iterator, as well as minecraft related concepts like block, entity, nbt

bool(expr)

Returns a boolean context of the expression. Note that there are no true/false values in scarpet. true is alias of 1, and false is 0. Bool is also interpreting string values as boolean, which is different from other places where boolean context can be used. This can be used in places where API functions return string values to represent binary values

bool(pi) => 1
bool(false) => 0
bool('') => 0
bool(l()) => 0
bool(l('')) => 1
bool('foo') => 1
bool('false') => 0
bool('nulL') => 0
if('false',1,0) => 1

number(expr)

Returns a numeric context of the expression. Can be used to read numbers from strings

number(null) => null
number(false) => 0
number('') => null
number('3.14') => 3.14
number(l()) => 0
number(l('')) => 1
number('foo') => null
number('3bar') => null
number('2')+number('2') => 4

str(expr, params? ... ), str(expr, param_list)

Returns a formatted string representing expression. Accepts formatting style accepted by String.format. Supported types (with "%?" syntax):

• d, o, x: integers, octal, hex
• a, e, f, g: floats
• b: booleans
• s: strings

str(null) => null
str(false) => 0
str('') => null
str('3.14') => 3.14
str(l()) => 0
str(l('')) => 1
str('foo') => null
str('3bar') => null
str(2)+str(2) => 22
str('pi: %.2f',pi) => 'pi: 3.14'
str('player at: %d %d %d',pos(player())) => 'player at: 567, -2423, 124'

---

Threading and Parallel Execution

Scarpet allows to run threads of execution in parallel to the main script execution thread. In Minecraft, the main thread scripts are executed on the main server thread. Since Minecraft is inherently NOT thread safe, it is not that beneficial to parallel execution in order to access world resources faster. Both GetBlockState and setBlockState are not thread safe and require the execution to park on the server thread, where these requests can be executed in the off-tick time in between ticks that didn't take 50ms. There are however benefits of running things in parallel, like fine time control not relying on the tick clock, or running things independent on each other. You can still run your actions on tick-by-tick basis, either taking control of the execution using game_tick() API function (nasty solution), or scheduling tick using schedule() function (much nicer solution), but threading often gives the neatest solution to solve problems in parallel (see scarpet camera).

Due to limitations with the game, there are some limits to the threading as well. You cannot for instance join_task() at all from the main script and server thread, because any use of Minecraft specific function that require any world access, will require to park and join on the main thread to get world access, meaning that calling join on that task would inevitably lead to a typical deadlock. You can still join tasks from other threads, just because the only possibility of a deadlock in this case would come explicitly from your bad code, not the internal world access behaviour. Some things tough like player or entity manipulation, can be effectively parallelized.

task(function, ?args...., ?executor)

Creates and runs a parallel task, returning the handle to the task object. Task will return the return value of the function when its completed, or will return null immediately if task is still in progress, so grabbing a value of a task object is non-blocking. Function can be either function value, or function lambda, or a name of an existing defined function. In case function needs arguments to be called with, they should be supplied after the function name, or value. Optional executor identifier is to place the task in a specific queue identified by this value. The default thread value is the null thread. There is no limits on number of parallel tasks for any executor, so using different queues is solely for synchronization purposes. Also, since task function knows how many extra arguments it requires, the use of tailing optional parameter for the custom executor thread pool is not ambiguous.

task( _() -> print('Hello Other World') )  => Runs print command on a separate thread
foo(a, b) -> print(a+b); task('foo',2,2)  => Uses existing function definition to start a task
task('foo',3,5,'temp');  => runs function foo with a different thread executor, identified as 'temp'
a = 3; task( _(outer(a), b) -> foo(a,b), 5, 'temp')  
    => Another example of running the same thing passing arguments using closure over anonymous function as well as passing a parameter.

task_count(executor?)

If no argument provided, returns total number of tasks being executed in parallel at this moment using scarpet threading system. If the executor is provided, returns number of active tasks for that provider. Use task_count(null) to get the task count of the default executor only.

task_value(task)

Returns the task return value, or null if task hasn't finished yet. Its a non-blocking operation. Unlike join_task, can be called on any task at any point

task_join(task)

Waits for the task completion and returns its computed value. If the task has already finished returns it immediately. Unless taking the task value directly, i.e. via task_value, this operation is blocking. Since Minecraft has a limitation that all world access operations have to be performed on the main game thread in the off-tick time, joining any tasks that use Minecraft API from the main thread would mean automatic lock, so joining from the main thread is not allowed. Join tasks from other threads, if you really need to, or communicate asynchronously with the task via globals or function data / arguments to monitor its progress, communicate, get partial results, or signal termination.

task_completed(task)

Returns true if task has completed, or false otherwise.

synchronize(lock, expression)

Evaluates expression synchronized with respect to the lock lock. Returns the value of the expression.

---

Auxiliary functions

lower(expr), upper(expr), title(expr)

Returns lowercase, uppercase or titlecase representation of a string representation of the passed expression

lower('aBc') => 'abc'
upper('aBc') => 'ABC'
title('aBc') => 'Abc'

replace(string, regex, repl?); replace_first(string, regex, repl?)

Replaces all, or first occurence of a regular expression in the string with repl expression, or nothing, if not specified

replace('abbccddebfg','b+','z')  // => azccddezfg
replace_first('abbccddebfg','b+','z')  // => azccddebfg

length(expr)

Returns length of the expression, the length of the string, the length of the integer part of the number, or length of the list

length(pi) => 1
length(pi*pi) => 1
length(pi^pi) => 2
length(l()) => 0
length(l(1,2,3)) => 3
length('') => 0
length('foo') => 3

rand(expr), rand(expr, seed)

returns a random number from 0.0 (inclusive) to expr (exclusive). In boolean context (in conditions, boolean functions, or bool), returns false if the randomly selected value is less than 1. This means that rand(2) returns true half of the time and rand(5) returns true for 80% (4/5) of the time. If seed is not provided, uses a random seed. If seed is provided, each consecutive call to rand() will act like 'next' call to the same random object. Scarpet keeps track of up to 1024 custom random number generators, so if you exceed this number (per app), then your random sequence will revert to the beginning.

map(range(10), floor(rand(10))) => [5, 8, 0, 6, 9, 3, 9, 9, 1, 8]
map(range(10), bool(rand(2))) => [1, 1, 1, 0, 0, 1, 1, 0, 0, 0]
map(range(10), str('%.1f',rand(_))) => [0.0, 0.4, 0.6, 1.9, 2.8, 3.8, 5.3, 2.2, 1.6, 5.6]

perlin(x), perlin(x, y), perlin(x, y, z), perlin(x, y, z, seed)

returns a noise value from 0.0 to 1.0 (roughly) for 1, 2 or 3 dimensional coordinate. The default seed it samples from is 0, but seed can be specified as a 4th argument as well. In case you need 1D or 2D noise values with custom seed, use null for z, or y and z arguments respectively.

Perlin noise is based on a square grid and generates rougher maps comparing to Simplex, which is creamier. Querying for lower-dimensional result, rather than affixing unused dimensions to constants has a speed benefit,

Thou shall not sample from noise changing seed frequently. Scarpet will keep track of the last 256 perlin seeds used for sampling providing similar speed comparing to the default seed of 0. In case the app engine uses more than 256 seeds at the same time, switching between them can get much more expensive.

simplex(x, y), simplex(x, y, z), simplex(x, y, z, seed)

returns a noise value from 0.0 to 1.0 (roughly) for 2 or 3 dimensional coordinate. The default seed it samples from is 0, but seed can be specified as a 4th argument as well. In case you need 2D noise values with custom seed, use null for z argument.

Simplex noise is based on a triangular grid and generates smoother maps comparing to Perlin. To sample 1D simplex noise, affix other coordinate to a constant.

Thou shall not sample from noise changing seed frequently. Scarpet will keep track of the last 256 simplex seeds used for sampling providing similar speed comparing to the default seed of 0. In case the app engine uses more than 256 seeds at the same time, switching between them can get much more expensive.

print(expr)

prints the value of the expression to chat. Passes the result of the argument to the output unchanged, so print-statements can be weaved in code to debug programming issues

print('foo') => results in foo, prints: foo
a = 1; print(a = 5) => results in 5, prints: 5
a = 1; print(a) = 5 => results in 5, prints: 1
print('pi = '+pi) => prints: pi = 3.141592653589793
print(str('pi = %.2f',pi)) => prints: pi = 3.14

sleep(expr)

Halts the execution of the program (and the game itself) for expr milliseconds. All in all, its better to use game_tick(expr) to let the game do its job while the program waits

sleep(50)

time()

Returns the number of milliseconds since 'some point', like Java's System.nanoTime(). It returns a float, which has 1 microsecond precision (0.001 ms)

start_time = time();
flip_my_world_upside_down();
print(str('this took %d milliseconds',time()-start_time))

profile_expr(expression)

Returns number of times given expression can be run in 50ms time. Useful to profile and optimize your code. Note that, even if its only a number, it WILL run these commands, so if they are destructive, you need to be careful.

---

Access to variables and stored functions (use with caution)

var(expr)

Returns the variable under the name of the string value of the expression. Allows to manipulate variables in more programmatic manner, which allows to use local variable set with a hash map type key-value access, can also be used with global variables

a = 1; var('a') = 'foo'; a => a == 'foo'

undef(expr)

Removes all bindings of a variable with a name of expr. Removes also all function definitions with that name. It can affect global variable pool, and local variable set for a particular function.

inc(i) -> i+1; foo = 5; inc(foo) => 6
inc(i) -> i+1; foo = 5; undef('foo'); inc(foo) => 1
inc(i) -> i+1; foo = 5; undef('inc'); undef('foo'); inc(foo) => Error: Function inc is not defined yet at pos 53
undef('pi')  => bad idea - removes hidden variable holding the pi value
undef('true')  => even worse idea, unbinds global true value, all references to true would now refer to the default 0

vars(prefix)

It returns all names of variables from local scope (if prefix does not start with 'global') or global variables (otherwise). Here is a larger example that uses combination of vars and var functions to be used for object counting

/script run
$ count_blocks(ent) -> (
$   l(cx, cy, cz) = query(ent, 'pos');
$   scan(cx, cy, cz, 16, 16, 16, var('count_'+_) += 1);
$   for ( sort_key( vars('count_'), -var(_)),
$     print(str( '%s: %d', slice(_,6), var(_) ))
$   )
$ )
/script run count_blocks(player())

---

System key-value storage

Scarpet runs apps in isolation. The can share code via use of shared libraries, but each library that is imported to each app is specific to that app. Apps can store and fetch state from disk, but its restricted to specific locations meaning apps cannot interact via disk either. To facilitate communication for interappperability, scarpet hosts its own key-value storage that is shared between all apps currently running on the host, providing methods for getting an associated value with optional setting it if not present, and an operation of modifying a content of a system global value.

system_variable_get(key, default_value ?)

Returns the variable from the system shared key-value storage keyed with a key value, optionally if value is not present, and default expression is provided, sets a new value to be associated with that key

system_variable_set(key, new_value)

Returns the variable from the system shared key-value storage keyed with a key value, and sets a new mapping for the key# Loops, and higher order functions

Efficient use of these functions can greatly simplify your programs and speed them up, as these functions will internalize most of the operations that need to be applied on multiple values at the same time. Most of them take a list argument which can be any iterable structure in scarpet, including generators, like rect, or range, and maps, where the iterator returns all the map keys

Loops

break(), break(expr), continue(), continue(expr)

These allow to control execution of a loop either skipping current iteration code, using continue, or finishing the current loop, using break. break and continue can only be used inside for, c_for, while, loop, map, filter, reduce as well as Minecraft API block loops, scan and volume functions, while break can be used in first as well. Outside of the internal expressions of these functions, calling break or continue will cause an error. In case of the nested loops, and more complex setups, use custom try and throw setup.

