The last time Hackerfall tried to access this page, it returned a not found error. A cached version of the page is below, or click here to continue anyway

Nim Tutorial (Part I)

Author: Andreas Rumpf Version: 0.13.0

"Der Mensch ist doch ein Augentier -- schöne Dinge wünsch ich mir."

This document is a tutorial for the programming language Nim. This tutorial assumes that you are familiar with basic programming concepts like variables, types or statements but is kept very basic. The manual contains many more examples of the advanced language features. All code examples in this tutorial, as well as the ones found in the rest of Nim's documentation, follow the Nim style guide <nep1.html>.

We start the tour with a modified "hello world" program:

echo("What's your name? ")
var name: string = readLine(stdin)
echo("Hi, ", name, "!")

Save this code to the file "greetings.nim". Now compile and run it:

nim compile --run greetings.nim

With the --run switch Nim executes the file automatically after compilation. You can give your program command line arguments by appending them after the filename:

nim compile --run greetings.nim arg1 arg2

Commonly used commands and switches have abbreviations, so you can also use:

nim c -r greetings.nim

To compile a release version use:

nim c -d:release greetings.nim

By default the Nim compiler generates a large amount of runtime checks aiming for your debugging pleasure. With -d:release these checks are turned off and optimizations are turned on.

Though it should be pretty obvious what the program does, I will explain the syntax: statements which are not indented are executed when the program starts. Indentation is Nim's way of grouping statements. Indentation is done with spaces only, tabulators are not allowed.

String literals are enclosed in double quotes. The var statement declares a new variable named name of type string with the value that is returned by the readLine procedure. Since the compiler knows that readLine returns a string, you can leave out the type in the declaration (this is called local type inference). So this will work too:

var name = readLine(stdin)

Note that this is basically the only form of type inference that exists in Nim: it is a good compromise between brevity and readability.

The "hello world" program contains several identifiers that are already known to the compiler: echo, readLine, etc. These built-ins are declared in the system module which is implicitly imported by any other module.

Let us look at Nim's lexical elements in more detail: like other programming languages Nim consists of (string) literals, identifiers, keywords, comments, operators, and other punctuation marks.

String literals are enclosed in double quotes; character literals in single quotes. Special characters are escaped with \: \n means newline, \t means tabulator, etc. There are also raw string literals:

r"C:\program files\nim"

In raw literals the backslash is not an escape character.

The third and last way to write string literals are long string literals. They are written with three quotes: """ ... """; they can span over multiple lines and the \ is not an escape character either. They are very useful for embedding HTML code templates for example.

Comments start anywhere outside a string or character literal with the hash character #. Documentation comments start with ##:

var myVariable: int 

Documentation comments are tokens; they are only allowed at certain places in the input file as they belong to the syntax tree! This feature enables simpler documentation generators.

You can also use the discard statement together with long string literals to create block comments:

discard """ You can have any Nim code text commented
out inside this with no indentation restrictions.
      yes("May I ask a pointless question?") """

Numerical literals are written as in most other languages. As a special twist, underscores are allowed for better readability: 1_000_000 (one million). A number that contains a dot (or 'e' or 'E') is a floating point literal: 1.0e9 (one billion). Hexadecimal literals are prefixed with 0x, binary literals with 0b and octal literals with 0o. A leading zero alone does not produce an octal.

The var statement declares a new local or global variable:

var x, y: int 

Indentation can be used after the var keyword to list a whole section of variables:

  x, y: int
  a, b, c: string

The assignment statement assigns a new value to a variable or more generally to a storage location:

var x = "abc" 
x = "xyz"     

= is the assignment operator. The assignment operator cannot be overloaded, overwritten or forbidden, but this might change in a future version of Nim. You can declare multiple variables with a single assignment statement and all the variables will have the same value:

var x, y = 3  
echo "x ", x  
echo "y ", y  
x = 42        
echo "x ", x  
echo "y ", y  

Note that declaring multiple variables with a single assignment which calls a procedure can have unexpected results: the compiler will unroll the assignments and end up calling the procedure several times. If the result of the procedure depends on side effects, your variables may end up having different values! For safety use only constant values.

Constants are symbols which are bound to a value. The constant's value cannot change. The compiler must be able to evaluate the expression in a constant declaration at compile time:

const x = "abc" 

Indentation can be used after the const keyword to list a whole section of constants:

  x = 1
  y = 2
  z = y + 5 

The let statement works like the var statement but the declared symbols are single assignment variables: After the initialization their value cannot change:

let x = "abc" 
x = "xyz"     

The difference between let and const is: let introduces a variable that can not be re-assigned, const means "enforce compile time evaluation and put it into a data section":

const input = readLine(stdin) 
let input = readLine(stdin)   

The greetings program consists of 3 statements that are executed sequentially. Only the most primitive programs can get away with that: branching and looping are needed too.

