Skip to the content.


Getting Started with Go course notes by Oliver Frolovs, 2020.

Setting up the environment

I use Ubuntu 20.04 LTS under WSL and the Go version in Ubuntu apt repo was too old for my liking.

Installing Go

I followed the Download and Install instructions on the Go website to install a more recent version.

$ curl -O
$ sudo tar -C /usr/local -xzf go1.15.2.linux-amd64.tar.gz
$ export PATH=$PATH:/usr/local/go/bin

I made a Go workspace ($GOPATH), although see the note below on modules.

$ mkdir ~/go

I’ve also updated my interactive shell configuration file .bashrc:

# Go as installed using official installation method
if [ -x /usr/local/go/bin/go ]; then
    # Go binaries
    export PATH=$PATH:/usr/local/go/bin

    # Go workspace
    export GOPATH=~/go

Some points for further investigation:

Setting up Visual Studio Code

I’ve added VSCode Go extension to my WSL Remote Target. I would like to use modules, so following the advice given during installation, I’ve enabled Go Language Server option by setting go.useLanguageServer to true in Go extension settings. This has prompted to install gopls to which I have agreed.

Once I opened some .go file, VSCode has asked me to install the missing modules. I responded with Install All.

I was able to run test in one of the [REDACTED] unit tests, which I accept as a sign of success in setting up the environment.

Code organisation

I started out using Go workspace with the intention to move to module-based workflow later.

My Go workspace ($GOPATH) is set to ~/go and structured according to the official convention into:

Say, I wanted to have a Go “project” for experimentation. I name it go-study and place it in $GOPATH/src/go-study/. There is a file main.go there, with the following code.

package main

import (

func main() {
	fmt.Println("Hello, Gruffalo!")
	fmt.Printf("%f", math.Trunc(1.98))

This way, when I add a new entry to import, VSCode does not complain about missing libraries, so I guess it’s OK?

Update: actually, I don’t even need to add anything to import because IntelliSense manages the import based on what’s actually used in the .go file. The import statement gets updated on file save. Neat.

Test Go installation

Separately from VSCode, I test if Go runs from the WSL Windows Terminal.

$ go version
go version go1.15.2 linux/amd64


This concludes the yak shaving part and I can finally begin the course.

Why Go

User-defined types

Group together data and code. Go does not use the term class. Instead, it has struct.



Most, but not all, motivation stems from performance limits.

Moore’s Law doesn’t work anymore. Can’t crank up the clock anymore — power constraints are real. We are at the limits of air cooling. Solution? Increase the number of cores.

Concurrent programming enables parallelism.

Concurrency in Go

Go has a lot of concurrency primitives built into the language.

These are the high level yet basic keywords in the language. The concurrency is built into the language.

Code organisation in Go


A workspace is a directory. There is typically a hierarchy of directories inside the workspace.

Recommended to have inside the workspace directory:

The $GOPATH may be defined by the Go installer, but don’t rely on it. The recommended installation method on Linux un-tar-ring into /usr/local/ certainly cannot do that.


A package is a group of related source code files. Each package can be imported by other packages.

The import keyword is used to access other packages. The standard library include many packages. The $GOROOT and $GOPATH directories are searched for packages.

import "fmt"
import (

There must be one package named main. When it’s build (compiled), an executable is made. When other packages are built, an executable is not made.

package main
import "fmt"
func main() {
    fmt.Printf("Hello, Gruffalo!\n")

FIXME I suspect that packages are searched differently in the context of using modules.

The go tool

The tool can do many different things.

Variable declarations

Follow the format: var name type

var x int
var s string

Basic types: integers, floats, strings.

Type declarations

Aliases can be defined for type names. Useful for writing self-documenting code.

type Celsius float64
type IDNum int

Then variables can be declared using these aliases.

var growboxTemp Celsius
var pid IDNum

Go linters don’t like underscores ( _ ) in variable names. Use capitalisation instead: tempDiff, not temp_diff.

