Learning Go

Assumptions

It is assumed you already know some Unix and Python and are familiar with Conda.

Installing Go

We use Conda to manage our golang distribution. To create a new environment for Go development start in base and then make a new enviornment:

conda create --name godev
conda activate godev
mamba install -c conda-forge go

$HOME/go

By default, Go will create a go directory in your home directory with the following directory structure.

go/
	bin/
	pkg/

If you go install programs, they will be copied to ~/go/bin. So it's a good idea to have that in your $PATH. If you followed the KorfLab/setup, it already is in your $PATH.

You will see some suggestions to put a src directory inside your go directory and do development there, but we will be using our Code directory as usual.

Hello Go

Start by making a new directory in your Code directory.

cd ~/Code
mkdir hellogo

Initialize the module by providing its location. This creates a file called go.mod whose contents are used for Go administrative tasks. We will see more a bit later.

go mod init github.com/iankorf/hellogo

Now create a really simple hello world program called helloworld.go. Type it in exactly as below (for reasons that will be explained later).

// A program to demonstrate go
package main
import "fmt"
func main() {
fmt.Println("Hello, world")
}

To run this program on the command line, type the following:

go run helloworld.go

As you can see, this works sort of like a python3 hello.py except that you have to tell go that you want to run. There are a bunch of other commands you can use. For example, build creates an executable.

go build

Now look at the directory. There is a new executable program called hellogo. Note that the name of the program is the name of the project directory and not the source file. The program you built is a standalone executable that can be run by itself.

./hellogo

What you might not appreciate until now is that go run did 3 things:

compiled the program
ran the program
deleted the program

When you do a go build it only performs step 1. To clean up your code directory, for example, before doing a git push, use the go clean command. Try that now.

go clean

Your program, hellogo is now deleted. Most of the time when you are developing software, you will probably use go run. Later, once you have a working program that you want to deploy, you can go install it.

go install

It doesn't look like anything happened. Hoever, your program was compiled and copied to ~/go/bin. Take a look.

ls ~/go/bin

If your $PATH is set up properly, you will be able to type hellogo from anywhere on your system.

cd
hellogo

If this doesn't work, please set up your environment properly.

Recall that your helloworld.go program looks rather ugly. There are no vertical whitespaces or indentation. The go community is pretty rabid about consistent style and provides go fmt to make sure your code follows that style.

cat helloworld.go
go fmt
cat helloworld.go

Note that the initial comment did not get separated from the package statement. That's because comments that are directly in front of top-line statements are special: they are documentation. Here's a taste of that.

go doc

Go commands

Frequently used

go run prog.go: compile and run your program

Sometimes used

go doc: get documentation
go fmt: reformat code before git push
go install: compile and install program in ~/go/bin
go test: run your unit tests

Infrequently used

go mod init: create a new mod.go file
go mod tidy: update your mod.go file

Not usually needed

go build compile executable
go clean remove executable

Operators

The Go operators are mostly the same as Python. Like many languages, Go offers ++ and -- for increasing or decreasing by one. Very unlike Python, Go uses * and &, like C for pointers and references. More on this later. There is also a shortcut assignment operator := that is used very frequently.

Variables

Unlike Python, which uses function scope, Go uses lexical scope. Thank god.

Unlike Python, Go variables are all typed. When you create a variable, you must also assign its type. There are a lot of flavors of integers based on their size, and whether or not they are signed or not. Some of these have convenient names like byte (which is like the char type in C) and rune (which is used for unicode). There are also two flavors of floats. Most of the time, you will simply use int for integers and float64 for floats.

var b bool          // boolean
var c byte          // unsigned 8-bit integer
var r rune          // 32-bit integer
var i int           // integer
var f float64       // floating point
var s string        // string
fmt.Println(b, c, r, i, f, s)

Variables that aren't assigned an initial value, like those above, are automatically set to something zero-ish (strings are empty, bools are false).