The if statement is one way to branch the control flow:

let name = readLine(stdin)
if name == "":
  echo("Poor soul, you lost your name?")
elif name == "name":
  echo("Very funny, your name is name.")
  echo("Hi, ", name, "!")

There can be zero or more elif parts, and the else part is optional. The keyword elif is short for else if, and is useful to avoid excessive indentation. (The "" is the empty string. It contains no characters.)

Another way to branch is provided by the case statement. A case statement is a multi-branch:

let name = readLine(stdin)
case name
of "":
  echo("Poor soul, you lost your name?")
of "name":
  echo("Very funny, your name is name.")
of "Dave", "Frank":
  echo("Cool name!")
  echo("Hi, ", name, "!")

As it can be seen, for an of branch a comma separated list of values is also allowed.

The case statement can deal with integers, other ordinal types and strings. (What an ordinal type is will be explained soon.) For integers or other ordinal types value ranges are also possible:

from strutils import parseInt

echo("A number please: ")
let n = parseInt(readLine(stdin))
case n
of 0..2, 4..7: echo("The number is in the set: {0, 1, 2, 4, 5, 6, 7}")
of 3, 8: echo("The number is 3 or 8")

However, the above code does not compile: the reason is that you have to cover every value that n may contain, but the code only handles the values 0..8. Since it is not very practical to list every other possible integer (though it is possible thanks to the range notation), we fix this by telling the compiler that for every other value nothing should be done:

case n
of 0..2, 4..7: echo("The number is in the set: {0, 1, 2, 4, 5, 6, 7}")
of 3, 8: echo("The number is 3 or 8")
else: discard

The empty discard statement is a do nothing statement. The compiler knows that a case statement with an else part cannot fail and thus the error disappears. Note that it is impossible to cover all possible string values: that is why string cases always need an else branch.

In general the case statement is used for subrange types or enumerations where it is of great help that the compiler checks that you covered any possible value.

The while statement is a simple looping construct:

echo("What's your name? ")
var name = readLine(stdin)
while name == "":
  echo("Please tell me your name: ")
  name = readLine(stdin)

The example uses a while loop to keep asking the users for their name, as long as the user types in nothing (only presses RETURN).

The for statement is a construct to loop over any element an iterator provides. The example uses the built-in countup iterator:

echo("Counting to ten: ")
for i in countup(1, 10):

The built-in $ operator turns an integer (int) and many other types into a string. The variable i is implicitly declared by the for loop and has the type int, because that is what countup returns. i runs through the values 1, 2, .., 10. Each value is echo-ed. This code does the same:

echo("Counting to 10: ")
var i = 1
while i <= 10:

Counting down can be achieved as easily (but is less often needed):

echo("Counting down from 10 to 1: ")
for i in countdown(10, 1):

Since counting up occurs so often in programs, Nim also has a .. iterator that does the same:

for i in 1..10:

Control flow statements have a feature not covered yet: they open a new scope. This means that in the following example, x is not accessible outside the loop:

while false:
  var x = "hi"

A while (for) statement introduces an implicit block. Identifiers are only visible within the block they have been declared. The block statement can be used to open a new block explicitly:

block myblock:
  var x = "hi"

The block's label (myblock in the example) is optional.

A block can be left prematurely with a break statement. The break statement can leave a while, for, or a block statement. It leaves the innermost construct, unless a label of a block is given:

block myblock:
  echo("entering block")
  while true:
  echo("still in block")

block myblock2:
  echo("entering block")
  while true:
    break myblock2 
  echo("still in block")

Like in many other programming languages, a continue statement starts the next iteration immediately:

while true:
  let x = readLine(stdin)
  if x == "": continue


when system.hostOS == "windows":
  echo("running on Windows!")
elif system.hostOS == "linux":
  echo("running on Linux!")
elif system.hostOS == "macosx":
  echo("running on Mac OS X!")
  echo("unknown operating system")

The when statement is almost identical to the if statement with some differences:

The when statement is useful for writing platform specific code, similar to the #ifdef construct in the C programming language.

Note: To comment out a large piece of code, it is often better to use a when false: statement than to use real comments. This way nesting is possible.

Now that we covered the basic control flow statements, let's return to Nim indentation rules.