Please check corresponding loop function description what continue and break do in their contexts, but in general case, passed values to break and continue will be used in place of the return value of the internal iteration expression.

c_for(init, condition, increment, body)

c_for Mimics c-style tri-arg (plus body) for loops. Return value of c_for is number of iterations performed in the loop. Unlike other loops, the body is not provided with pre-initialized _ style variables - all initialization and increments has to be handled by the programmers themselves. break and continue statements are handled within body expression only, and not in condition or increment.

 c_for(x=0, x<10, x+=1,
    c_for(y=0, y<10, y+=1,
        print(str('%d * %d = %d', x, y, x*y))
    )
 )
 

for(list,expr(_,_i))

Evaluates expression over list of items from the list. Supplies _(value) and _i(iteration number) to the expr.

Returns the number of times expr was successful. Uses continue and break argument in place of the returned value from the expr(if supplied), to determine if the iteration was successful.

check_prime(n) -> !first( range(2, sqrt(n)+1), !(n % _) );
for(range(1000000,1100000),check_prime(_))  => 7216

From which we can learn that there is 7216 primes between 1M and 1.1M

while(cond, limit)

Evaluates expression expr repeatedly until condition cond becomes false, but not more than limit times. Returns the result of the last expr evaluation, or null if nothing was successful. Both expr and cond will recveived a bound variable _ indicating current iteration, so its a number.

while(a<100,10,a=_*_)  => 81 // loop exhausted via limit
while(a<100,20,a=_*_)  => 100 // loop stopped at condition, but a has already been assigned
while(_*_<100,20,a=_*_)  => 81 // loop stopped at condition, before a was assigned a value

loop(num,expr(_),exit(_)?)

Evaluates expression expr, num number of times. code>expr receives _ system variable indicating the iteration.

loop(5, game_tick())  => repeat tick 5 times
list = l(); loop(5, x = _; loop(5, list += l(x, _) ) ); list
  // double loop, produces: [[0, 0], [0, 1], [0, 2], [0, 3], [0, 4], [1, 0], [1, 1], ... , [4, 2], [4, 3], [4, 4]]

In this small example we will search for first 10 primes, apparently including 0:

check_prime(n) -> !first( range(2, sqrt(n)+1), !(n % _) );
primes = l();
loop(10000, if(check_prime(_), primes += _ ; if (length(primes) >= 10, break())));
primes
// outputs: [0, 1, 2, 3, 5, 7, 11, 13, 17, 19]

Higher Order Functions

map(list,expr(_,_i))

Converts a list of values, to another list where each value is result of an expression v = expr(_, _i) where _ is passed as each element of the list, and _i is the index of such element. If break is called the map returns whatever collected thus far. If continue and break are used with supplied argument, it is used in place of the resulting map element, otherwise current element is skipped.

map(range(10), _*_)  => [0, 1, 4, 9, 16, 25, 36, 49, 64, 81]
map(players('*'), _+' is stoopid') [gnembon is stoopid, herobrine is stoopid]

filter(list,expr(_,_i))

filters list elements returning only these that return positive result of the expr. With break and continue statements, the supplied value can be used as a boolean check instead.

filter(range(100), !(_%5), _*_>1000)  => [0, 5, 10, 15, 20, 25, 30]
map(filter(entity_list('*'),_=='Witch'), query(_,'pos') )  => [[1082.5, 57, 1243.5]]

first(list,expr(_,_i))

Finds and returns the first item in the list that satisfies expr. It sets _ for current element value, and _i for index of that element. break can be called inside the iteration code, using its argument value instead of the current item. continue has no sense and cannot be called inside first call.

first(range(1000,10000), n=_; !first( range(2, sqrt(n)+1), !(n % _) ) )  => 1009 // first prime after 1000

Notice in the example above, that we needed to rename the outer _ to be albe to use in in the inner first call

all(list,expr(_,_i))

Returns true if all elements on the list satisfy the condition. Its roughly equivalent to all(list,expr) <=> for(list,expr)==length(list). expr also receives bound _ and _i variables. break and continue have no sense and cannot be used inside of expr body.

all([1,2,3], check_prime(_))  => 1
all(neighbours(x,y,z), _=='stone')  => 1 // if all neighbours of [x, y, z] are stone
map(filter(rect(0,4,0,1000,0,1000), l(x,y,z)=pos(_); all(rect(x,y,z,1,0,1),_=='bedrock') ), pos(_) )
  => [[-298, 4, -703], [-287, 4, -156], [-269, 4, 104], [242, 4, 250], [-159, 4, 335], [-208, 4, 416], [-510, 4, 546], [376, 4, 806]]
    // find all 3x3 bedrock structures in the top bedrock layer
map( filter( rect(0,4,0,1000,1,1000,1000,0,1000), l(x,y,z)=pos(_);
        all(rect(x,y,z,1,0,1),_=='bedrock') && for(rect(x,y-1,z,1,1,1,1,0,1),_=='bedrock')<8),
   pos(_) )  => [[343, 3, -642], [153, 3, -285], [674, 3, 167], [-710, 3, 398]]
    // ditto, but requiring at most 7 bedrock block in the 18 blocks below them

reduce(list,expr(_a,_,_i), initial)

Applies expr for each element of the list and saves the result in _a accumulator. Consecutive calls to expr can access that value to apply more values. You also need to specify the initial value to apply for the accumulator. break can be used to terminate reduction prematurely. If a value is provided to break or continue, it will be used from now on as a new value for the accumulator.

reduce([1,2,3,4],_a+_,0)  => 10
reduce([1,2,3,4],_a*_,1)  => 24

# User-defined functions and program control flow

Writing programs with more than 1 line

Operator ;

To effectively write programs that have more than one line, a programmer needs way to specify a sequence of commands that execute one after another. In scarpet this can be achieved with ;. Its an operator, and by separating statements with semicolons. And since whitespaces and $ (commandline visible newline separator) sign are all treats as whitespaces, how you layout your code doesn't matter, as long as it is readable to everyone involved.

expr;
expr;
expr;
expr

Notice that the last expression is not followed by a semicolon. Since instruction separation is functional in scarpet, and not barely an instruction delimiter, terminating the code with a dangling operator wouldn't be valid. Having said that, since many programming languages don't care about the number of op terminators programmers use, carpet preprocessor will remove all unnecessary semicolons from scripts when compiled.

In general expr; expr; expr; expr is equivalent to (((expr ; expr) ; expr) ; expr).

Result of the evaluated expression is the same as the result of the second expression, but first expression is also evaluated for side-effects

expr1 ; expr2 => expr2  // with expr1 as a side-effect

Global variables

All defined functions are compiled, stored persistently, and available globally - accessible to all other scripts. Functions can only be undefined via call to undef('fun'), which would erase global entry for function fun. Since all variables have local scope inside each function, one way to share large objects is via global variables

Any variable that is used with a name that starts with 'global_' will be stored and accessible globally, not, inside current scope. It will also persist across scripts, so if a procedure needs to use its own construct, it needs to define it, or initialize it explicitly, or undefine it via undef

a() -> global_list+=1; global_list = l(1,2,3); a(); a(); global_list  // => [1,2,3,1,1]

Operator ->

To organize code better than a flat sequence of operations, one can define functions. Definition is correct if has the following form

fun(args, ...) -> expr

Where fun(args, ...) is a function signature indicating function name, number of arguments, and their names, and expr is an expression (can be complex) that is evaluated when fun is called. Names in the signature don't need to be used anywhere else, other occurrences of these names will be masked in this function scope. Function call creates new scope for variables inside expr, so all non-global variables are not visible from the caller scope. All parameters are passed by value to the new scope, including lists and other containers, however their copy will be shallow.

The function returns its name as a string, which means it can be used to call it later with the call function

Using _ as the function name creates anonymous function, so each time _ function is defined, it will be given a unique name, which you can pass somewhere else to get this function called.

a(lst) -> lst+=1; list = l(1,2,3); a(list); a(list); list  // => [1,2,3]

In case the inner function wants to operate and modify larger objects, lists from the outer scope, but not global, it needs to use outer function in function signature

outer(arg)

outer function can only be used in the function signature, and it will cause an error everywhere else. It saves the value of that variable from the outer scope and allows its use in the inner scope. This is a similar behaviour to using outer variables in lambda function definitions from Java, except here you have to specify which variables you want to use, and borrow

This mechanism can be used to use static mutable objects without the need of using global_... variables

list = l(1,2,3); a(outer(list)) -> list+=1;  a(); a(); list  // => [1,2,3,1,1]

The return value of a function is the value of the last expression. This as the same effect as using outer or global lists, but is more expensive

a(lst) -> lst+=1; list = l(1,2,3); list=a(list); list=a(list); list  // => [1,2,3,1,1]

Ability to combine more statements into one expression, with functions, passing parameters, and global and outer scoping allow to organize even larger scripts

import(module_name, symbols ...)

Imports symbols from other apps and libraries into the current one: global variables or functions, allowing to use them in the current app. This include other symbols imported by these modules. Scarpet supports cicular dependencies, but if symbols are used directly in the module body rather than functions, it may not be able to retrieve them. Returns full list of available symbols that could be imported from this module, which can be used to debug import issues, and list contents of libraries.

call(function, args.....)

calls a user defined function with specified arguments. It is equivalent to calling function(args...) directly except you can use it with function value, or name instead. This means you can pass functions to other user defined functions as arguments and call them with call internally. And since function definitions return the defined function, they can be defined in place as anonymous functions.

Little technical note: the use of _ in expression passed to built in functions is much more efficient due to not creating new call stacks for each invoked function, but anonymous functions is the only mechanism available for programmers with their own lambda arguments

my_map(list, function) -> map(list, call(function, _));
my_map(l(1,2,3), _(x) -> x*x);    // => [1,4,9]
profile_expr(my_map(l(1,2,3), _(x) -> x*x));   // => ~32000
sq(x) -> x*x; profile_expr(my_map(l(1,2,3), 'sq'));   // => ~36000
sq = (_(x) -> x*x); profile_expr(my_map(l(1,2,3), sq));   // => ~36000
profile_expr(map(l(1,2,3), _*_));   // => ~80000

Control flow

return(expr?)

Sometimes its convenient to break the organized control flow, or it is not practical to pass the final result value of a function to the last statement, in this case a return statement can be used

If no argument is provided - returns null value.

def() -> (
   expr1;
   expr2;
   return(expr3); // function terminates returning expr3
   expr4;     // skipped
   expr5      // skipped
)

In general its cheaper to leave the last expression as a return value, rather than calling returns everywhere, but it would often lead to a messy code.

exit(expr?)

It terminates entire program passing expr as the result of the program execution, or null if omitted.

try(expr, catch_expr(_)?) ... throw(value?)

try function evaluates expression, and continues further unless throw function is called anywhere inside expr. In that case the catch_expr is evaluates with _ set to the argument throw was called with. This mechanic accepts skipping thrown value - it throws null instead, and catch expression - then try returns null as well This mechanism allows to terminate large portion of a convoluted call stack and continue program execution. There is only one level of exceptions currently in carpet, so if the inner function also defines the try catchment area, it will received the exception first, but it can technically rethrow the value its getting for the outer scope. Unhandled throw acts like an exit statement.

if(cond, expr, cond?, expr?, ..., default?)