Although the code above shows the use of var to declare variables, this is an uncommon usage. Most of the time we use the := operator to assign a variable both a type and value. This variable type is generally implicit from the assignment.

j := 1               // int
x := 1.0             // float64
t := "hello world"   // string

Variables can be created and assigned in a list context. Since functions can return multiple values (see below), this syntax is quite common.

k, v := "key", 1.0   // string, float64

Variables can be declared as a constant (read-only). Here, you use the = operator even though you are doing assignment and declaration at the same time.

const AUTHOR = "Ian"
const LINE = 80

Strings in Go are immutable, just like in Python. However, you can make reassignments that sort of look like you are making changes.

var s string // declare an empty string
s += "A"
s += "C"
fmt.Println(s)

Go has a slice operator very much like Python. You can use blanks for implicit start and end, but unlike Python you cannot use negative indices or indicies that are out of bounds.

seq := "ACGT"
fmt.Println(seq[0:])   // ACGT
fmt.Println(seq[0:1])  // A
fmt.Println(seq[0])    // 65 - fmt.Println() does ASCII with singles

Conditonals

There is no elif in Go. Go if-else syntax is mostly like C.

if condition1 {
	stuff1
} else if condition2 {
	stuff2
} else {
	whatever
}

Unlike Python, Go has a switch statement. The Go switch is a little different from C in that you don't need break statements because there is no fall-through. Use it if you like, but there's nothing wrong with chaining if-else.

switch j {
case 0:
	fmt.Println("zero")
case 1:
	fmt.Println("one")
default:
	fmt.Println("other")
}

Loops

The Go for loop has a few different forms.

seq := "ACGT"

// C-like syntax
for i := 0; i < len(seq); i++ {
	fmt.Println(i, seq[i])
}

// Python-like syntax for containers
for i, v := range(seq) {
	fmt.Println(i, v)
}

// while-like syntax
i := 0
for i < 3 {
	fmt.Println(i)
	i++
}

for {} // endless loop

Arrays and Slices

Both Python and Go have fixed-length arrays that we never use, so let's ignore them. In Python, dynamic arrays are called lists and in Go they are called slices. I find it confusing that the word "slice" is used both for an operator and a dynamic array. As a result, I'm just going to use "array" to mean slices.

Arrays can be created with several different syntaxes.

var a []int            // empty array with initial size 0
b := make([]int, 0)    // empty array with initial size 0
c := make([]int, 3)    // array containing 3 zeroes
d := make([]int, 3, 4) // as above, but a hard limit of 4 values
e := []int{1, 2, 3}    // array containging [1, 2, 3]

Even though arrays (ok, actually slices) are dynamic, you can set the capacity as shown in d above. All of the other arrays can grow infinitely. If you want to append to an array, you use the append() function. It looks a little different from Python with the array on both sides of the function call. The reason for this is because the memory location may have to change in order to accomodate resizing. It's not as pretty as Python.

a = append(a, 4)

As you might expect, there is a strings.Join() to turn arrays into strings and a strings.Split() to turn strings into arrays.

Maps / Dictionaries

Maps are like Python dictionaries or Perl hashes. I don't know why there is the make() syntax when the composite literal is easier.

m := make(map[string]int, 0) // empty map
n := map[string]int{}        // empty map
o := map[string]bool{        // initialized map
	"go": true,
	"py": false,
}

Like Perl, but not Python, map keys have automagic assignment. For example, imagine you are counting k-mers:

seq := "ACGATATCGGATATCT"
count := map[string]int{}
k := 3
for i := 0; i < len(seq) -k +1; i++ {
	kmer := string(seq[i:i+k])
	count[kmer] += 1    // assumes existence of key with value of 0
}

Unlike modern versions of Python, keys come out in random-ish order.

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

To delete keys, there is a delete() function.

Functions

Go function declaration is a little different from Python because you have to state the return value. In the example below, the max() function takes an array of integers as input and returns an integer as output. Note the use of the throwaway variable _ in the range() function.

func max(vals []int) int {
	m := vals[0]
	for _, v := range(vals[1:]) {
		if v > m {
			m = v
		}
	}
	return m
}

Go functions can also return multiple values. Here, you have to put parentheses around the multiple return values in the function declaration (but not in the return).

func minmax(vals []int) (int, int) {
	min := vals[0]
	max := vals[0]
	for _, v := range(vals[1:]) {
		if v > max {
			max = v
		}
		if v < min {
			min = v
		}
	}
	return min, max
}

Unlike Python, you cannot mix positional and named arguments. In Go, there are only positional arguments. If you have a function with really complex inputs and outputs, pass structs. If you want to pass a variable number of arguments, like fmt.Println(), you can make a variadic fuction, but should you? Given that it's not explained here... probably not.