In Nim there is a distinction between simple statements and complex statements. Simple statements cannot contain other statements: Assignment, procedure calls or the return statement belong to the simple statements. Complex statements like if, when, for, while can contain other statements. To avoid ambiguities, complex statements always have to be indented, but single simple statements do not:

if x: x = false

if x:
  if y:
    y = false
    y = true

if x:
  x = false
  y = false

Expressions are parts of a statement which usually result in a value. The condition in an if statement is an example for an expression. Expressions can contain indentation at certain places for better readability:

if thisIsaLongCondition() and
       2, 3, 4):
  x = true

As a rule of thumb, indentation within expressions is allowed after operators, an open parenthesis and after commas.

With parenthesis and semicolons (;) you can use statements where only an expression is allowed:

const fac4 = (var x = 1; for i in 1..4: x *= i; x)

To define new commands like echo and readLine in the examples, the concept of a procedure is needed. (Some languages call them methods or functions.) In Nim new procedures are defined with the proc keyword:

proc yes(question: string): bool =
  echo(question, " (y/n)")
  while true:
    case readLine(stdin)
    of "y", "Y", "yes", "Yes": return true
    of "n", "N", "no", "No": return false
    else: echo("Please be clear: yes or no")

if yes("Should I delete all your important files?"):
  echo("I'm sorry Dave, I'm afraid I can't do that.")
  echo("I think you know what the problem is just as well as I do.")

This example shows a procedure named yes that asks the user a question and returns true if they answered "yes" (or something similar) and returns false if they answered "no" (or something similar). A return statement leaves the procedure (and therefore the while loop) immediately. The (question: string): bool syntax describes that the procedure expects a parameter named question of type string and returns a value of type bool. Bool is a built-in type: the only valid values for bool are true and false. The conditions in if or while statements should be of the type bool.

Some terminology: in the example question is called a (formal) parameter, "Should I..." is called an argument that is passed to this parameter.

A procedure that returns a value has an implicit result variable declared that represents the return value. A return statement with no expression is a shorthand for return result. The result value is always returned automatically at the end a procedure if there is no return statement at the exit.

proc sumTillNegative(x: varargs[int]): int =
  for i in x:
    if i < 0:
    result = result + i

echo sumTillNegative() 
echo sumTillNegative(3, 4, 5) 
echo sumTillNegative(3, 4 , -1 , 6) 

The result variable is already implicitly declared at the start of the function, so declaring it again with 'var result', for example, would shadow it with a normal variable of the same name. The result variable is also already initialised with the type's default value. Note that referential data types will be nil at the start of the procedure, and thus may require manual initialisation.

Parameters are constant in the procedure body. By default, their value cannot be changed because this allows the compiler to implement parameter passing in the most efficient way. If a mutable variable is needed inside the procedure, it has to be declared with var in the procedure body. Shadowing the parameter name is possible, and actually an idiom:

proc printSeq(s: seq, nprinted: int = -1) =
  var nprinted = if nprinted == -1: s.len else: min(nprinted, s.len)
  for i in 0 .. <nprinted:
    echo s[i]

If the procedure needs to modify the argument for the caller, a var parameter can be used:

proc divmod(a, b: int; res, remainder: var int) =
  res = a div b        
  remainder = a mod b  

  x, y: int
divmod(8, 5, x, y) 

In the example, res and remainder are var parameters. Var parameters can be modified by the procedure and the changes are visible to the caller. Note that the above example would better make use of a tuple as a return value instead of using var parameters.

To call a procedure that returns a value just for its side effects and ignoring its return value, a discard statement has to be used. Nim does not allow to silently throw away a return value:

discard yes("May I ask a pointless question?")

The return value can be ignored implicitly if the called proc/iterator has been declared with the discardable pragma:

proc p(x, y: int): int {.discardable.} =
  return x + y

p(3, 4) 

The discard statement can also be used to create block comments as described in the Comments section.

Often a procedure has many parameters and it is not clear in which order the parameters appear. This is especially true for procedures that construct a complex data type. Therefore the arguments to a procedure can be named, so that it is clear which argument belongs to which parameter:

proc createWindow(x, y, width, height: int; title: string;
                  show: bool): Window =

var w = createWindow(show = true, title = "My Application",
                     x = 0, y = 0, height = 600, width = 800)

Now that we use named arguments to call createWindow the argument order does not matter anymore. Mixing named arguments with ordered arguments is also possible, but not very readable:

var w = createWindow(0, 0, title = "My Application",
                     height = 600, width = 800, true)

The compiler checks that each parameter receives exactly one argument.