If statement is a function that takes a number of conditions that are evaluated one after another and if any of them turns out true, its expr gets returned, otherwise, if all conditions fail, the return value is default expression, or null if default is skipped

if function is equivalent to if (cond) expr; else if (cond) expr; else default; from Java, just in a functional form

Lists, Maps and API support for Containers

Scarpet supports basic container types: lists and maps (aka hashmaps, dicts etc..)

Container manipulation

Here is a list of operations that work on all types of containers: lists, maps, as well as other Minecraft specific modifyable containers, like NBTs

get(container, address, ...), get(lvalue), ':' operator

Returns the value at address element from the value. For lists it indicates an index, use negative numbers to reach elements from the end of the list. get call will always be able to find the index. In case there is few items, it will loop over

for maps, retrieves the value under the key specified in the address or null otherwise

[Minecraft specific usecase]: In case value is of nbt type, uses address as the nbt path to query, returning null, if path is not found, one value if there was one match, or list of values if result is a list. Returned elements can be of numerical type, string texts, or another compound nbt tags

In case to simplify the access with nested objects, you can add chain of addresses to the arguments of get rather than calling it multiple times. get(get(foo,a),b) is equivalent to get(foo, a, b), or foo:a:b.

get(l(range(10)), 5)  => 5
get(l(range(10)), -1)  => 9
get(l(range(10)), 10)  => 0
l(range(10)):93  => 3
get(player() ~ 'nbt', 'Health') => 20 // inefficient way to get player health, use player() ~ 'health' instead
get(m( l('foo',2), l('bar',3), l('baz',4) ), 'bar')  => 3

has(container, address, ...), has(lvalue)

Similar to get, but returns boolean value indicating if the given index / key / path is in the container. Can be used to determine if get(...)==null means the element doesn't exist, or the stored value for this address is null, and is cheaper to run than get.

Like get, it can accept multiple addresses for chains in nested containers. In this case has(foo:a:b) is equivalent to has(get(foo,a), b) or has(foo, a, b)

delete(container, address, ...), delete(lvalue)

Removes specific entry from the container. For the lists - removes the element and shrinks it. For maps, it removes the key from the map, and for nbt - removes content from a given path. For lists and maps returns previous entry at the address, for nbt's - number of removed objects, with 0 indicating that the original value was unaffected.

Like with the get and has, delete can accept chained addresses, as well as l-value container access, removing the value from the leaf of the path provided, so delete(foo, a, b) is the same as delete(get(foo,a),b) or delete(foo:a:b)

Returns true, if container was changed, false, if it was left unchanged, and null if operation was invalid.

put(container, address, value), put(container, address, value, mode), put(lvalue, value)

Lists

Modifies the container by replacing the value under the address with the supplied value. For lists, a valid index is required, but can be negative as well to indicate positions from the end of the list. If null is supplied as the address, it always means - add to the end of the list.

There are three modes that lists can have items added to them:

• replace(default): Replaces item under given index(address). Doesn't change the size of the array unless null address is used, which is an exception and then it appends to the end
• insert: Inserts given element at a specified index, shifting the rest of the array to make space for the item. Note that index of -1 points to the last element of the list, thus inserting at that position and moving the previous last element to the new last element position. To insert at the end, use += operator, or null address in put
• extend: treats the supplied value as an iterable set of values to insert at a given index, extending the list by this amount of items. Again use null address/index to point to the end of the list

Due to the extra mode parameter, there is no chaining for put, but you can still use l-value container access to indicate container and address, so put(foo, key, value) is the same as put(foo:key, value) or foo:key=value

Returns true, if container got modified, false otherwise, and null if operation was invalid.

Maps

For maps there are no modes available (yet, seems there is no reason to). It replaces the value under the supplied key (address), or sets it if not currently present.

NBT Tags

The address for nbt values is a valid nbt path that you would use with /data command, and tag is any tag that would be applicable for a given insert operation. Note that to distinguish between proper types (like integer types, you need to use command notation, i.e. regular ints is 123, while byte size int would be 123b and an explicit string would be "5", so it helps that scarpet uses single quotes in his strings. Unlike for lists and maps, it returns the number of affected nodes, or 0 if none were affected.

There are three modes that NBT tags can have items added to them:

• replace(default): Replaces item under given path(address). Removes them first if possible, and then adds given element to the supplied position. The target path can indicate compound tag keys, lists, or individual elements of the lists.
• <N>: Index for list insertions. Inserts given element at a specified index, inside a list specified with the path address. Fails if list is not specified. It behaves like insert mode for lists, i.e. it is not removing any of the existing elements. Use replace to remove and replace existing element.
• merge: assumes that both path and replacement target are of compound type (dictionaries, maps, {} types), and merges keys from value with the compound tag under the path

a = l(1, 2, 3); put(a, 1, 4); a  => [1, 4, 3]
a = l(1, 2, 3); put(a, null, 4); a  => [1, 2, 3, 4]
a = l(1, 2, 3); put(a, 1, 4, 'insert'); a  => [1, 4, 2, 3]
a = l(1, 2, 3); put(a, null, l(4, 5, 6), 'extend'); a  => [1, 2, 3, 4, 5, 6]
a = l(1, 2, 3); put(a, 1, l(4, 5, 6), 'extend'); a  => [1, 4, 5, 6, 2, 3]
a = l(l(0,0,0),l(0,0,0),l(0,0,0)); put(a.1, 1, 1); a  => [[0, 0, 0], [0, 1, 0], [0, 0, 0]]
a = m(1,2,3,4); put(a, 5, null); a  => {1: null, 2: null, 3: null, 4: null, 5: null}
tag = nbt('{}'); put(tag, 'BlockData.Properties', '[1,2,3,4]'); tag  => {BlockData:{Properties:[1,2,3,4]}}
tag = nbt('{a:[{lvl:3},{lvl:5},{lvl:2}]}'); put(tag, 'a[].lvl', 1); tag  => {a:[{lvl:1},{lvl:1},{lvl:1}]}
tag = nbt('{a:[{lvl:[1,2,3]},{lvl:[3,2,1]},{lvl:[4,5,6]}]}'); put(tag, 'a[].lvl', 1, 2); tag
     => {a:[{lvl:[1,2,1,3]},{lvl:[3,2,1,1]},{lvl:[4,5,1,6]}]}
tag = nbt('{a:[{lvl:[1,2,3]},{lvl:[3,2,1]},{lvl:[4,5,6]}]}'); put(tag, 'a[].lvl[1]', 1); tag
     => {a:[{lvl:[1,1,3]},{lvl:[3,1,1]},{lvl:[4,1,6]}]}

List operations

l(values ...), l(iterator)

Creates a list of values of the expressions passed as parameters. It can be used as an L-value and if all elements are variables, you coujld use it to return multiple results from one function call, if that function returns a list of results with the same size as the l call uses. In case there is only one argument and it is an iterator (vanilla expression specification has range, but Minecraft API implements a bunch of them, like diamond), it will convert it to a proper list. Iterators can only be used in high order functions, and are treated as empty lists, unless unrolled with l

l(1,2,'foo') => [1, 2, foo]
l() => [] (empty list)
l(range(10)) => [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
l(1, 2) = l(3, 4) => Error: l is not a variable
l(foo, bar) = l(3,4); foo==3 && bar==4 => 1
l(foo, bar, baz) = l(2, 4, 6); l(min(foo, bar), baz) = l(3, 5); l(foo, bar, baz)  => [3, 4, 5]

In the last example l(min(foo, bar), baz) creates a valid L-value, as min(foo, bar) finds the lower of the variables (in this case foo) creating a valid assignable L-list of [foo, baz], and these values will be assigned new values

join(delim, list), join(delim, values ...)

Returns a string that contains joined elements of the list, iterator, or all values, concatenated with delim delimiter

join('-',range(10))  => 0-1-2-3-4-5-6-7-8-9
join('-','foo')  => foo
join('-', 'foo', 'bar')  => foo-bar

split(delim, expr)

Splits a string undr expr by delim which can be a regular expression

split('',foo)  => [f, o, o]
split('.','foo.bar')  => []
split('\\.','foo.bar')  => [foo, bar]

slice(expr, from, to?)

extracts a substring, or sublist (based on the type of the result of the expression under expr with starting index of from, and ending at to if provided, or the end, if omitted

slice(l(0,1,2,3,4,5), 1, 3)  => [1, 2, 3]
slice('foobar', 0, 1)  => 'f'
slice('foobar', 3)  => 'bar'
slice(range(10), 3, 5)  => [3, 4, 5]
slice(range(10), 5)  => [5, 6, 7, 8, 9]

sort(list), sort(values ...)

Sorts in the default sortographical order either all arguments, or a list if its the only argument. It returns a new sorted list, not affecting the list passed to the argument

sort(3,2,1)  => [1, 2, 3]
sort('a',3,11,1)  => [1, 3, 11, 'a']
list = l(4,3,2,1); sort(list)  => [1, 2, 3, 4]

sort_key(list, key_expr)

Sorts a copy of the list in the order or keys as defined by the key_expr for each element

sort_key([1,3,2],_)  => [1, 2, 3]
sort_key([1,3,2],-_)  => [3, 2, 1]
sort_key(l(range(10)),rand(1))  => [1, 0, 9, 6, 8, 2, 4, 5, 7, 3]
sort_key(l(range(20)),str(_))  => [0, 1, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 2, 3, 4, 5, 6, 7, 8, 9]

range(to), range(from, to), range(from, to, step)

Creates a range of numbers from from, no greater/larger than to. The step parameter dictates not only the increment size, but also direction (can be negative). The returned value is not a proper list, just the iterator but if for whatever reason you need a proper list with all items evaluated, use l(range(to)). Primarily to be used in higher order functions

range(10)  => [...]
l(range(10))  => [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
map(range(10),_*_)  => [0, 1, 4, 9, 16, 25, 36, 49, 64, 81]
reduce(range(10),_a+_, 0)  => 45
range(5,10)  => [5, 6, 7, 8, 9]
range(20, 10, -2)  => [20, 18, 16, 14, 12]

element(list, index)(deprecated)

Legacy support for older method that worked only on lists. Please use get(...) for equivalent support, or . operator. Also previous unique behaviours with put on lists has been removed to support all type of container types.

Map operations

Scarpet supports map structures, aka hashmaps, dicts etc. Map structure can also be used, with null values as sets. Apart from container access functions, (. , get, put, has, delete), the following functions:

m(values ...), m(iterator), m(key_value_pairs)

creates and initializes a map with supplied keys, and values. If the arguments contains a flat list, these are all treated as keys with no value, same goes with the iterator - creates a map that behaves like a set. If the arguments is a list of lists, they have to have two elements each, and then first is a key, and second, a value

m(1,2,'foo') => {1: null, 2: null, foo: null}
m() => {} (empty map)
m(range(10)) => {0: null, 1: null, 2: null, 3: null, 4: null, 5: null, 6: null, 7: null, 8: null, 9: null}
m(l(1, 2), l(3, 4)) => {1: 2, 3: 4}
reduce(range(10), put(_a, _, _*_); _a, m())
     => {0: 0, 1: 1, 2: 4, 3: 9, 4: 16, 5: 25, 6: 36, 7: 49, 8: 64, 9: 81}

keys(map), values(map), pairs(map)

Returns full lists of keys, values and key-value pairs (2-element lists) for all the entries in the map

Minecraft specific API and scarpet language add-ons and commands

Here is the gist of the Minecraft related functions. Otherwise the CarpetScript could live without Minecraft.