Variable initialisation in Go

Initialise after declaration

We’ve seen that many times before in C.

var x int
x = 100

Initialise in declaration

The literal constant 100 is an integer, so the type is inferred as integer.

var x = 100

Type inference is optional. The type can be specified as well.

var x int32 = 100

Uninitialised variables have a zero value. I suppose the zero value is different for each type.

var x int    // x == 0
var s string // s == ""

TODO what’s the zero value of a custom type?

Short variable declaration

Declare and initialise in one line using the := operator.

x := 100

The variable type is inferred from the expression on the right hand side. Can do this only inside a function.

Pointers in Go

A pointer value is an address in memory. So, a pointer points to some data in memory.

Operations on pointers:

They are the opposites of one another.

var x int = 1
var y int

var ip *int   // ip is a pointer to int

ip = &x       // ip now points to x
y  = *ip      // y is now 1

Now y has value 1.

FIXME What if x is assigned a different value now?

Creating variables with new()

The new() function creates a variable and returns a pointer to it. The variable is initialised to zero by default.

ptr := new(int)
*ptr = 42

A variable of type int is created and the pointer to it is returned. The pointer is then dereferenced, that is the value 42 is put into the memory location pointed to by ptr.

Variable scoping

In Go, variable scoping is done using blocks — they are a sequence of declarations and statements within matching brackets, {}

    // scope A
        // scope B
        // variables declared in scope A are also accessible
    // scope A only

There is also a hierarchy of implicit blocks:

The main point is that there is a hierarchy of blocks. At each level the block can have its own environment of variables associated with it.

Using the set notation, {clauses in switch or select} ⊆ {if, for, switch} ⊆ File ⊆ Package ⊆ Universe.

Lexical scoping

This defines how variables references are resolved. Go is lexically scoped using blocks.

I’ll just say that it’s “intuitive” based on the model I learned from C <3 – go to the greater including scope, until the variable definition is found.

Traditional view on memory management

Stack vs heap

Traditionally, the stack is where the function local variables are allocated. They are deallocated once the function completes (returns).

The heap is a persistent area of memory which does not get deallocated automatically. Memory on the heap must be allocated and deallocated manually when it’s no longer needed.

In C <3:

x = malloc(32);
// the variable x has a life 
// full of adventure and suffering...

It’s error-prone but fast.

Go does things differently. It has garbage collection.

Garbage collection

When using pointers, it’s very hard to figure out when a variable is no longer in use so it can be deallocated safely.

Example in Go, returning the pointer to a local variable. This is not legal in C, but Go can do it:

func foo() *int {
    x := 1
    return &x

func main() {
    var y *int
    y = foo()
    fmt.Printf("%d", *y)

This is possible thanks to garbage collection.

Go is a compiled language which enables garbage collection!

FIXME Does GC pause the program?! I remember Azul Java GC does not, so it’s possible.

Comments in Go

Like in C <3

Single-line comments

// Mamka Tvoya
var x int // Боже Праведный!

Block comments

  Mamka Tvoya
  Иисус Христос был русским и родился в Твери.
  2020 (с) Православная Тверь

var x int


Using fmt package:

String concatenation

x := "Gruffalo"
fmt.Printf("Hello, " + x)

Format strings

Very similar to C <3 Use a conversion character for each argument.

x := "Gruffalo"
fmt.Printf("Hello, %s", x)

Integers in Go

Generic integer declaration is int:

var x int

But there is a wide variety of integer types, based on width in bits and sign:

Finally (!) some sensible names, no more short, and long long, haha.

Usually it’s best to declare as int and leave it to the compiler to figure out, unless there is a reason to control the size manually.

Operations on integers

As one would expect in C <3

Type conversions in Go

Type conversions are not always possible. The following would fail (thankfully).

var x int32 = 1
var y int16 = 2

x = y

The compiler views int32 and int16 as different types so the assignment would fail.

T() operator

Convert type with T() operator.

var x int32
var y int16 = 2
x = int32(y)

Converting int16 to int32 is always possible — it sign-extends the value in y.