To make the createWindow proc easier to use it should provide default values, these are values that are used as arguments if the caller does not specify them:

proc createWindow(x = 0, y = 0, width = 500, height = 700,
                  title = "unknown",
                  show = true): Window =

var w = createWindow(title = "My Application", height = 600, width = 800)

Now the call to createWindow only needs to set the values that differ from the defaults.

Note that type inference works for parameters with default values; there is no need to write title: string = "unknown", for example.

Nim provides the ability to overload procedures similar to C++:

proc toString(x: int): string = ...
proc toString(x: bool): string =
  if x: result = "true"
  else: result = "false"


(Note that toString is usually the $ operator in Nim.) The compiler chooses the most appropriate proc for the toString calls. How this overloading resolution algorithm works exactly is not discussed here (it will be specified in the manual soon). However, it does not lead to nasty surprises and is based on a quite simple unification algorithm. Ambiguous calls are reported as errors.

The Nim library makes heavy use of overloading - one reason for this is that each operator like + is a just an overloaded proc. The parser lets you use operators in infix notation (a + b) or prefix notation (+ a). An infix operator always receives two arguments, a prefix operator always one. Postfix operators are not possible, because this would be ambiguous: does a @ @ b mean (a) @ (@b) or (a@) @ (b)? It always means (a) @ (@b), because there are no postfix operators in Nim.

Apart from a few built-in keyword operators such as and, or, not, operators always consist of these characters: + - * \ / < > = @ $ ~ & % ! ? ^ . |

User defined operators are allowed. Nothing stops you from defining your own @!?+~ operator, but readability can suffer.

The operator's precedence is determined by its first character. The details can be found in the manual.

To define a new operator enclose the operator in backticks "``":

proc `$` (x: myDataType): string = ...

The "``" notation can also be used to call an operator just like any other procedure:

if `==`( `+`(3, 4), 7): echo("True")

Every variable, procedure, etc. needs to be declared before it can be used. (The reason for this is that it is non-trivial to do better than that in a language that supports meta programming as extensively as Nim does.) However, this cannot be done for mutually recursive procedures:

proc even(n: int): bool
proc odd(n: int): bool =
  assert(n >= 0) 
  if n == 0: false
    n == 1 or even(n-1)

proc even(n: int): bool =
  assert(n >= 0) 
  if n == 1: false
    n == 0 or odd(n-1)

Here odd depends on even and vice versa. Thus even needs to be introduced to the compiler before it is completely defined. The syntax for such a forward declaration is simple: just omit the = and the procedure's body. The assert just adds border conditions, and will be covered later in Modules section.

Later versions of the language will weaken the requirements for forward declarations.

The example also shows that a proc's body can consist of a single expression whose value is then returned implicitly.

Let's return to the boring counting example:

echo("Counting to ten: ")
for i in countup(1, 10):

Can a countup proc be written that supports this loop? Lets try:

proc countup(a, b: int): int =
  var res = a
  while res <= b:
    return res

However, this does not work. The problem is that the procedure should not only return, but return and continue after an iteration has finished. This return and continue is called a yield statement. Now the only thing left to do is to replace the proc keyword by iterator and there it is - our first iterator:

iterator countup(a, b: int): int =
  var res = a
  while res <= b:
    yield res

Iterators look very similar to procedures, but there are several important differences:

However, you can also use a closure iterator to get a different set of restrictions. See first class iterators for details. Iterators can have the same name and parameters as a proc, essentially they have their own namespace. Therefore it is common practice to wrap iterators in procs of the same name which accumulate the result of the iterator and return it as a sequence, like split from the strutils module.

This section deals with the basic built-in types and the operations that are available for them in detail.

The boolean type is named bool in Nim and consists of the two pre-defined values true and false. Conditions in while, if, elif, when statements need to be of type bool.

The operators not, and, or, xor, <, <=, >, >=, !=, == are defined for the bool type. The and and or operators perform short-cut evaluation. Example:

while p != nil and != "xyz":
  p =

The character type is named char in Nim. Its size is one byte. Thus it cannot represent an UTF-8 character, but a part of it. The reason for this is efficiency: for the overwhelming majority of use-cases, the resulting programs will still handle UTF-8 properly as UTF-8 was specially designed for this. Character literals are enclosed in single quotes.

Chars can be compared with the ==, <, <=, >, >= operators. The $ operator converts a char to a string. Chars cannot be mixed with integers; to get the ordinal value of a char use the ord proc. Converting from an integer to a char is done with the chr proc.