Dimension warning

One note, which is important is that most of the calls for entities and blocks would refer to the current dimension of the caller, meaning, that if we for example list all the players using player('all') function, if a player is in the other dimension, calls to entities and blocks around that player would point incorrectly. Moreover, running commandblocks in the spawn chunks would mean that commands will always refer to the overworld blocks and entities. In case you would want to run commands across all dimensions, just run three of them, using /execute in overworld/the_nether/the_end run script run ... and query players using player('*'), which only returns players in current dimension, or use in_dimension(expr) function.# Blocks / World API

Specifying blocks

block(x, y, z), block(l(x,y,z)), block(state)

Returns either a block from specified location, or block with a specific state (as used by /setblock command), so allowing for block properties, block entity data etc. Blocks otherwise can be referenced everywhere by its simple string name, but its only used in its default state.

block('air')  => air
block('iron_trapdoor[half=top]')  => iron_trapdoor
block(0,0,0) == block('bedrock')  => 1
block('hopper[facing=north]{Items:[{Slot:1b,id:"minecraft:slime_ball",Count:16b}]}') => hopper

Retrieving a block with block function has also a side-effect of evaluating its current state and data. So if you use it later it will reflect block state and data of the block that was when block was called, rather than when it was used. Block values passed in various places like scan functions, etc, are not fully evaluated unless its properties are needed. This means that if the block at the location changes before its queried in the program this might result in getting the later state, which might not be desired. Consider the following example:

set(10,10,10,'stone');
scan(10,10,10,0,0,0, b = _);
set(10,10,10,'air');
print(b); // 'air', block was remembered 'lazily', and evaluated by `print`, when it was already set to air
set(10,10,10,'stone');
scan(10,10,10,0,0,0, b = block(_));
set(10,10,10,'air');
print(b); // 'stone', block was evaluated 'eagerly' but call to `block`

World Manipulation

All the functions below can be used with block value, queried with coord triple, or 3-long list. All pos in the functions referenced below refer to either method of passing block position.

set(pos, block, property?, value?, ...)

First part of the set function is either a coord triple, list of three numbers, or other block with coordinates. Second part, block is either block value as a result of block() function string value indicating the block name, and optional property - value pairs for extra block properties. If block is specified only by name, then if a destination block is the same the set operation is skipped, otherwise is executed, for other potential extra properties.

The returned value is either the block state that has been set, or false if block setting was skipped

set(0,5,0,'bedrock')  => bedrock
set(l(0,5,0), 'bedrock')  => bedrock
set(block(0,5,0), 'bedrock')  => bedrock
scan(0,5,0,0,0,0,set(_,'bedrock'))  => 1
set(pos(players()), 'bedrock')  => bedrock
set(0,0,0,'bedrock')  => 0   // or 1 in overworlds generated in 1.8 and before
scan(0,100,0,20,20,20,set(_,'glass'))
    // filling the area with glass
scan(0,100,0,20,20,20,set(_,block('glass')))
    // little bit faster due to internal caching of block state selectors
b = block('glass'); scan(0,100,0,20,20,20,set(_,b))
    // yet another option, skips all parsing
set(x,y,z,'iron_trapdoor')  // sets bottom iron trapdoor
set(x,y,z,'iron_trapdoor[half=top]')  // Incorrect. sets bottom iron trapdoor - no parsing of properties
set(x,y,z,'iron_trapdoor','half','top') // correct - top trapdoor
set(x,y,z, block('iron_trapdoor[half=top]')) // also correct, block() provides extra parsing
set(x,y,z,'hopper[facing=north]{Items:[{Slot:1b,id:"minecraft:slime_ball",Count:16b}]}') // extra block data

without_updates(expr)

Evaluates subexpression without causing updates when blocks change in the world

Consider following scenario: We would like to generate a bunch of terrain in a flat world following a perling noise generator. The following code causes cascading effect as blocks placed on chunk borders will cause other chunks to get loaded to full, thus generated:

__config() -> m(l('scope', 'global'));
__on_chunk_generated(x, z) -> (
  scan(x,0,z,0,0,0,15,15,15,
    if (perlin(_x/16, _y/8, _z/16) > _y/16,
      set(_, 'black_stained_glass');
    )
  )
)

The following addition resolves this issue, by not allowing block updates pass chunk borders:

__config() -> m(l('scope', 'global'));
__on_chunk_generated(x, z) -> (
  without_updates(
    scan(x,0,z,0,0,0,15,15,15,
      if (perlin(_x/16, _y/8, _z/16) > _y/16,
        set(_, 'black_stained_glass');
      )
    )
  )
)

place_item(item, pos, facing?, sneak?)

Places a given item in the world like it was placed by a player. Item names are default minecraft item name, less the minecraft prefix. Default facing is 'up', but there are other options: 'down', 'north', 'east', 'south', 'west', but also there are other secondary directions important for placement of blocks like stairs, doors, etc. Try experiment with options like 'north-up' which is placed facing north with cursor pointing to the upper part of the block, or 'up-north', which means a block placed facing up (player looking down) and placed smidge away of the block center towards north. Optional sneak is a boolean indicating if a player would be sneaking while placing the block - this option only affects placement of chests and scaffolding at the moment. Returns true if placement was successful, false otherwise.

place_item('stone',x,y,z) // places a stone block on x,y,z block
place_item('piston,x,y,z,'down') // places a piston facing down
place_item('carrot',x,y,z) // attempts to plant a carrot plant. Returns true if could place carrots at that position.

set_poi(pos, type, occupancy?)

Sets a Point of Interest (POI) of a specified type with optional custom occupancy. By default new POIs are not occupied. If type is null, POI at position is removed. In any case, previous POI is also removed. Available POI types are:

• 'unemployed', 'armorer', 'butcher', 'cartographer', 'cleric', 'farmer', 'fisherman', 'fletcher', 'leatherworker', 'librarian', 'mason', 'nitwit', 'shepherd', 'toolsmith', 'weaponsmith', 'home', 'meeting', 'beehive', 'bee_nest', 'nether_portal'

Interestingly, unemployed, and nitwit are not used in the game, meaning, they could be used as permanent spatial markers for scarpet apps. meeting is the only one with increased max occupancy of 32.

set_biome(pos, biome_name)

changes biome at that block position.

update(pos)

Causes a block update at position.

block_tick(pos)

Causes a block to tick at position.

random_tick(pos)

Causes a random tick at position.

destroy(pos), destroy(pos, -1), destroy(pos, <N>), destroy(pos, tool, nbt?)

Destroys the block like it was mined by a player. Add -1 for silk touch, and positive number for fortune level. If tool is specified, and optionally its nbt, it will use that tool and will attempt to mine the block with this tool. If called without item context, this function, unlike harvest, will affect all kinds of blocks. If called with item in context, it will fail to break blocks that cannot be broken by a survival player.

Without item context it returns false if failed to destroy the block and true if block breaking was successful. In item context, true means that breaking item has no nbt to use, null indicating that the tool should be considered broken in process, and nbt type value, for a resulting NBT tag on a hypothetical tool. Its up to the programmer to use that nbt to apply it where it belong

Here is a sample code that can be used to mine blocks using items in player inventory, without using player context for mining. Obviously, in this case the use of harvest would be much more applicable:

mine(x,y,z) ->
(
   p = player();
   slot = p~'selected_slot';
   item_tuple = inventory_get(p, slot);
   if (!item_tuple, destroy(x,y,z,'air'); return()); // empty hand, just break with 'air'
   l(item, count, tag) = item_tuple;
   tag_back = destroy(x,y,z, item, tag);
   if (tag_back == false, // failed to break the item
	     return(tag_back)
   );
   if (tag_back == true, // block broke, tool has no tag
	     return(tag_back)
   );
   if (tag_back == null, //item broke
	     delete(tag:'Damage');
	     inventory_set(p, slot, count-1, item, tag);
	     return(tag_back)
   );
   if (type(tag_back) == 'nbt', // item didn't break, here is the effective nbt
	     inventory_set(p, slot, count, item, tag_back);
	     return(tag_back)
   );
   print('How did we get here?');
)

harvest(player, pos)

Causes a block to be harvested by a specified player entity. Honors player item enchantments, as well as damages the tool if applicable. If the entity is not a valid player, no block gets destroyed. If a player is not allowed to break that block, a block doesn't get destroyed either.

Block and World querying

pos(block), pos(entity)

Returns a triple of coordinates of a specified block or entity. Technically entities are queried with query function and the same can be achieved with query(entity,'pos'), but for simplicity pos allows to pass all positional objects.

pos(block(0,5,0))  => l(0,5,0)
pos(players()) => l(12.3, 45.6, 32.05)
pos(block('stone'))  => Error: Cannot fetch position of an unrealized block

pos_offset(pos, direction, amount?)

Returns a coords triple that is offset in a specified direction by amount of blocks. The default offset amount is 1 block. To offset into opposite facing, use negative numbers for the amount.

pos_offset(block(0,5,0), 'up', 2)  => l(0,7,0)
pos_offset(l(0,5,0), 'up', -2 ) => l(0,3,0)

block_properties(pos)

Returns a list of available block properties for a particular block. If a block has no properties, returns an empty list.

property(pos, name)

Returns property of block at pos, or specified by block argument. If a block doesn't have that property, null value is returned. Returned values are always strings. It is expected from the user to know what to expect and convert values to numbers using number() function or booleans using bool() function.

set(x,y,z,'iron_trapdoor','half','top'); property(x,y,z,'half')  => top
set(x,y,z,'air'); property(x,y,z,'half')  => null
property(block('iron_trapdoor[half=top]'),'half')  => top
property(block('iron_trapdoor[half=top]'),'powered')  => false
bool(property(block('iron_trapdoor[half=top]'),'powered'))  => 0

block_data(pos)

Return NBT string associated with specific location, or null if the block does not carry block data. Can be currently used to match specific information from it, or use it to copy to another block

    block_data(x,y,z) => '{TransferCooldown:0,x:450,y:68, ... }'

poi(pos), poi(pos, radius), poi(pos, radius, status)

Queries a POI (Point of Interest) at a given position, returning null if none is found, or tuple of poi type and its occupancy load. With optional radius and status, returns a list of POIs around pos within a given radius. If status is specified (either 'available', or 'occupied') returns only POIs with that status. The return format is again, poi type, occupancy load, and extra tripple of coordinates.

