Go Maps

External

• https://go.dev/ref/spec#Map_types
• https://go.dev/blog/maps

Internal

• Go Language
• Go Slices
• Go Sets

Overview

A map is an unordered collection of key/value pairs. The map implementation uses the key as an index and retrieves the associated value in O(1). In order to store a key/value pair in the map, the map hash function is applied to the key, which produces the index of a bucket. The purpose of the hash function is to generate the index that is evenly distributed across the bucket population. Once a bucket is selected, the map stores the key/value pair in the bucket. In order to retrieve the value corresponding to a key, the same hash function is applied to select the bucket, and then the map iterates over the keys stored in that bucket.

Only comparable types can be keys. For more details see Key Identity below.

Like slices, maps hold references to the underlying data structures. The map instances must be initialized before attempting to use into them, either by using the make() function or a composite literal, otherwise you will get a runtime error, because a zero value for a map is nil, which panics when it is written.

The Go documentation used to refer to maps as reference types, but not anymore. The "reference type" terminology was removed from Go documentation.

Maps and Pass-by-Value

Go uses pass-by-value, so when a map argument is passed to a function, the internal fields are copied across on the function stack, including the pointer to the underlying storage. Therefore, the underlying storage data structure is intrinsically shared, even in case of pass-by-value. Both the original map header and the copy of the header passed to the function describe the same underlying storage. If the function changes the storage via its copy of the map, the modified elements can be seen outside the function through the original map variable.

m := map[string]int{"A": 1} // map[A:1]
changeMap(m)
fmt.Println(m) // map[A:2]

func changeMap(m map[string]int) {
     m["A"] = 2
}

The map variables are small, so their content can be copied fast. The fact that they include a pointer to the underlying storage prevents copying the storage, which has important performance implications.

Note that slices are not reference variables, as Go does not support reference variables, so no two different maps variables may point to the same map instance.

Key Identity

The key can be of any comparable type: a type for which the equality operator == is defined: integer, floating point and complex number, string, pointer, interface (as long as the dynamic type supports equality), structs and arrays. Slices, functions and struct types that contain slices cannot be used as map keys because equality is not defined on them. An attempt will generate a compiler error.

The identity of a key in the map is given by its value and its type. We can use the same underlying value, but if the type is different, the keys will count as different. Two different aliases of the same type count as two different types.

type A int
type B int

m := map[interface{}]string{}
m[A(1)] = "red"
m[B(1)] = "blue"
fmt.Printf("%v", m)

will display:

map[1:red 1:blue]

This behavior is used to implement a technique that prevents key collisions when context.Context is used to store request-scoped key/value pairs.

Also see:

Identity, Equality, Comparability in Go

nil and Empty Map

nil Map

Declaring a map with the long declaration and without initialization yields a nil map:

var mm map[string]int

A nil map is also returned by the new() function, but new() returns the pointer to the map, not the map handle itself.

var mmp = new(map[string]int)

fmt.Printf("%p", mm) prints 0x0. The underlying data structures of the map handle are initialized with zero values.

A nil map behaves like an empty map when reading. An attempt to read a key with the indexing operator [] operator from a nil map will not panic, but it will return the zero value for the map type.

v, ok := mm["A"] // v is 0 and ok is false

However, an attempt to write a nil map will cause a runtime panic:

m["A"] = 1

will result in:

panic: assignment to entry in nil map

A non-nil map can be obtained by initializing the map with make() or with a composite literal.

Test nil Map

var mm map[string]int // nil map
if mm == nil {
  // nil map
}

Empty Map

An empty map has zero elements, but it is different from a nil map in that the underlying storage is allocated. For an empty map, fmt.Printf("%p", mm) prints a non-zero pointer.

An empty map can be initialized with an empty composite literal:

mm := map[string]int{}

This declaration and initialization is equivalent with using make():

mm := make(string]int)

Declaration and Initialization

A map can be declared with the long variable declaration.

var mapVarName map[KeyType]ValueType

var mm [string]int

This declaration creates a nil map. Because nil maps cannot be written, this form has a limited usefulness.

Inside functions, the short variable declaration can be used with make() or a composite literal.

Initialization with make()

make() allocates a the underlying data structures required by the map, and creates a new map header to describe it, all at once. Note that unlike new(), make() returns the map, not a pointer. The returned map is empty, but ready to use.

mm := make([string]int)

Initialization with a Composite Literal

Composite literals can be used for map initialization.

Long variable declaration and initialization with type inference:

var mm = map[string]int{"A": 1, "B": 2}

Short variable declaration:

mm := map[string]int{"A": 1, "B": 2}

Operators

Indexing Operator []

Values stored in a map can be referenced with the indexing operator [], using the [key] syntax.

The indexing operator returns two values: a copy of the value associated with the key, or the zero-value for the map type if the key does not exist in map, and a boolean that says whether the key exists or not in the map. This is called the "comma ok" idiom.

The second returned value can be syntactically omitted, but given the fact that the indexing operator returns the zero value for the type and not nil on missing keys, it is impossible to say without it whether the key is missing, or a zero-value was associated with the key.

mm := map[string]int{"A": 1, "B": 2}
v, ok := mm["A"]
fmt.Printf("%d, %t\n", v, ok) // will display 1, true

Idiom:

if v, ok := mm["A"]; ok {
  // use the value
}

If the key does not exist in map, the retuned value is not nil, but the zero value for the type.

The indexing operator can be used to change the value associated with a key, or to add a new key/value pair:

mm["A"] = 7
mm["C"] = 9

Map Functions

delete()

delete() is a built-in function that deletes a key/value pair from the map:

mm := map[string]int{"A": 1, "B": 2}
delete(mm, "A")

If the key does not exist in the map, delete() is a no-op, and safe.

len()

len() is a built-in function that returns the number of keys in the map. Note that the built-in cap() function does not work on maps. Unlike slices, maps do not have the concept of capacity.

make()

make() is a built-in function that creates ready-to-use map. See Initialization with make() above.

new()

new() will initialize a nil map, but it will return the pointer to the map, instead of the value.

A map returned by new() can be positively tested as nil (careful to test the map, not the pointer to it):

mm := new(map[string]int)
fmt.Println(*mm == nil) // true

Iterating over a Map

Use the range keyword.

for k, v := range mm {
    fmt.Printf("%s=%d\n", k, v)
}

If we only want to iterate over keys and the values are not needed, this syntax is valid:

for k := range mm {
  fmt.Printf("key: %s\n", k)
}