String variables in Nim are mutable, so appending to a string is quite efficient. Strings in Nim are both zero-terminated and have a length field. One can retrieve a string's length with the builtin len procedure; the length never counts the terminating zero. Accessing the terminating zero is no error and often leads to simpler code:

if s[i] == 'a' and s[i+1] == 'b':

The assignment operator for strings copies the string. You can use the & operator to concatenate strings and add to append to a string.

Strings are compared by their lexicographical order. All comparison operators are available. Per convention, all strings are UTF-8 strings, but this is not enforced. For example, when reading strings from binary files, they are merely a sequence of bytes. The index operation s[i] means the i-th char of s, not the i-th unichar.

String variables are initialized with a special value, called nil. However, most string operations cannot deal with nil (leading to an exception being raised) for performance reasons. One should use empty strings "" rather than nil as the empty value. But "" often creates a string object on the heap, so there is a trade-off to be made here.

Nim has these integer types built-in: int int8 int16 int32 int64 uint uint8 uint16 uint32 uint64.

The default integer type is int. Integer literals can have a type suffix to mark them to be of another integer type:

  x = 0     
  y = 0'i8  
  z = 0'i64 
  u = 0'u   

Most often integers are used for counting objects that reside in memory, so int has the same size as a pointer.

The common operators + - * div mod < <= == != > >= are defined for integers. The and or xor not operators are defined for integers too and provide bitwise operations. Left bit shifting is done with the shl, right shifting with the shr operator. Bit shifting operators always treat their arguments as unsigned. For arithmetic bit shifts ordinary multiplication or division can be used.

Unsigned operations all wrap around; they cannot lead to over- or underflow errors.

Automatic type conversion is performed in expressions where different kinds of integer types are used. However, if the type conversion loses information, the EOutOfRange exception is raised (if the error cannot be detected at compile time).

Nim has these floating point types built-in: float float32 float64.

The default float type is float. In the current implementation, float is always 64 bit wide.

Float literals can have a type suffix to mark them to be of another float type:

  x = 0.0      
  y = 0.0'f32  
  z = 0.0'f64  

The common operators + - * / < <= == != > >= are defined for floats and follow the IEEE standard.

Automatic type conversion in expressions with different kinds of floating point types is performed: the smaller type is converted to the larger. Integer types are not converted to floating point types automatically and vice versa. The toInt and toFloat procs can be used for these conversions.

Conversion between basic types in nim is performed by using the type as a function:

  x: int32 = 1.int32   
  y: int8  = int8('a') 
  z: float = 2.5       
  sum: int = int(x) + int(y) + int(z) 

As mentioned earlier, the built-in $ (stringify) operator turns any basic type into a string, which you can then print to the screen with the echo proc. However, advanced types, or types you may define yourself won't work with the $ operator until you define one for them. Sometimes you just want to debug the current value of a complex type without having to write its $ operator. You can use then the repr proc which works with any type and even complex data graphs with cycles. The following example shows that even for basic types there is a difference between the $ and repr outputs:

  myBool = true
  myCharacter = 'n'
  myString = "nim"
  myInteger = 42
  myFloat = 3.14
echo($myBool, ":", repr(myBool))

echo($myCharacter, ":", repr(myCharacter))

echo($myString, ":", repr(myString))

echo($myInteger, ":", repr(myInteger))

echo($myFloat, ":", repr(myFloat))

In Nim new types can be defined within a type statement:

  biggestInt = int64      
  biggestFloat = float64  

Enumeration and object types cannot be defined on the fly, but only within a type statement.

A variable of an enumeration type can only be assigned a value of a limited set. This set consists of ordered symbols. Each symbol is mapped to an integer value internally. The first symbol is represented at runtime by 0, the second by 1 and so on. Example:

  Direction = enum
    north, east, south, west

var x = south      

All comparison operators can be used with enumeration types.

An enumeration's symbol can be qualified to avoid ambiguities: Direction.south.

The $ operator can convert any enumeration value to its name, the ord proc to its underlying integer value.

For better interfacing to other programming languages, the symbols of enum types can be assigned an explicit ordinal value. However, the ordinal values have to be in ascending order. A symbol whose ordinal value is not explicitly given is assigned the value of the previous symbol + 1.