Querying for POIs using the radius is intended use of POI mechanics and ability of accessing individual POIs via poi(pos) in only provided for completness.

poi(x,y,z) => null  // nothing set at position
poi(x,y,z) => ['meeting',3]  // its a bell-type meeting point occupied by 3 villagers
poi(x,y,z,5) => []  // nothing around
poi(x,y,z,5) => [['nether_portal',0,[7,8,9]],['nether_portal',0,[7,9,9]]] // two portal blocks in the range

biome(pos)

returns biome at that block position.

solid(pos)

Boolean function, true if the block is solid

air(pos)

Boolean function, true if a block is air.... or cave air... or void air.... or any other air they come up with.

liquid(pos)

Boolean function, true if the block is liquid, or liquidlogged

flammable(pos)

Boolean function, true if the block is flammable

transparent(pos)

Boolean function, true if the block is transparent

opacity(pos)

Numeric, returning opacity level of a block

blocks_daylight(pos)

Boolean function, true if the block blocks daylight

emitted_light(pos)

Numeric, returning light level emitted from block

light(pos)

Integer function, returning total light level at position

block_light(pos)

Integer function, returning block light at position. From torches and other light sources.

sky_light(pos)

Numeric function, returning sky light at position. From the sky access.

see_sky(pos)

Boolean function, returning true if the block can see sky.

hardness(pos)

Numeric function, indicating hardness of a block.

blast_resistance(pos)

Numeric function, indicating blast_resistance of a block.

in_slime_chunk(pos)

Boolean indicating if the given block position is in a slime chunk.

top(type, pos)

Returns the Y value of the topmost block at given x, z coords (y value of a block is not important), according to the heightmap specified by type. Valid options are:

• light: topmost light blocking block (1.13 only)
• motion: topmost motion blocking block
• terrain: topmost motion blocking block except leaves
• ocean_floor: topmost non-water block
• surface: topmost surface block

top('motion', x, y, z)  => 63
top('ocean_floor', x, y, z)  => 41

loaded(pos)

Boolean function, true if the block is accessible for the game mechanics. Normally scarpet doesn't check if operates on loaded area - the game will automatically load missing blocks. We see this as advantage. Vanilla fill/clone commands only check the specified corners for loadness.

To check if block is truly loaded, I mean in memory, use generation_status(x) != null, as chunks can still be loaded outside of the playable area, just are not used any of the game mechanics processes.

loaded(pos(players()))  => 1
loaded(100000,100,1000000)  => 0

(Deprecated) loaded_ep(pos)

Boolean function, true if the block is loaded and entity processing, as per 1.13.2

Deprecated as of scarpet 1.6, use loaded_status(x) > 0, or just loaded(x) with the same effect

loaded_status(pos)

Returns loaded status as per new 1.14 chunk ticket system, 0 for inaccessible, 1 for border chunk, 2 for ticking, 3 for entity ticking

generation_status(pos), generation_status(pos, true)

Returns generation status as per new 1.14 chunk ticket system. Can return any value from several available but chunks can only be valid in a few states: full, features, liquid_carvers, and structure_starts. Returns null if the chunk is not in memory unless called with optional true.

structures(pos), structures(pos, structure_name)

Returns structure information for a given block position. Note that structure information is the same for all the blocks from the same chunk. structures function can be called with a block, or a block and a structure name. In the first case it returns a map of structures at a given position, keyed by structure name, with values indicating the bounding box of the structure - a pair of two 3-value coords (see examples). When called with an extra structure name, returns list of components for that structure, with their name, direction and two sets of coordinates indicating the bounding box of the structure piece.

structure_references(pos), structure_references(pos, structure_name)

Returns structure information that a chunk with a given block position is part of. structure_references function can be called with a block, or a block and a structure name. In the first case it returns a list of structure names that give chunk belongs to. When called with an extra structure name, returns list of positions pointing to the lowest block position in chunks that hold structure starts for these structures. You can query that chunk structures then to get its bounding boxes.

set_structure(pos, structure_name), set_structure(pos, structure_name, null)

Creates or removes structure information of a structure associated with a chunk of pos. Unlike plop, blocks are not placed in the world, only structure information is set. For the game this is a fully functional structure even if blocks are not set. To remove structure a given point is in, use structure_references to find where current structure starts.

suffocates(pos)

Boolean function, true if the block causes suffocation.

power(pos)

Numeric function, returning redstone power level at position.

ticks_randomly(pos)

Boolean function, true if the block ticks randomly.

blocks_movement(pos)

Boolean function, true if block at position blocks movement.

block_sound(pos)

Returns the name of sound type made by the block at position. One of:

• wood
• gravel
• grass
• stone
• metal
• glass
• wool
• sand
• snow
• ladder
• anvil
• slime
• sea_grass
• coral

material(pos)

Returns the name of material of the block at position. very useful to target a group of blocks. One of:

• air
• void
• portal
• carpet
• plant
• water_plant
• vine
• sea_grass
• water
• bubble_column
• lava
• snow_layer
• fire
• redstone_bits
• cobweb
• redstone_lamp
• clay
• dirt
• grass
• packed_ice
• sand
• sponge
• wood
• wool
• tnt
• leaves
• glass
• ice
• cactus
• stone
• iron
• snow
• anvil
• barrier
• piston
• coral
• gourd
• dragon_egg
• cake

map_colour(pos)

Returns the map colour of a block at position. One of:

• air
• grass
• sand
• wool
• tnt
• ice
• iron
• foliage
• snow
• clay
• dirt
• stone
• water
• wood
• quartz
• adobe
• magenta
• light_blue
• yellow
• lime
• pink
• gray
• light_gray
• cyan
• purple
• blue
• brown
• green
• red
• black
• gold
• diamond
• lapis
• emerald
• obsidian
• netherrack
• white_terracotta
• orange_terracotta
• magenta_terracotta
• light_blue_terracotta
• yellow_terracotta
• lime_terracotta
• pink_terracotta
• gray_terracotta
• light_gray_terracotta
• cyan_terracotta
• purple_terracotta
• blue_terracotta
• brown_terracotta
• green_terracotta
• red_terracotta
• black_terracotta

Iterating over larger areas of blocks

These functions help scan larger areas of blocks without using generic loop functions, like nested loop.

scan(cx, cy, cz, dx, dy, dz, px?, py?, pz?, expr)

Evaluates expression over area of blocks defined by its center (cx, cy, cz), expanded in all directions by dx, dy, dz blocks, or optionally in negative with d coords, and p coords in positive values. expr receives _x, _y, _z as coords of current analyzed block and _, which represents the block itself.

Returns number of successful evaluations of expr (with true boolean result) unless called in void context, which would cause the expression not be evaluated for their boolean value.

scan also handles continue and break statements, using continue's return value to use in place of expression return value. break return value has no effect.

volume(x1, y1, z1, x2, y2, z2, expr)

Evaluates expression for each block in the area, the same as the scanfunction, but using two opposite corners of the rectangular cuboid. Any corners can be specified, its like you would do with /fill command.

For return value and handling break and continue statements, see scan function above.

neighbours(x, y, z), neighbours(block), neighbours(l(x,y,z))

Returns the list of 6 neighbouring blocks to the argument. Commonly used with other loop functions like for.

for(neighbours(x,y,z),air(_)) => 4 // number of air blocks around a block

rect(cx, cy, cz, dx?, dy?, dz?, px?, py?, pz?)

returns an iterator, just like range function that iterates over rectangular cubarea of blocks. If only center point is specified, it iterates over 27 blocks. If d arguments are specified, expands selection of respective number of blocks in each direction. If p arguments are specified, it uses d for negative offset, and p for positive.

diamond(cx, cy, cz, radius?, height?)

Iterates over a diamond like area of blocks. With no radius and height, its 7 blocks centered around the middle (block + neighbours). With a radius specified, it expands shape on x and z coords, and with a custom height, on y. Any of these can be zero as well. radius of 0 makes a stick, height of 0 makes a diamond shape pad.

Entity API

Entity Selection

Entities have to be fetched before using them. Entities can also change their state between calls to the script if game happens either in between separate calls to the programs, or if the program calls game_tick on its own. In this case - entities would need to be re-fetched, or the code should account for entities getting dead.

player(), player(type), player(name)

With no arguments, it returns the calling player or the player closest to the caller. Note that the main context will receive p variable pointing to this player. With type or name specified it will try first to match a type, returning a list of players matching a type, and if this fails, will assume its player name query retuning player with that name, or null if no player was found. With 'all', list of all players in the game, in all dimensions, so end user needs to be cautious, that you might be referring to wrong blocks and entities around the player in question. With type = '*' it returns all players in caller dimension, 'survival' returns all survival and adventure players, 'creative' returns all creative players, 'spectating' returns all spectating players, and '!spectating', all not-spectating players. If all fails, with name, the player in question, if is logged in.

entity_id(uuid), entity_id(id)

Fetching entities wither by their ID obtained via entity ~ 'id', which is unique for a dimension and current world run, or by UUID, obtained via entity ~ 'uuid'. It returns null if no such entity is found. Safer way to 'store' entities between calls, as missing entities will be returning null. Both calls using UUID or numerical ID are O(1), but obviously using UUIDs takes more memory and compute.

entity_list(type)

Returns global lists of entities in the current dimension of a specified type. Currently the following selectors are available:

• *: all entites
• living
• items
• players
• !players

entity_area(type, cx, cy, cz, dx, dy, dz)

Returns entities of a specified type in an area centered on cx, cy, cz and at most dx, dy, dz blocks away from the center point. Uses same selectors as entities_list

entity_selector(selector)

Returns entities satisfying given vanilla entity selector. Most complex among all the methods of selecting entities, but the most capable. Selectors are cached so should be as fast as other methods of selecting entities.

spawn(name, pos, nbt?)

Spawns and places an entity in world, like /summon vanilla command. Requires a position to spawn, and optional extra nbt data to merge with the entity. What makes it different from calling run('summon ...'), is the fact that you get the entity back as a return value, which is swell.

Entity Manipulation

Unlike with blocks, that use plethora of vastly different querying functions, entities are queried with query function and altered via modify function. Type of information needed or values to be modified are different for each call.

Using ~ (in) operator is an alias for query. Especially useful if a statement has no arguments, which in this case can be radically simplified

query(p, 'name') <=> p ~ 'name'     // much shorter and cleaner
query(p, 'holds', 'offhand') <=> p ~ l('holds', 'offhand')    // not really but can be done

query(e,'removed')

Boolean. True if the entity is removed

query(e,'id')

Returns numerical id of the entity. Most efficient way to keep track of entites in a script. Ids are only unique within current game session (ids are not preserved between restarts), and dimension (each dimension has its own ids which can overlap.

query(e,'uuid')

Returns UUID (unique id) of the entity. Can be used to access entities with the other vanilla commands and remains unique regardless of the dimension, and is preserved between game restarts. Apparently players cannot be accessed via UUID, but name instead.

map(entities_area('*',x,y,z,30,30,30),run('kill '+query(_,'id'))) // doesn't kill the player

query(e,'pos')

Triple of entity position

query(e,'location')

Quin-tuple of entity position (x, y, and z coords), and rotation (yaw, pitch)

query(e,'x'), query(e,'y'), query(e,'z')

Respective entity coordinate

query(e,'pitch'), query(e,'yaw')

Pitch and Yaw or where entity is looking.

query(e,'look')

Returns a 3d vector where the entity is looking.

query(e,'motion')

Triple of entity motion vector, l(motion_x, motion_y, motion_z)

query(e,'motion_x'), query(e,'motion_y'), query(e,'motion_z')

Respective component of the motion vector

query(e,'name'), query(e,'custom_name'), query(e,'type')

String of entity name

query(e,'name')  => Leatherworker
query(e,'custom_name')  => null
query(e,'type')  => villager

query(e,'is_riding')

Boolean. True if riding another entity.

query(e,'is_ridden')

Boolean. True if another entity is riding it.

query(e,'passengers')

List of entities riding the entity.

query(e,'mount')

Entity that e rides.

query(e,'tags')

List of entity tags.

query(e,'has_tags',tag)

Boolean, True if the entity is marked with tag.

query(e,'is_burning')

Boolean, True if the entity is burning.

query(e,'fire')

Number of remaining ticks of being on fire.

query(e,'silent')

Boolean, True if the entity is silent.

query(e,'gravity')

Boolean, True if the entity is affected by gravity, like most entities do.

query(e,'immune_to_fire')

Boolean, True if the entity is immune to fire.

query(e,'dimension')

Name of the dimension entity is in.

query(e,'height')

Height of the entity.