Not all conversions are this easy. FIXME more information?

Floating point numbers

Floating-point numbers can be expressed as decimals or using scientific notation.

var g float64 = 9.81           // acceleration of free fall on Earth [m/s^2]
var h float64 = 6.62607e-34    // Plank's constant [J.s]

Complex numbers in Go

Rejoice, mortals! Natively supported, the complex numbers are.

var z complex128 = complex(2,3)

FIXME is this some kind of constructor? But Go doesn’t have them, right?!

ASCII and Unicode

Characters are not the same as bytes (and they haven’t been for a long time).

Unicode characters are 32-bit, so can represent a lot more characters.

UTF-8 is a variable Unicode encoding which matches ASCII for lower 2^8 code points. It is the default encoding in go.

Code point is the Unicode term for a character. There can be 2^32 code points. In Go, the code point is called a rune.

Strings in Go

String literals

Are notated by double quotes.

x := "Gruffalo"

Bytes are not runes!

The string in the following example is made of Cyrillic runes. Each of these runes takes more than a single byte. Thus, byte-addressing the string prints rubbish and not the corresponding runes.

x := "Православная Тверь!\n"
fmt.Printf("%c", x[0])
fmt.Printf("%c", x[3])

Outputs ASCII symbols Ѐ, instead of runes П and в.

String packages in Go

Working at individual rune level

Given that bytes are not equivalent to runes, for working at the runes level (of abstraction), the Unicode package treats strings as “arrays of runes”.

Working at the whole string level

There is also the strings package. It provides functions to manipulate UTF-8 encoded strings. It does not look at individual runes, but at the whole string instead.

String manipulation

Strings are immutable, but the strings package provides some functions to create new strings based on some existing string.

Another useful package is strconv containing conversions to and from other basic datatypes.

Constants in Go

A constant is an expression whose value is known at compile time. A type is inferred from the right-hand side of the assignment: boolean, string, number.

const x = 1.3
const (
    y = 4
    z = "Gruffalo"

Enumerations in Go

Just like an enumerated type in other languages, like in C <3

FIXME Why did they pick up an unfamiliar name?!

type Grades int

const (
    A Grades = iota

In these example, we want five different grades, represented by integers, but we don’t care what actual value each grade has, we just want something different.

We don’t have to repeat the type and iota, it is implied! Each constant is assigned to a unique integer.

Control flow in Go

The control flow describes the order in which the statements are executed inside a program.

The most basic control flow

Is executing one statement at a time, one after another, top to bottom in terms of mapping it to the source code file.

Control flow might change from the basic one because the programmer inserted statements which alter the control flow.

Conditional statement

if <condition> {

Unlike C <3, the brackets {} are a must. There is also an optional else clause, as one would expect.


for <init>; <cond>; <update> {

There are several forms of for statements. The three most common ones are:

// Traditional form, seen before
for i:=0; i<10; i++ {

// The for loop itself has no initialisation and no update, only the condition
i = 0
for i < 10 {

// Infinite loop, maybe for an embedded system?
for {

Switch/case statement

switch x {
    case 1:
        // ...
    case 2:
        // ...
        // optional part if no other case match

Note, that break is not necessary to skip other cases, this is not C <3.

Tagless switch

Sometimes one does not want a tag for each case, but an expression instead. The first case for which the expression evaluates to true is executed.

The case contains a boolean expression to evaluate.

switch {
    case x > 1:
        // ...
    case x < -1:
        // ...
        // optional

A tagless switch can be used in place of a sequence of chained if else statements. And this looks good, I must add! Chained if-else never looked right.

Break and continue

Reading user input in Go

Most trivial way

The scan function is the most basic way for reading user input. It is in fmt package.

var appleNum int

fmt.Printf("Number of apples?")

num, err := fmt.Scan(&appleNum)

Note the multiple assignment!

