learnxinyminutes-docs/nix.html.markdown

---
language: nix
filename: learn.nix
contributors:
    - ["Chris Martin", "http://chris-martin.org/"]
    - ["Rommel Martinez", "https://ebzzry.io"]
    - ["Javier Candeira", "https://candeira.com/"]
---

Nix is a simple functional language developed for the
[Nix package manager](https://nixos.org/nix/) and
[NixOS](https://nixos.org/).

You can evaluate Nix expressions using
[nix-instantiate](https://nixos.org/nix/manual/#sec-nix-instantiate)
or [`nix repl`](https://nixos.org/nix/manual/#ssec-relnotes-2.0).

```
with builtins; [

  #  Comments
  #=========================================

  # Inline comments look like this.

  /* Multi-line comments
     look like this. */


  #  Booleans
  #=========================================

  (true && false)               # And
  #=> false

  (true || false)               # Or
  #=> true

  (if 3 < 4 then "a" else "b")  # Conditional
  #=> "a"


  #  Integers and Floats
  #=========================================

  # There are two numeric types: integers and floats

  1 0 42 (-3)       # Some integers

  123.43 .27e13     # A couple of floats

  # Operations will preserve numeric type

  (4 + 6 + 12 - 2)  # Addition
  #=> 20
  (4 - 2.5)
  #=> 1.5

  (7 / 2)           # Division
  #=> 3
  (7 / 2.0)
  #=> 3.5


  #  Strings
  #=========================================

  "Strings literals are in double quotes."

  "
    String literals can span
    multiple lines.
  "

  ''
    This is called an "indented string" literal.
    It intelligently strips leading whitespace.
  ''

  ''
    a
      b
  ''
  #=> "a\n  b"

  ("ab" + "cd")   # String concatenation
  #=> "abcd"

  # Antiquotation lets you embed values into strings.
  ("Your home directory is ${getEnv "HOME"}")
  #=> "Your home directory is /home/alice"


  #  Paths
  #=========================================

  # Nix has a primitive data type for paths.
  /tmp/tutorials/learn.nix

  # A relative path is resolved to an absolute path at parse
  # time, relative to the file in which it occurs.
  tutorials/learn.nix
  #=> /the-base-path/tutorials/learn.nix

  # A path must contain at least one slash, so a relative
  # path for a file in the same directory needs a ./ prefix,
  ./learn.nix
  #=> /the-base-path/learn.nix

  # The / operator must be surrounded by whitespace if
  # you want it to signify division.

  7/2        # This is a path literal
  (7 / 2)    # This is integer division


  #  Imports
  #=========================================

  # A nix file contains a single top-level expression with no free
  # variables. An import expression evaluates to the value of the
  # file that it imports.
  (import /tmp/foo.nix)

  # Imports can also be specified by strings.
  (import "/tmp/foo.nix")

  # Import paths must be absolute. Path literals
  # are automatically resolved, so this is fine.
  (import ./foo.nix)

  # But this does not happen with strings.
  (import "./foo.nix")
  #=> error: string ‘foo.nix’ doesn't represent an absolute path


  #  Let
  #=========================================

  # `let` blocks allow us to bind values to variables.
  (let x = "a"; in
    x + x + x)
  #=> "aaa"

  # Bindings can refer to each other, and their order does not matter.
  (let y = x + "b";
       x = "a"; in
    y + "c")
  #=> "abc"

  # Inner bindings shadow outer bindings.
  (let a = 1; in
    let a = 2; in
      a)
  #=> 2


  #  Functions
  #=========================================

  (n: n + 1)      # Function that adds 1

  ((n: n + 1) 5)  # That same function, applied to 5
  #=> 6

  # There is no syntax for named functions, but they
  # can be bound by `let` blocks like any other value.
  (let succ = (n: n + 1); in succ 5)
  #=> 6

  # A function has exactly one argument.
  # Multiple arguments can be achieved with currying.
  ((x: y: x + "-" + y) "a" "b")
  #=> "a-b"

  # We can also have named function arguments,
  # which we'll get to later after we introduce sets.


  #  Lists
  #=========================================

  # Lists are denoted by square brackets.

  (length [1 2 3 "x"])
  #=> 4

  ([1 2 3] ++ [4 5])
  #=> [1 2 3 4 5]

  (concatLists [[1 2] [3 4] [5]])
  #=> [1 2 3 4 5]

  (head [1 2 3])
  #=> 1
  (tail [1 2 3])
  #=> [2 3]

  (elemAt ["a" "b" "c" "d"] 2)
  #=> "c"

  (elem 2 [1 2 3])
  #=> true
  (elem 5 [1 2 3])
  #=> false

  (filter (n: n < 3) [1 2 3 4])
  #=> [ 1 2 ]


  #  Sets
  #=========================================

  # A "set" is an unordered mapping with string keys.
  { foo = [1 2]; bar = "x"; }

  # The . operator pulls a value out of a set.
  { a = 1; b = 2; }.a
  #=> 1

  # The ? operator tests whether a key is present in a set.
  ({ a = 1; b = 2; } ? a)