query(e,'width')

Width of the entity.

query(e,'eye_height')

Eye height of the entity.

query(e,'age')

Age, in ticks, of the entity, i.e. how long it existed.

query(e,'breeding_age')

Breeding age of passive entity, in ticks. If negative it it time to adulthood, if positive, breeding cooldown

query(e,'despawn_timer')

For living entities - the number of ticks they fall outside of immediate player presence.

query(e,'item')

The item triple (name, count, nbt) if its an item entity, null otherwise

query(e,'count')

Number of items in a stack from item entity, null otherwise

query(e,'pickup_delay')

Retrieves pickup delay timeout for an item entity, null otherwise

query(e,'is_baby')

Boolean, true if its a baby.

query(e,'target')

Returns mob's attack target or null if none or not applicable.

query(e,'home')

Returns creature's home position or null if none or not applicable.

query(e, 'pose')

Returns a pose of an entity, one of the following options

• 'standing'
• 'fall_flying'
• 'sleeping'
• 'swimming'
• 'spin_attack'
• 'crouching'
• 'dying'

query(e,'sneaking')

Boolean, true if entity is sneaking.

query(e,'sprinting')

Boolean, true if entity is sprinting.

query(e,'swimming')

Boolean, true if entity is swimming.

query(e,'jumping')

Boolean, true if entity is jumping.

query(e,'gamemode')

String with gamemode, or null if not a player.

query(e,'gamemode_id')

Good'ol gamemode id, or null if not a player.

query(e,'player_type')

Returns null if the argument is not a player, otherwise:

• singleplayer: for singleplayer game
• multiplayer: for players on a dedicated server
• lan_host: for singleplayer owner that opened the game to LAN
• lan_player: for all other players that connected to a LAN host
• fake: any carpet-spanwed fake player
• shadow: any carpet-shadowed real player
• realms: ?

query(e,'permission_level')

Player's permission level, or null if not applicable for this entity.

query(e,'effect',name?)

Without extra arguments, it returns list of effect active on a living entity. Each entry is a triple of short effect name, amplifier, and remaining duration. With an argument, if the living entity has not that potion active, returns null, otherwise return a tuple of amplifier and remaining duration.

query(p,'effect')  => [[haste, 0, 177], [speed, 0, 177]]
query(p,'effect','haste')  => [0, 177]
query(p,'effect','resistance')  => null

query(e,'health')

Number indicating remaining entity health, or null if not applicable.

query(e,'holds',slot?)

Returns triple of short name, stack count, and NBT of item held in slot. Available options for slot are:

• mainhand
• offhand
• head
• chest
• legs
• feet

If slot is not specified, it defaults to the main hand.

query(e,'selected_slot')

Number indicating the selected slot of entity inventory. Currently only applicable to players.

query(e,'facing', order?)

Returns where the entity is facing. optional order (number from 0 to 5, and negative), indicating primary directions where entity is looking at. From most prominent (order 0) to opposite (order 5, or -1)

query(e,'trace', reach?, options?...)

Returns the result of ray tracing from entity perspective, indicating what it is looking at. Default reach is 4.5 blocks (5 for creative players), and by default it traces for blocks and entities, identical to player attack tracing action. This can be customized with options, use 'blocks' to trace for blocks, 'liquids' to include liquid blocks as possible results, and 'entities' to trace entities. Any combination of the above is possible. When tracing entities and blocks, blocks will take over the priority even if transparent or non-colliding (aka fighting chickens in tall grass).

Regardless of the options selected, the result could be null if nothing is in reach, entity, if look targets an entity, and block value if block is in reach.

query(e,'nbt',path?)

Returns full NBT of the entity. If path is specified, it fetches only that portion of the NBT, that corresponds to the path. For specification of path attribute, consult vanilla /data get entity command.

Note that calls to nbt are considerably more expensive comparing to other calls in Minecraft API, and should only be used when there is no other option. Returned value is of type nbt, which can be further manipulated with nbt type objects via get, put, has, delete, so try to use API calls first for that.

Entity Modification

Like with entity querying, entity modifications happen through one function. Most position and movements modifications don't work currently on players as their position is controlled by clients.

Currently there is no ability to modify NBT directly, but you could always use run('data modify entity

modify(e,'remove')

Removes (not kills) entity from the game.

modify(e,'kill')

Kills the entity.

modify(e, 'pos', x, y, z), modify(e, 'pos', l(x,y,z) )

Moves the entity to a specified coords.

modify(e, 'location', x, y, z, yaw, pitch), modify(e, 'location', l(x, y, z, yaw, pitch) )

Changes full location vector all at once.

modify(e, 'x', x), modify(e, 'y', y), modify(e, 'z', z)

Moves the entity in.... one direction.

modify(e, 'pitch', pitch), modify(e, 'yaw', yaw)

Changes entity's pitch or yaw.

modify(e, 'move', x, y, z), modify(e, 'move', l(x,y,z) )

Moves the entity by a vector from its current location.

modify(e, 'motion', x, y, z), modify(e, 'motion', l(x,y,z) )

Sets the motion vector (where and how much entity is moving).

modify(e, 'motion_z', x), modify(e, 'motion_y', y), modify(e, 'motion_z', z)

Sets the corresponding component of the motion vector.

modify(e, 'accelerate', x, y, z), modify(e, 'accelerate', l(x, y, z) )

Adds a vector to the motion vector. Most realistic way to apply a force to an entity.

modify(e, 'custom_name'), modify(e, 'custom_name', name )

modify(e, 'age', number)

Modifies entity's internal age counter. Fiddling with this will affect directly behaviours of complex entities, so use it with caution.

modify(e, 'pickup_delay', number)

modify(e, 'breeding_age', number)

modify(e, 'despawn_timer', number)

Sets a custom despawn timer value.

modify(e, 'dismount')

Dismounts riding entity.

modify(e, 'mount', other)

Mounts the entity to the other.

modify(e, 'drop_passengers')

Shakes off all passengers.

modify(e, 'mount_passengers', passenger, ? ...), modify(e, 'mount_passengers', l(passengers) )

Mounts on all listed entities on e.

modify(e, 'tag', tag, ? ...), modify(e, 'tag', l(tags) )

Adds tag / tags to the entity.

modify(e, 'clear_tag', tag, ? ...), modify(e, 'clear_tag', l(tags) )

Removes tag from entity.

modify(e, 'talk')

Make noises.

modify(e, 'ai', boolean)

If called with false value, it will disable AI in the mob. true will enable it again.

modify(e, 'no_clip', boolean)

Sets if the entity obeys any collisions, including collisions with the terrain and basic physics. Not affecting players, since they are controlled client side

modify(e, 'effect', name, duration?, amplifier?, show_particles?, show_icon?)

Applies status effect to the living entity. Takes several optional parameters, which default to 0, true and true. If no duration is specified, or it is null or 0, the effect is removed.

modify(e, 'home', null), modify(e, 'home', block, distance?), modify(e, 'home', x, y, z, distance?)

Sets AI to stay around the home position, within distance blocks from it. distance defaults to 16 blocks. null removes it. May not work fully with mobs that have this AI built in, like Villagers.

Entity Events

There is a number of events that happen to entities that you can attach your own code to in the form of event handlers. The event handler is any function that runs in your package that accepts certain expected parameters, which you can expand with your own arguments. When it comes to the moment when the given command needs to be executed, it does so providing that number of arguments it accepts is equal number of event arguments, and extra arguments passed when defining the callback with entity_event

The following events can be handled by entities.

• 'on_tick': executes every tick right before entity is ticked in the game. Required arguments: entity
• 'on_death': executes once when a living entity dies. Required arguments: entity, reason
• 'on_removed': execute once when an entity is removed. Required arguments: entity
• 'on_damaged': executed every time a living entity is about to receive damage. Required arguments: entity, amount, source, attacking_entity

It doesn't mean that all entity types will have a chance to execute a given event, but entities will not error when you will attach inapplicable event to it.

entity_event(e, event, call_name, args...)

Attaches specific function from the current package to be called upon the event, with extra args curried to the original required arguments for the event handler

protect_villager(entity, amount, source, source_entity, healing_player) ->
(
   if(source_entity && source_entity~'type' != 'player',
      modify(entity, 'health', amount + entity~'health' );
      particle('end_rod', pos(entity)+l(0,3,0));
      print(str('%s healed thanks to %s', entity, healing_player))
   )
);
__on_player_interacts_with_entity(player, entity, hand) ->
(
   if (entity~'type' == 'villager',
      entity_event(entity, 'on_damage', 'protect_villager', player~'name')
   )
)

In this case this will protect a villager from entity damage (zombies, etc.) except players by granting all the health back to the villager after being harmed.

Inventory and Items API

Manipulating inventories of blocks and entities

Most functions in this category require inventory as the first argument. Inventory could be specified by an entity, or a block, or position (three coordinates) of a potential block with inventory. Player enderchest inventory require two arguments, keyword 'enderchest', followed by the player entity argument. If the entity or a block doesn't have an inventory, they typically do nothing and return null.

Most items returned are in the form of a triple of item name, count, and nbt or the extra data associated with an item. Manipulating of the nbt data can be costly, but retrieving them from the tuple to match other aspects is cheap

stack_limit(item)

Returns number indicating what is the stack limit for the item. Its typically 1 (non-stackable), 16 (like buckets), or 64 - rest. It is recommended to consult this, as other inventory API functions ignore normal stack limits, and it is up to the programmer to keep it at bay. As of 1.13, game checks for negative numbers and setting an item to negative is the same as empty.

stack_limit('wooden_axe') => 1
stack_limit('ender_pearl') => 16
stack_limit('stone') => 64

item_category(item)

Returns the string representing the category of a given item, like building_blocks, combat, or tools.

item_category('wooden_axe') => tools
item_category('ender_pearl') => misc
item_category('stone') => building_blocks

inventory_size(inventory)

Returns the size of the inventory for the entity or block in question. Returns null if the block or entity don't have an inventory

inventory_size(player()) => 41
inventory_size('enderchest', player()) => 27 // enderchest
inventory_size(x,y,z) => 27 // chest
inventory_size(block(pos)) => 5 // hopper

inventory_has_items(inventory)

Returns true, if the inventory is not empty, false if it is empty, and null, if its not an inventory.

    inventory_has_items(player()) => true
    inventory_has_items(x,y,z) => false // empty chest
    inventory_has_items(block(pos)) => null // stone

inventory_get(inventory, slot)

Returns the item in the corresponding inventory slot, or null if slot empty or inventory is invalid. You can use negative numbers to indicate slots counted from 'the back'.

inventory_get(player(), 0) => null // nothing in first hotbar slot
inventory_get(x,y,z, 5) => ['stone', 1, {}]
inventory_get(player(), -1) => ['diamond_pickaxe', 1, {Damage:4}] // slightly damaged diamond pick in the offhand

inventory_set(inventory, slot, count, item?, nbt?)

Modifies or sets a stack in inventory. specify count 0 to empty the slot. If item is not specified, keeps existing item, just modifies the count. If item is provided - replaces current item. If nbt is provided - adds a tag to the stack at slot. Returns previous stack in that slot.

inventory_set(player(), 0, 0) => ['stone', 64, {}] // player had a stack of stone in first hotbar slot
inventory_set(player(), 0, 6) => ['diamond', 64, {}] // changed stack of diamonds in player slot to 6
inventory_set(player(), 0, 1, 'diamond_axe','{Damage:5}') => null //added slightly damaged pick to first player slot

inventory_find(inventory, item, start_slot?, ), inventory_find(inventory, null, start_slot?)