More sophisticated: handling whole lines

If the input is supposed to have whitespace, best to use bufio.

package main

import (

func main() {
    file, err := os.Open("/path/to/file.txt")
    if err != nil {
    defer file.Close()

    scanner := bufio.NewScanner(file)
    for scanner.Scan() {

    if err := scanner.Err(); err != nil {

From StackOverflow

I also liked another answer, with different options considered.

Composite datatypes in Go

Aggregate data types we’ll look at: arrays, slices, maps, and structs.

Arrays in Go

Fixed-length sequence of elements of a chosen type.

Key points:

As in C, indixing starts at 0. Elements are accessed using subscript notation []. Unlike C, the arrays are initialised to zero by default.

var x [5]int

x[0] = 2
fmt.Printf(x[1])  // is going to be zero

Array literals in Go

Is a sequence of pre-defined values that make up an array.

var x [5]int = [5]{1, 2, 3, 4, 5}

The left-hand side declaration is familiar. The right-hand side syntax is new. This is how you describe array literals. I note that the size of the literal is explicitly provided, which is great. Length of the literal must be length of array.

The size of the array can be inferred from the size of array literal using &hellip; keyword

x := [...]int {1, 2, 3, 4}

Iterating through array in Go

Using for loop.

x := [3]int {1, 2, 3}

for i, v range x {
    fmt.Printf("index %d, value %d \n", i, v)

range is a keyword. It returns two values: index and element value at that index.

Slices in Go

A slice is a window on an underlying (possibly larger) array. They can be variable size, up to a the size of the whole array. A slice has three main properties:

arr := [...]string {"a", "b", "c", "d", "e", "f", "g"}

s1 := arr[1:3] // includes elements 1 and 2: [b,c]
s2 := arr[2:5] // includes elements 2,3, and 4: [c,d,e]

Note, that slices can overlap. We use bracket and colon notation with slices. Similar to Python’s slices so far.

Length and capacity

For slices,

FIXME I got into the habit of always adding of an underlying array when I say capacity, at least until I remember the Go terminology.

Referring to slices

When writing to a slice, you are writing into the underlying array. Note, that different slices can be overlapping on the underlying array.

Slice literals

Note, that there is nothing in the brackets [], so the compiler infers that this is a slice. It creates the array and it makes the slice point to the beginning of that array.

sli := []int {1, 2, 3}

FIXME This looks like rather technical shenanigans which have to do with the language being a compiled one? Would like to see some uses of this which is not just like what you do with Python slices (with added technical complexity).

Creating slices

There is a function make() which can be used to create slices (and an underlying array).

The use case for it is when you want to make a slice because you have some data to store and you want to initialise the slice to a particular size in the beginning.

Two argument version: specify type and length or capacity:

// Initialise a slice to zero. Length == capacity == 10.
sli = make([]int, 10) 

Three argument version: specify length and capacity separately. So, the underlying array is bigger than the slice:

// Initialise a slice to zero. Length is 10, but capacity is 15.
sli = make([]int, 10, 15)

Increasing the size of a slice

Use append() to increase the size (length) of slice. It adds element to the end of the slice.

If the size of the underlying array is reached, it will make a new array. There is a time penalty for that.

sli = make([int, 0, 3]) // slice length is zero, but underlying array is size 3
sli = append(sli, 100)  // the underlying array is increased to size 100

Hash tables

Allow fast access to large bodies of data. Contains key/value pairs. Each value is associated with unique key. Hash function is used to compute the slot for a key. All of this is familiar territory.

Tradeoffs of hash tables

Advantages of hash tables over lists:

Disadvantages of hash tables over lists:

Maps in Go

A map is a Golang’s implementation of a hash table.

Creating maps

var idMap map[string]int   // [key type]value type
idMap = make(map[string]int)  
idMap := map[string]int {
    "joe": 123

Accessing maps

Reference a value with [key]. Returns zero if key is not present.

FIXME i don’t like this, zero is a valid integer value, wtf?


Adding a key/value pair

idMap["mouse"] = 456

Deleting a key/value pair

delete(idMap, "joe")

Key existence test

Two-value assignment tests for existence of the key:

id, p := idMap["joe"]

A boolean p would indicate if the key "joe" is in the map.