An explicit ordered enum can have holes:

  MyEnum = enum
    a = 2, b = 4, c = 89

Enumerations without holes, integer types, char and bool (and subranges) are called ordinal types. Ordinal types have quite a few special operations:

OperationComment ord(x)returns the integer value that is used to represent x's value inc(x)increments x by one inc(x, n)increments x by n; n is an integer dec(x)decrements x by one dec(x, n)decrements x by n; n is an integer succ(x)returns the successor of x succ(x, n)returns the n'th successor of x pred(x)returns the predecessor of x pred(x, n)returns the n'th predecessor of x

The inc, dec, succ and pred operations can fail by raising an EOutOfRange or EOverflow exception. (If the code has been compiled with the proper runtime checks turned on.)

A subrange type is a range of values from an integer or enumeration type (the base type). Example:

  Subrange = range[0..5]

Subrange is a subrange of int which can only hold the values 0 to 5. Assigning any other value to a variable of type Subrange is a compile-time or runtime error. Assignments from the base type to one of its subrange types (and vice versa) are allowed.

The system module defines the important Natural type as range[0..high(int)] (high returns the maximal value). Other programming languages mandate the usage of unsigned integers for natural numbers. This is often wrong: you don't want unsigned arithmetic (which wraps around) just because the numbers cannot be negative. Nim's Natural type helps to avoid this common programming error.

The set type models the mathematical notion of a set. The set's

basetype can only be an ordinal type of a certain size, namely:

or equivalent. The reason is that sets are implemented as high performance bit vectors. Attempting to declare a set with a larger type will result in an error:

var s: set[int64] 

Sets can be constructed via the set constructor: {} is the empty set. The empty set is type compatible with any concrete set type. The constructor can also be used to include elements (and ranges of elements):

  CharSet = set[char]
  x: CharSet
x = {'a'..'z', '0'..'9'} 

These operations are supported by sets:

operationmeaning A + Bunion of two sets A * Bintersection of two sets A - Bdifference of two sets (A without B's elements) A == Bset equality A <= Bsubset relation (A is subset of B or equal to B) A < Bstrong subset relation (A is a real subset of B) e in Aset membership (A contains element e) e notin AA does not contain element e contains(A, e)A contains element e card(A)the cardinality of A (number of elements in A) incl(A, elem)same as A = A + {elem} excl(A, elem)same as A = A - {elem}

Sets are often used to define a type for the flags of a procedure. This is a much cleaner (and type safe) solution than just defining integer constants that should be or'ed together.

An array is a simple fixed length container. Each element in the array has the same type. The array's index type can be any ordinal type.

Arrays can be constructed via []:

  IntArray = array[0..5, int] 
  x: IntArray
x = [1, 2, 3, 4, 5, 6]
for i in low(x)..high(x):

The notation x[i] is used to access the i-th element of x. Array access is always bounds checked (at compile-time or at runtime). These checks can be disabled via pragmas or invoking the compiler with the --bound_checks:off command line switch.

Arrays are value types, like any other Nim type. The assignment operator copies the whole array contents.

The built-in len proc returns the array's length. low(a) returns the lowest valid index for the array a and high(a) the highest valid index.

  Direction = enum
    north, east, south, west
  BlinkLights = enum
    off, on, slowBlink, mediumBlink, fastBlink
  LevelSetting = array[north..west, BlinkLights]
  level: LevelSetting
level[north] = on
level[south] = slowBlink
level[east] = fastBlink
echo repr(level)  
echo low(level)   
echo len(level)   
echo high(level)  

The syntax for nested arrays (multidimensional) in other languages is a matter of appending more brackets because usually each dimension is restricted to the same index type as the others. In Nim you can have different dimensions with different index types, so the nesting syntax is slightly different. Building on the previous example where a level is defined as an array of enums indexed by yet another enum, we can add the following lines to add a light tower type subdivided in height levels accessed through their integer index:

  LightTower = array[1..10, LevelSetting]
  tower: LightTower
tower[1][north] = slowBlink
tower[1][east] = mediumBlink
echo len(tower)     
echo len(tower[1])  
echo repr(tower)    

Note how the built-in len proc returns only the array's first dimension length. Another way of defining the LightTower to show better its nested nature would be to omit the previous definition of the LevelSetting type and instead write it embedded directly as the type of the first dimension:

  LightTower = array[1..10, array[north..west, BlinkLights]]

It is quite frequent to have arrays start at zero, so there's a shortcut syntax to specify a range from zero to the specified index minus one:

  IntArray = array[0..5, int] 
  QuickArray = array[6, int]  
  x: IntArray
  y: QuickArray
x = [1, 2, 3, 4, 5, 6]
y = x
for i in low(x)..high(x):
  echo(x[i], y[i])

Sequences are similar to arrays but of dynamic length which may change during runtime (like strings). Since sequences are resizable they are always allocated on the heap and garbage collected.