Finds the first slot with a corresponding item in the inventory, or if queried with null: the first empty slot. Returns slot number if found, or null otherwise. Optional start_slot argument allows to skip all preceeding slots allowing for efficient (so not slot-by-slot) inventory search for items.

inventory_find(player(), 'stone') => 0 // player has stone in first hotbar slot
inventory_find(player(), null) => null // player's inventory has no empty spot
while( (slot = inventory_find(p, 'diamond', slot)) != null, 41, drop_item(p, slot) )
    // spits all diamonds from player inventory wherever they are
inventory_drop(x,y,z, 0) => 64 // removed and spawned in the world a full stack of items

inventory_remove(inventory, item, amount?)

Removes amount (defaults to 1) of item from inventory. If the inventory doesn't have the defined amount, nothing happens, otherwise the given amount of items is removed wherever they are in the inventory. Returns boolean whether the removal operation was successful. Easiest way to remove a specific item from player inventory without specifying the slot.

inventory_remove(player(), 'diamond') => 1 // removed diamond from player inventory
inventory_remove(player(), 'diamond', 100) => 0 // player doesn't have 100 diamonds, nothing happened

drop_item(inventory, slot, amount?, )

Drops the items from indicated inventory slot, like player that Q's an item or villager, that exchanges food. You can Q items from block inventories as well. default amount is 0 - which is all from the slot. NOTE: hoppers are quick enough to pick all the queued items from their inventory anyways. Returns size of the actual dropped items.

inventory_drop(player(), 0, 1) => 1 // Q's one item on the ground
inventory_drop(x,y,z, 0) => 64 // removed and spawned in the world a full stack of items

Scarpet events system

Provides the ability to execute specific function whenever an event occurs. The functions to be registered need to conform with the arguments to the event specification. When loading module functions, each function that starts with __on_... and has the required arguments, will be bound automatically. In case of player specific modules, all player action events will be directed to the appropriate player space, and all tick events will be executed in the global context, so its not a good idea to mix these two, so use either of these, or use commands to call tick events directly, or handle player specific data inside a module. In near future using global events (on_tick, on_chunk_generated) will be not allowed in player scoped apps.

Event list

Here is a list of events that can be handled by scarpet. This list includes prefixes required by modules to autoload them, but you can add any function to any event if it accepts required number of parameters:

__on_tick()         // can access blocks and entities in the overworld
__on_tick_nether()  // can access blocks and entities in the nether
__on_tick_ender()   // can access blocks and entities in the end
__on_chunk_generated(x,z) // called after a chunk is promoted to the full chunk,
                          // prodiving lowest x and z coords in the chunk
                          // event will not work with optifine installed in the game
// player specific callbacks
__on_player_uses_item(player, item_tuple, hand)  // right click action
__on_player_releases_item(player, item_tuple, hand)  // client action (e.g. bow)
__on_player_finishes_using_item(player, item_tuple, hand))  // server action (e.g. food), called item is from before it is used.
__on_player_clicks_block(player, block, face)  // left click attack on a block
__on_player_breaks_block(player, block) // called after block is broken (the caller receives previous blockstate)
__on_player_right_clicks_block(player, item_tuple, hand, block, face, hitvec)  // player right clicks block with anything
__on_player_interacts_with_block(player, hand, block, face, hitvec)  //right click on a block resulted in activation of said block
__on_player_places_block(player, item_tuple, hand, block) // player have just placed the block.
__on_player_interacts_with_entity(player, entity, hand)
__on_player_attacks_entity(player, entity)
__on_player_rides(player, forward, strafe, jumping, sneaking)
__on_player_jumps(player)
__on_player_deploys_elytra(player)
__on_player_wakes_up(player)
__on_player_starts_sneaking(player)
__on_player_stops_sneaking(player)
__on_player_starts_sprinting(player)
__on_player_stops_sprinting(player)
__on_player_drops_item(player)
__on_player_drops_stack(player)
__on_statistic(player, category, event, value) // player statistic changes

/script event command

used to display current events and bounded functions. use add_to ro register new event, or remove_from to unbind a specific function from an event. Function to be bounded to an event needs to have the same number of parameters as the action is attempting to bind to (see list above). All calls in modules loaded via /script load that have functions listed above will be automatically bounded and unbounded when script is unloaded.

Scoreboard

scoreboard(), scoreboard(objective), scoreboard(objective, key), scoreboard(objective, key, value)

Displays or modifies individual scoreboard values. With no arguments, returns the list of current objectives. With specified objective, lists all keys (players) associated with current objective. With specified objective, key, returns current value of the objective for a given player (key). With additional value sets a new scoreboard value, returning previous value associated with the key.

scoreboard_add(objective, criterion?)

Adds a new objective to scoreboard. If criterion is not specified, assumes 'dummy'. Returns false if objective already existed, true otherwise.

scoreboard_add('counter')
scoreboard_add('lvl','level')

scoreboard_remove(objective)

Removes an objective. Returns true if objective has existed and has been removed.

scoreboard_display(objective, place)

sets display location for a specified objective. If place is null, then display is cleared.

Auxiliary aspects

Collection of other methods that control smaller, yet still important aspects of the game

Sounds

sound(name, pos, volume?, pitch?)

Plays a specific sound name, at block or position pos, with optional volume and modified pitch. pos can be either a block, triple of coords, or a list of thee numbers. Uses the same options as a corresponding playsound command.

Particles

particle(name, pos, count?. spread?, speed?, playername?)

Renders a cloud of particles name centered around pos position, by default count 10 of them, default speed of 0, and to all players nearby, but these options can be changed via optional arguments. Follow vanilla /particle command on details on those options. Valid particle names are for example 'angry_villager', 'item diamond', 'block stone', 'dust 0.8 0.1 0.1 4'.

particle_line(name, pos, pos2, density?)

Renders a line of particles from point pos to pos2 with supplied density (defaults to 1), which indicates how far apart you would want particles to appear, so 0.1 means one every 10cm.

particle_rect(name, pos, pos2, density?)

Renders a cuboid of particles between point pos to pos2 with supplied density.

Markers

create_marker(text, pos, rotation?, block?)

Spawns a (permanent) marker entity with text or block at position. Returns that entity for further manipulations. Unloading the app that spawned them will cause all the markers from the loaded portion of the world to be removed. Also - if the game loads that marker in the future and the app is not loaded, it will be removed as well.

remove_all_markers()

Removes all scarpet markers from the loaded portion of the world created by this app, in case you didn't want to do the proper cleanup.

System function

nbt(expr)

Treats the argument as a nbt serializable string and returns its nbt value. In case nbt is not in a correct nbt compound tag format, it will return null value.

Consult section about container operations in Expression to learn about possible operations on nbt values.

escape_nbt(expr)

Excapes all the special characters in the string or nbt tag and returns a string that can be stored in nbt directly as a string value.

print(expr)

Displays the result of the expression to the chat. Overrides default scarpet behaviour of sending everyting to stderr.

logger(expr)

Prints the message to system logs, and not to chat.

run(expr)

Runs a vanilla command from the string result of the expr and returns its success count

run('fill 1 1 1 10 10 10 air') -> 123 // 123 block were filled, this operation was successful 123 times out of a possible 1000 blocks volume
run('give @s stone 4') -> 1 // this operation was successful once

save()

Performs autosave, saves all chunks, player data, etc. Useful for programs where autosave is disabled due to performance reasons and saves the world only on demand.

load_app_data(), load_app_data(file)

Loads the app data associated with the app from the world /scripts folder. Without argument returns the memory managed and buffered / throttled NBT tag. With a file name - reads explicitly a file with that name from the scripts folder.

You can use app data to save non-vanilla information separately from the world and other scripts.

store_app_data(tag), store_app_data(tag, file)

Stores the app data associated with the app from the world /scripts folder. With the file parameter saves immediately and with every call, without file parameter, it may take up to 10 seconds for the output file to sync preventing flickering in case this tag changes frequently. It will be synced when server closes.

tick_time()

Returns game tick counter. Can be used to run certain operations every n-th ticks, or to count in-game time

game_tick(mstime?)

Causes game to run for one tick. By default runs it and returns control to the program, but can optionally accept expected tick length, in milliseconds. You can't use it to permanently change the game speed, but setting longer commands with custom tick speeds can be interrupted via /script stop command

loop(1000,game_tick())  // runs the game as fast as it can for 1000 ticks
loop(1000,game_tick(100)) // runs the game twice as slow for 1000 ticks

seed()

Returns current world seed.

current_dimension()

Returns current dimension that the script runs in.

in_dimension(smth, expr)

Evaluates the expression expr with different dimension execution context. smth can be an entity, world-localized block, so not block('stone'), or a string representing a dimension like: 'nether', 'the_nether', 'end' or 'overworld', etc.

schedule(delay, function, args...)

Schedules a user defined function to run with a specified delay ticks of delay. Scheduled functions run at the end of the tick, and they will run in order they were scheduled.

statistic(player, category, entry)

Queries in-game statistics for certain values. Categories include:

• mined: blocks mined
• crafted: items crafted
• used: items used
• broken: items broken
• picked_up: items picked up
• dropped: items dropped
• killed: mobs killed
• killed_by: blocks mined
• custom: various random stats

For the options of entry, consult your statistics page, or give it a guess.

The call will return null if the statistics options are incorrect, or player didn't get these in their history. If the player encountered the statistic, or game created for him empty one, it will return a number. Scarpet will not affect the entries of the statistics, even if it is just creating empty ones. With null response it could either mean your input is wrong, or statistic has effectively a value of 0.

plop(pos, what)

Plops a structure or a feature at a given pos, so block, triple position coordinates or a list of coordinates. To what gets plopped and exactly where it often depends on the feature or structure itself. For example, all structures are chunk aligned, and often span multiple chunks. Repeated calls to plop a structure in the same chunk would result either in the same structure generated on top of each other, or with different state, but same position. Most structures generate at specific altitudes, which are hardcoded, or with certain blocks around them. API will cancel all extra position / biome / random requirements for structure / feature placement, but some hardcoded limitations may still cause some of structures/features not to place. Some features require special blocks to be present, like coral -> water or ice spikes -> snow block, and for some features, like fossils, placement is all sorts of messed up. This can be partially avoided for structures by setting their structure information via set_structure, which sets it without looking into world blocks, and then use plop to fill it with blocks. This may, or may not work.

All generated structures will retain their properties, like mob spawning, however in many cases the world / dimension itself has certain rules to spawn mobs, like plopping a nether fortress in the overworld will not spawn nether mobs, because nether mobs can spawn only in the nether, but plopped in the nether - will behave like a valid nether fortress.

plop will not use world random number generator to generate structures and features, but its own. This has a benefit that they will generate properly randomly, not the same time every time.

Structure list:

• monument: Ocean Monument. Generates at fixed Y coordinate, surrounds itself with water.
• fortress: Nether Fortress. Altitude varies, but its bounded by the code.
• mansion: Woodland Mansion
• jungle_temple: Jungle Temple
• desert_temple: Desert Temple. Generates at fixed Y altitude.
• end_city: End City with Shulkers
• igloo: Igloo
• shipwreck: Shipwreck, version1?
• shipwreck2: Shipwreck, version2?
• witch_hut
• ocean_ruin, ocean_ruin_small, ocean_ruin_tall: Stone variants of ocean ruins.
• ocean_ruin_warm, ocean_ruin_warm_small, ocean_ruin_warm_tall: Sandstone variants of ocean ruins.
• treasure: A treasure chest. Yes, its a whole structure.
• pillager_outpost: A pillager outpost.
• mineshaft: A mineshaft.
• mineshaft_mesa: A Mesa (Badlands) version of a mineshaft.
• village: Plains, oak village.
• village_desert: Desert, sandstone village.
• village_savanna: Savanna, acacia village.
• village_taiga: Taiga, spruce village.
• village_snowy: Resolute, Canada.