  #=> true
  ({ a = 1; b = 2; } ? c)
  #=> false

  # The // operator merges two sets.
  ({ a = 1; } // { b = 2; })
  #=> { a = 1; b = 2; }

  # Values on the right override values on the left.
  ({ a = 1; b = 2; } // { a = 3; c = 4; })
  #=> { a = 3; b = 2; c = 4; }

  # The rec keyword denotes a "recursive set",
  # in which attributes can refer to each other.
  (let a = 1; in     { a = 2; b = a; }.b)
  #=> 1
  (let a = 1; in rec { a = 2; b = a; }.b)
  #=> 2

  # Nested sets can be defined in a piecewise fashion.
  {
    a.b   = 1;
    a.c.d = 2;
    a.c.e = 3;
  }.a.c
  #=> { d = 2; e = 3; }

  # Sets are immutable, so you can't redefine an attribute:
  {
    a = { b = 1; };
    a.b = 2;
  }
  #=> attribute 'a.b' at (string):3:5 already defined at (string):2:11

  # However, an attribute's set members can also be defined piecewise
  # way even if the attribute itself has been directly assigned.
  {
    a = { b = 1; };
    a.c = 2;
  }
  #=> { a = { b = 1; c = 2; }; }


  #  With
  #=========================================

  # The body of a `with` block is evaluated with
  # a set's mappings bound to variables.
  (with { a = 1; b = 2; };
    a + b)
  # => 3

  # Inner bindings shadow outer bindings.
  (with { a = 1; b = 2; };
    (with { a = 5; };
      a + b))
  #=> 7

  # This first line of tutorial starts with "with builtins;"
  # because builtins is a set that contains all of the built-in
  # functions (length, head, tail, filter, etc.). This saves
  # us from having to write, for example, "builtins.length"
  # instead of just "length".


  #  Set patterns
  #=========================================

  # Sets are useful when we need to pass multiple values
  # to a function.
  (args: args.x + "-" + args.y) { x = "a"; y = "b"; }
  #=> "a-b"

  # This can be written more clearly using set patterns.
  ({x, y}: x + "-" + y) { x = "a"; y = "b"; }
  #=> "a-b"

  # By default, the pattern fails on sets containing extra keys.
  ({x, y}: x + "-" + y) { x = "a"; y = "b"; z = "c"; }
  #=> error: anonymous function called with unexpected argument ‘z’

  # Adding ", ..." allows ignoring extra keys.
  ({x, y, ...}: x + "-" + y) { x = "a"; y = "b"; z = "c"; }
  #=> "a-b"

  # The entire set can be bound to a variable using `@`
  (args@{x, y}: args.x + "-" + args.y) { x = "a"; y = "b"; }
  #=> "a-b"

  #  Errors
  #=========================================

  # `throw` causes evaluation to abort with an error message.
  (2 + (throw "foo"))
  #=> error: foo

  # `tryEval` catches thrown errors.
  (tryEval 42)
  #=> { success = true; value = 42; }
  (tryEval (2 + (throw "foo")))
  #=> { success = false; value = false; }

  # `abort` is like throw, but it's fatal; it cannot be caught.
  (tryEval (abort "foo"))
  #=> error: evaluation aborted with the following error message: ‘foo’

  # `assert` evaluates to the given value if true;
  # otherwise it throws a catchable exception.
  (assert 1 < 2; 42)
  #=> 42
  (assert 1 > 2; 42)
  #=> error: assertion failed at (string):1:1
  (tryEval (assert 1 > 2; 42))
  #=> { success = false; value = false; }


  #  Impurity
  #=========================================

  # Because repeatability of builds is critical to the Nix package
  # manager, functional purity is emphasized in the Nix language
  # used to describe Nix packages. But there are a few impurities.

  # You can refer to environment variables.
  (getEnv "HOME")
  #=> "/home/alice"

  # The trace function is used for debugging. It prints the first
  # argument to stderr and evaluates to the second argument.
  (trace 1 2)
  #=> trace: 1
  #=> 2

  # You can write files into the Nix store. Although impure, this is
  # fairly safe because the file name is derived from the hash of
  # its contents. You can read files from anywhere. In this example,
  # we write a file into the store, and then read it back out.
  (let filename = toFile "foo.txt" "hello!"; in
    [filename (readFile filename)])
  #=> [ "/nix/store/ayh05aay2anx135prqp0cy34h891247x-foo.txt" "hello!" ]

  # We can also download files into the Nix store.
  (fetchurl "https://example.com/package-1.2.3.tgz")
  #=> "/nix/store/2drvlh8r57f19s9il42zg89rdr33m2rm-package-1.2.3.tgz"

]
```

### Further Reading

* [Nix Manual - Nix expression language]
  (https://nixos.org/nix/manual/#ch-expression-language)

* [James Fisher - Nix by example - Part 1: The Nix expression language]
  (https://medium.com/@MrJamesFisher/nix-by-example-a0063a1a4c55)

* [Susan Potter - Nix Cookbook - Nix By Example]
  (https://ops.functionalalgebra.com/nix-by-example/)
  
* [Rommel Martinez - A Gentle Introduction to the Nix Family]
  (https://web.archive.org/web/20210121042658/https://ebzzry.io/en/nix/#nix)