Number of values in the map

The len function returns the number of key/value pairs in the map.


Iterating through a map

Use two-value assignment with a range keyword.

for key, val := range idMap {
    fmt.Println(key, val)

No surprises here.

Structs in Go

A struct is short for structure and it comes from C <3. It’s an aggregate data type, grouping together values of arbitrary data types.

Example: Person struct. Each person has name, address, and a phone no.

Can define a struct type, analogous to type Celsius int seen before.

type Person struct {
    name string
    addr string
    phone string

var p1 Person // contains data one person
var p2 Person // contains data for another person

Accessing struct fields

Use dot notation. = "Gruffalo"
x := p1.addr

Initialising struct

Use new():

p1 := new(Person)

FIXME Why not make()?

Or use a struct literal:

p1 := Person{name: "Gruffalo", addr: "Deep Dark Wood", phone: "123"}

TODO Revision examples


This example puts together quite a few concepts from Go syntax.

TODO annotate with comments

type P struct {
    x string
    y int

func main() {
  b := P{"x", -1}
  a := [...]P{
        P{"a", 10}, 
        P{"b", 2},
        P{"c", 3}
  for _, z := range a {
    if z.y > b.y {
      b = z

Length and capacity

The following prints 0 3.

append puts the item into the slice, or expands the slice if the need be.

func main() {
  s := make([]int, 0, 3)
  s = append(s, 100)
  fmt.Println(len(s), cap(s))

Protocols and formats in Go

Golang has packages to handle lots of different protocols and data formats.


Not news, just not something I check out a lot in the recent years))

Network protocols packages

net.Dial("tcp", "")

It’s very useful to have these packages ready for use.


A struct in Go:

p1 := Person{name: "Gruffalo", addr: "Deep Dark Wood", phone: "123"}

As a JSON object:

    "name": "Gruffalo",
    "addr": "Deep Dark Wood",
    "phone": "123"

JSON marshalling

Generating JSON representation from an object.

I remember, that marshalling and serialisation aren’t exactly the same thing!

A type definition in Go:

type Person struct {
    name string
    addr string
    phone string

Define a value of that type using field-value initialisers

p1 := Person{name: "Gruffalo", addr: "Deep Dark Wood", phone: "123"}

Note {} in the definition.

Could have used value initialisers only:

p1 := Person{"Gruffalo", "Deep Dark Wood", "123"}

Note, that field-value initalisers, and value initialisers cannot be mixed, that is either all or no field names must be provided.

Marshall into JSON format:

barr, err := json.Marshall(p1)

json.Marshall() returns JSON representation as []byte.

JSON unmarshalling:

Converting a JSON []byte into a Go object.

var p2 Person
err := json.Unmarshall(barr, &p2)

Constraint: p2 must “fit” the JSON []byte, that is must have the same attributes/fields.

File access in Go

Files are still commondly used to trade data between the programmes. File access is linear access (not random access), because originally they come from tapes.

(What a crap abstraction! Obsolete af.)

Basic operations:

Go has ioutil package for basic file-related operations.

Reading files in Go

To read, use ioutil.ReadFile(fname).

Writing files in Go

dat = "Hello, Gruffalo!"

err := ioutil.WriteFile("outfile.txt", dat, 0777)

os package file access

So, ioutil gives some very basic file-related functions. For more granular control, the os package has file-related functions.

Opening and reading example

Using the os package for opening and reading files.

f, err := os.Open("dt.txt")
barr := make([]byte, 10)
nb, err := f.Read(barr)

Writing a file example

Using the os package for creating the file and writing data into it.

f, err := os.Create("outfile.txt")

barr := []byte {1,2,3}
nb,err := f.Write(barr)
nb,err := f.WriteString("Hello, Gruffalo!")  // it has to be Unicode sequence

We could have appended the data to an existing file as well, os has functions for that.

The end

This concludes the course!