Feature list:

• oak
• oak_beehive: oak with a hive (1.15+).
• oak_large: oak with branches.
• oak_large_beehive: oak with branches and a beehive (1.15+).
• birch
• birch_large: tall variant of birch tree.
• shrub: low bushes that grow in jungles.
• shrub_acacia: low bush but configured with acacia (1.14 only)
• shrub_snowy: low bush with white blocks (1.14 only)
• jungle: a tree
• jungle_large: 2x2 jungle tree
• spruce
• spruce_large: 2x2 spruce tree
• pine: spruce with minimal leafage (1.15+)
• pine_large: 2x2 spruce with minimal leafage (1.15+)
• spruce_matchstick: see 1.15 pine (1.14 only).
• spruce_matchstick_large: see 1.15 pine_large (1.14 only).
• dark_oak
• acacia
• oak_swamp: oak with more leaves and vines.
• well: desert well
• grass: a few spots of tall grass
• grass_jungle: little bushier grass feature (1.14 only)
• lush_grass: grass with patchy ferns (1.15+)
• tall_grass: 2-high grass patch (1.15+)
• fern: a few random 2-high ferns
• cactus: random cacti
• dead_bush: a few random dead bushi
• fossils: underground fossils, placement little wonky
• mushroom_brown: large brown mushroom.
• mushroom_red: large red mushroom.
• ice_spike: ice spike. Require snow block below to place.
• glowstone: glowstone cluster. Required netherrack above it.
• melon: a patch of melons
• melon_pile: a pile of melons (1.15+)
• pumpkin: a patch of pumpkins
• pumpkin_pile: a pile of pumpkins (1.15+)
• sugarcane
• lilypad
• dungeon: Dungeon. These are hard to place, and fail often.
• iceberg: Iceberg. Generate at sea level.
• iceberg_blue: Blue ice iceberg.
• lake
• lava_lake
• end_island
• chorus: Chorus plant. Require endstone to place.
• sea_grass: a patch of sea grass. Require water.
• sea_grass_river: a variant.
• kelp
• coral_tree, coral_mushroom, coral_claw: various coral types, random color.
• coral: random coral structure. Require water to spawn.
• sea_pickle
• boulder: A rocky, mossy formation from a giant taiga biome. Doesn't update client properly, needs relogging.

/script run command

Primary way to input commands. The command executes in the context, position, and dimension of the executing player, commandblock, etc... The command receives 4 variables, x, y, z and p indicating position and the executing entity of the command. You will receive tab completion suggestions as you type your code suggesting functions and global variables. It is advisable to use /execute in ... at ... as ... run script run ... or similar, to simulate running commands in a different scope.

/script load / unload <app> (global?), /script in <app> commands

load / unload commands allow for very conventient way of writing your code, providing it to the game and distribute with your worlds without the need of use of commandblocks. Just place your scarpet code in the /scripts folder of your world files and make sure it ends with .sc extension. The good thing about editing that code is that you can no only use normal editing without the need of marking of newlines, but you can also use comments in your code.

a comment is anything that starts with a double slash, and continues to the end of the line:

foo = 1;
//This is a comment
bar = 2;
// This never worked, so I commented it out
// baz = foo()

/script load/unload <app> (?global)

Loading operation will load that script code from disk and execute it right away. You would probably use it to load some stored procedures to be used for later. To reload the module, just type /script load again. Reloading removes all the current global state (globals and functions) that were added later by the module.

Loading a module, if it contains a __command() method, will attempt to registed a command with that app name, and register all public (no underscore) functions available in the module as subcommands. It will also bind specific events to the event system (check Events section for details).

Loaded apps have the ability to store and load external files, especially their persistent tag state. For that check load_app_data and store_app_data functions.

If an app defines __config method, and that method returns a map, it will be used to apply custom settings for this app. Currently the following options are supported:

• scope: default scope for global variables for the app, Default is 'player' which means that globals and defined functions will be unique for each player so that apps for each player will run in isolation. This is useful in tool-like applications. Another option is 'global' which shares global state for all runs on the server - applicable to 'block' like solutions, where custom behaviours are applied to blocks.
• stay_loaded: defaults to false. If true, and /carpet scriptsAutoload is turned on, the following apps will stay loaded after startup. Otherwise, after reading a code, and fetching the config, server will drop them down. This is to allow to store multiple apps on the server/world and selectively decide which one should be running at startup. WARNING: all apps will run once at startup anyways, so be aware that their actions that are called statically, will be performed once anyways.

Unloading an app removes all of its state from the game, disables commands, and removes bounded events, and saves its global state. If more cleanup is needed, one can define an __on_close() function which will be executed when the module is unloaded, or server is closing, or crashing. However, there is no need to do that explicitly for the things mentioned in the previous statement.

Scripts can be loaded in shared(global) and player mode. Default is player, so all globals and stored functions are individual for each player, meaning scripts don't need to worry about making sure they store some intermittent data for each player independently. In global mode - all global values and stored functions are shared among all players. To access specific player data with commandblocks, use /execute as (player) run script in (app) run .... To access global/server state, you need to disown the command from any player, so use a commandblock, or any arbitrary entity: /execute as @e[type=bat,limit=1] run script in (module) globals for instance.

/script in <app> ...

Allows to run normal /script commands in a specific app, like run, invoke,..., globals etc...

/script invoke / invokepoint / invokearea, /script globals commands

invoke family of commands provide convenient way to invoke stored procedures (i.e. functions that has been defined previously by any running script. To view current stored procedure set, run /script globals(or /script globals all to display all functions even hidden ones), to define a new stored procedure, just run a /script run function(a,b) -> ( ... ) command with your procedure once, and to forget a procedure, use undef function: /script run undef('function')

/script invoke <fun> <args?> ...

Equivalent of running /script run fun(args, ...), but you get the benefit of getting the tab completion of the command name, and lower permission level required to run these (since player is not capable of running any custom code in this case, only this that has been executed before by an operator). Arguments will be checked for validity, and you can only pass simple values as arguments (strings, numbers, or null value). Use quotes to include whitespaces in argument strings.

Command will check provided arguments with required arguments (count) and fail if not enough or too much arguments are provided. Operators defining functions are advised to use descriptive arguments names, as these will be visible for invokers and form the base of understanding what each argument does.

invoke family of commands will tab complete any stored function that does not start with '_', it will still allow to run procedures starting with '_' but not suggest them, and ban execution of any hidden stored procedures, so ones that start with '__'. In case operator needs to use subroutines for convenience and don't want to expose them to the invoke callers, they can use this mechanic.

/script run example_function(const, phrase, price) -> print(const+' '+phrase+' '+price)
/script invoke example_function pi costs 5

/script invokepoint <fun> <coords x y z> <args?> ...

It is equivalent to invoke except it assumes that the first three arguments are coordinates, and provides coordinates tab completion, with looking at... mechanics for convenience. All other arguments are expected at the end

/script invokearea <fun> <coords x y z> <coords x y z> <args?> ...

It is equivalent to invoke except it assumes that the first three arguments are one set of ccordinates, followed by the second set of coordinates, providing tab completion, with looking at... mechanics for convenience, followed by any other required arguments

/script scan, /script fill and /script outline commands

These commands can be used to evaluate an expression over an area of blocks. They all need to have specified the origin of the analyzed area (which is used as referenced (0,0,0), and two corners of an area to analyzed. If you would want that the script block coordinates refer to the actual world coordinates, use origin of 0 0 0, or if it doesn't matter, duplicating coordinates of one of the corners is the easiest way.

These commands, unlike raw /script run are limited by vanilla fill / clone command limit of 32k blocks, which can be altered with carpet mod's own /carpet fillLimit command.

/script scan origin<x y z> corner<x y z> corner<x y z> expr

Evaluates expression for each point in the area and returns number of successes (result was positive). Since the command by itself doesn't affect the area, the effects would be in side effects.

/script fill origin<x y z> corner<x y z> corner<x y z> "expr" <block> (? replace <replacement>)

Think of it as a regular fill command, that sets blocks based on whether a result of the command was successful. Note that the expression is in quotes. Thankfully string constants in scarpet use single quotes. Can be used to fill complex geometric shapes.

/script outline origin<x y z> corner<x y z> corner<x y z> "expr" <block> (? replace <replacement>)

Similar to fill command it evaluates an expression for each block in the area, but in this case setting blocks where condition was true and any of the neighbouring blocks were evaluated negatively. This allows to create surface areas, like sphere for example, without resorting to various rounding modes and tricks.

Here is an example of seven ways to draw a sphere of radius of 32 blocks around 0 100 0:

/script outline 0 100 0 -40 60 -40 40 140 40 "x*x+y*y+z*z <  32*32" white_stained_glass replace air
/script outline 0 100 0 -40 60 -40 40 140 40 "x*x+y*y+z*z <= 32*32" white_stained_glass replace air
/script outline 0 100 0 -40 60 -40 40 140 40 "x*x+y*y+z*z <  32.5*32.5" white_stained_glass replace air
/script fill    0 100 0 -40 60 -40 40 140 40 "floor(sqrt(x*x+y*y+z*z)) == 32" white_stained_glass replace air
/script fill    0 100 0 -40 60 -40 40 140 40 "round(sqrt(x*x+y*y+z*z)) == 32" white_stained_glass replace air
/script fill    0 100 0 -40 60 -40 40 140 40 "ceil(sqrt(x*x+y*y+z*z)) == 32" white_stained_glass replace air
/draw sphere 0 100 0 32 white_stained_glass replace air // fluffy ball round(sqrt(x*x+y*y+z*z)-rand(abs(y)))==32

The last method is the one that world edit is using (part of carpet mod). It turns out that the outline method with 32.5 radius, fill method with round function and draw command are equivalent

script stop/script resume command

/script stop allows to stop execution of any script currently running that calls the game_tick() function which allows the game loop to regain control of the game and process other commands. This will also make sure that all current and future programs will stop their execution. Execution of all programs will be prevented until /script resume command is called.

Lets look at the following example. This is a program computes Fibonacci number in a recursive manner:

fib(n) -> if(n<3, 1, fib(n-1)+fib(n-2) ); fib(8)

That's really bad way of doing it, because the higher number we need to compute the compute requirements will rise exponentially with n. It takes a little over 50 milliseconds to do fib(24), so above one tick, but about a minute to do fib(40). Calling fib(40) will not only freeze the game, but also you woudn't be able to interrupt its execution. We can modify the script as follows

fib(n) -> ( game_tick(50); if(n<3, 1, fib(n-1)+fib(n-2) ) ); fib(40)

But this would never finish as such call would finish after ~ 2^40 ticks. To make our computations responsive, yet able to respond to user interactions, other commands, as well as interrupt execution, we could do the following:

fib(n) -> ( if(n==23, game_tick(50) ); if(n<3, 1, fib(n-1)+fib(n-2) ) ); fib(40)

This would slow down the computation of fib(40) from a minute to two, but allows the game to keep continue running and be responsive to commands, using about half of each tick to advance the computation. Obviously depending on the problem, and available hardware, certain things can take more or less time to execute, so portioning of work with calling gametick should be balanced in each case separately