Sequences are always indexed with an int starting at position 0. The len, low and high operations are available for sequences too. The notation x[i] can be used to access the i-th element of x.

Sequences can be constructed by the array constructor [] in conjunction with the array to sequence operator @. Another way to allocate space for a sequence is to call the built-in newSeq procedure.

A sequence may be passed to an openarray parameter.


  x: seq[int] 
x = @[1, 2, 3, 4, 5, 6] 

Sequence variables are initialized with nil. However, most sequence operations cannot deal with nil (leading to an exception being raised) for performance reasons. Thus one should use empty sequences @[] rather than nil as the empty value. But @[] creates a sequence object on the heap, so there is a trade-off to be made here.

The for statement can be used with one or two variables when used with a sequence. When you use the one variable form, the variable will hold the value provided by the sequence. The for statement is looping over the results from the items() iterator from the system module. But if you use the two variable form, the first variable will hold the index position and the second variable will hold the value. Here the for statement is looping over the results from the pairs() iterator from the system module. Examples:

for i in @[3, 4, 5]:

for i, value in @[3, 4, 5]:
  echo("index: ", $i, ", value:", $value)

Note: Openarrays can only be used for parameters.

Often fixed size arrays turn out to be too inflexible; procedures should be able to deal with arrays of different sizes. The openarray type allows this. Openarrays are always indexed with an int starting at position 0. The len, low and high operations are available for open arrays too. Any array with a compatible base type can be passed to an openarray parameter, the index type does not matter.

  fruits:   seq[string]       
  capitals: array[3, string]  

fruits = @[]                  

capitals = ["New York", "London", "Berlin"]   

proc openArraySize(oa: openArray[string]): int =

assert openArraySize(fruits) == 2     
assert openArraySize(capitals) == 3   

The openarray type cannot be nested: multidimensional openarrays are not supported because this is seldom needed and cannot be done efficiently.

A varargs parameter is like an openarray parameter. However, it is also a means to implement passing a variable number of arguments to a procedure. The compiler converts the list of arguments to an array automatically:

proc myWriteln(f: File, a: varargs[string]) =
  for s in items(a):
    write(f, s)
  write(f, "\n")

myWriteln(stdout, "abc", "def", "xyz")

myWriteln(stdout, ["abc", "def", "xyz"])

This transformation is only done if the varargs parameter is the last parameter in the procedure header. It is also possible to perform type conversions in this context:

proc myWriteln(f: File, a: varargs[string, `$`]) =
  for s in items(a):
    write(f, s)
  write(f, "\n")

myWriteln(stdout, 123, "abc", 4.0)

myWriteln(stdout, [$123, $"abc", $4.0])

In this example $ is applied to any argument that is passed to the parameter a. Note that $ applied to strings is a nop.

Slices look similar to subranges types in syntax but are used in a different context. A slice is just an object of type Slice which contains two bounds, a and b. By itself a slice is not very useful, but other collection types define operators which accept Slice objects to define ranges.

  a = "Nim is a progamming language"
  b = "Slices are useless."

echo a[7..12] 
b[11..^2] = "useful"
echo b 

In the previous example slices are used to modify a part of a string. The slice's bounds can hold any value supported by their type, but it is the proc using the slice object which defines what values are accepted.

A tuple type defines various named fields and an order of the fields. The constructor () can be used to construct tuples. The order of the fields in the constructor must match the order in the tuple's definition. Different tuple-types are equivalent if they specify fields of the same type and of the same name in the same order.

The assignment operator for tuples copies each component. The notation t.field is used to access a tuple's field. Another notation is t[i] to access the i'th field. Here i needs to be a constant integer.

  Person = tuple[name: string, age: int] 
  person: Person
person = (name: "Peter", age: 30)

person = ("Peter", 30)



var building: tuple[street: string, number: int]
building = ("Rue del Percebe", 13)

var teacher: tuple[name: string, age: int] = ("Mark", 42)
person = teacher

Even though you don't need to declare a type for a tuple to use it, tuples created with different field names will be considered different objects despite having the same field types.

Tuples can be unpacked during variable assignment (and only then!). This can be handy to assign directly the fields of the tuples to individually named variables. An example of this is the splitFile proc from the os module which returns the directory, name and extension of a path at the same time. For tuple unpacking to work you have to use parenthesis around the values you want to assign the unpacking to, otherwise you will be assigning the same value to all the individual variables! Example:

import os

  path = "usr/local/nimc.html"
  (dir, name, ext) = splitFile(path)
  baddir, badname, badext = splitFile(path)
echo dir      
echo name     
echo ext      

echo baddir
echo badname
echo badext

Tuple unpacking only works in var or let blocks. The following code won't compile:

import os

  path = "usr/local/nimc.html"
  dir, name, ext = ""

(dir, name, ext) = splitFile(path)

References (similar to pointers in other programming languages) are a way to introduce many-to-one relationships. This means different references can point to and modify the same location in memory.

Nim distinguishes between traced and untraced references. Untraced references are also called pointers. Traced references point to objects of a garbage collected heap, untraced references point to manually allocated objects or to objects somewhere else in memory. Thus untraced references are unsafe. However for certain low-level operations (accessing the hardware) untraced references are unavoidable.

Traced references are declared with the ref keyword, untraced references are declared with the ptr keyword.

The empty [] subscript notation can be used to derefer a reference, meaning to retrieve the item the reference points to. The . (access a tuple/object field operator) and [] (array/string/sequence index operator) operators perform implicit dereferencing operations for reference types:

  Node = ref NodeObj
  NodeObj = object
    le, ri: Node
    data: int
  n: Node
new(n) = 9

To allocate a new traced object, the built-in procedure new has to be used. To deal with untraced memory, the procedures alloc, dealloc and realloc can be used. The documentation of the system module contains further information.

If a reference points to nothing, it has the value nil.

A procedural type is a (somewhat abstract) pointer to a procedure. nil is an allowed value for a variable of a procedural type. Nim uses procedural types to achieve functional programming techniques.


proc echoItem(x: int) = echo(x)

proc forEach(action: proc (x: int)) =
    data = [2, 3, 5, 7, 11]
  for d in items(data):


A subtle issue with procedural types is that the calling convention of the procedure influences the type compatibility: procedural types are only compatible if they have the same calling convention. The different calling conventions are listed in the manual.

Nim supports splitting a program into pieces with a module concept. Each module is in its own file. Modules enable information hiding and separate compilation. A module may gain access to symbols of another module by the import statement. Only top-level symbols that are marked with an asterisk (*) are exported:

  x*, y: int

proc `*` *(a, b: seq[int]): seq[int] =
  newSeq(result, len(a))
  for i in 0..len(a)-1: result[i] = a[i] * b[i]

when isMainModule:
  assert(@[1, 2, 3] * @[1, 2, 3] == @[1, 4, 9])

The above module exports x and *, but not y.

The top-level statements of a module are executed at the start of the program. This can be used to initialize complex data structures for example.

Each module has a special magic constant isMainModule that is true if the module is compiled as the main file. This is very useful to embed tests within the module as shown by the above example.

Modules that depend on each other are possible, but strongly discouraged, because then one module cannot be reused without the other.

The algorithm for compiling modules is:

This is best illustrated by an example:

  T1* = int  
import B     

proc main() =
  var i = p(3) 


import A  

proc p*(x: A.T1): A.T1 =
  result = x + 1

A symbol of a module can be qualified with the module.symbol syntax. If the symbol is ambiguous, it even has to be qualified. A symbol is ambiguous if it is defined in two (or more) different modules and both modules are imported by a third one:

var x*: string

var x*: int

import A, B
write(stdout, x) 
write(stdout, A.x) 

var x = 4
write(stdout, x) 

But this rule does not apply to procedures or iterators. Here the overloading rules apply:

proc x*(a: int): string = $a

proc x*(a: string): string = $a

import A, B
write(stdout, x(3))   
write(stdout, x(""))  

proc x*(a: int): string = nil
write(stdout, x(3))   

The normal import statement will bring in all exported symbols. These can be limited by naming symbols which should be excluded with the except qualifier.

import mymodule except y

We have already seen the simple import statement that just imports all exported symbols. An alternative that only imports listed symbols is the from import statement:

from mymodule import x, y, z

The from statement can also force namespace qualification on symbols, thereby making symbols available, but needing to be qualified to be used.

from mymodule import x, y, z

from mymodule import nil



Since module names are generally long to be descriptive, you can also define a shorter alias to use when qualifying symbols.

from mymodule as m import nil


The include statement does something fundamentally different than importing a module: it merely includes the contents of a file. The include statement is useful to split up a large module into several files:

include fileA, fileB, fileC

So, now that we are done with the basics, let's see what Nim offers apart from a nice syntax for procedural programming: Part II

Continue reading on