Note: something about closures and recursion?

Pointers and References

One of the big differences between Go and Python is references and pointers. The address-of operator, & provides the memory address of a variable.

i := 1
fmt.Println(i, &i) // 1 0xc000014110

The variable i has a value of 1, but its memory address is that weird hexidecimal thing. You can store the memory address of a variable in another variable. For example, in the code below iptr contains the address of i. What type of variable holds the memory address of another variable? A pointer. It is common for pointers to have some kind of ptr suffix.

var iptr *int = &i
fmt.Println(iptr) // some weird hex value

Every variable type is paired with a pointer specific to that type. You cannot use *p for both pointers to integers and pointers to floating points, which is why people don't use p as the name of the variable.

Where you see pointers and references used most frequently is when you want to pass a variable to a function and have the function modify that variable. Does it sound dangerous to have functions with side-effects that might change the input arguments? It is, which is why it should generally be avoided. However, there are times when this is really useful (e.g. recursion).

The following function takes a pointer to integer, *int, as an argument and modifies the value by dereferencing the pointer with the * operator. Yes, this is also the multiplication operator, but the context is different as there is one operand, not two. Note that the function has no return value.

func modi(iptr *int) {
	*iptr++
}

Let's see what happens when we call it.

i := 1
fmt.Println(i) // 1
modi(&i)
fmt.Println(i) // 2

The value at the underyling memory address is incremented by one.

Structs and Methods

Go has several object-oriented programming features. I don't really love inheritence, encapsulation, and such, but I do like method syntax. Using structs and methods one can do struct-oriented programming, which I'm calling SOP to align with "standard operating procedures". Meaning: use structs and methods but think twice about interfaces and embeddings. The Go struct is very much like C except that you don't have to typedef it to give it a proper name.

type fasta struct {
	Def string    // definition line
	Seq string    // letters of sequence
}

Every struct should have a constructor that returns a pointer to the struct. The name of the constructor should begin with "new" and then follow with a capital letter version of the struct name as shown in newFasta() below. You don't have to write destructors because Go does garbage collection for you.

func newFasta(d string, s string) *fasta {
	p := fasta{Def: d, Seq: s}
	return &p
}

Many of your structs will be defined in libraries. In order for other packages to read your struct properties, their names must begin with a captial letter. This is a bit different from every other language. If you want to keep something private, start it with a lowercase letter.

type fasta struct {
	Def string    // exported definition
	Seq string    // exported sequence
	index int     // some private indexing thing
}

OOP-style function calls are supported by methods whose receiver type matches the pointer-to-struct type. Let's say we want to provide a pretty-printer for our fasta struct.

func (fa *fasta) Pretty() {
    fmt.Print(">")
    fmt.Println(fa.Def)
    fmt.Println(fa.Seq)
}

We can now use the convenient object syntax.

s := newFasta("Ian", "ATAGCGAAT")
s.Pretty()

Now that you know about methods, you may be thinking that you should provide getters and setters to encapsulate your data. Nah, just capitalize your struct fields and access them directly.

s.Seq = "ACGT"     // yes
fmt.Println(s.Seq) // yes

s.SetSeq("ACGT")        // no
fmt.Println(s.GetSeq()) // no

Error Handling

The error handling mechanism in Go is particularly clean. There is no try-catch. Instead, functions return errors as their final argument. Let's look at an example. Suppose you want a function that checks if an array of probabilities sums to 1.0 or close enough. The function should return true or false. However, if someone puts a negative number in there, the function should report some kind of error. The panic() function kills the program and reports the error.

func sumToOne(prob []float64, tol float64) (bool, error) {
	s := 0.0
	for _, p := range(prob) {
		if p < 0 {
			return false, errors.New("error: negative number")
		}
		s += p
	}
	if math.Abs(s -1.0) < tol {
		return true, nil
	} else {
		return false, nil
	}
}

func main() {
	p := []float64{0.5, 0.6, -0.1}
	if r, e := sumToOne(p, 0.001); e != nil {
		panic(e)
	} else {
		fmt.Println(r, e)
	}
}

Unfinished

Need to flesh out the following sections

Random Numbers
Sorting
Commandline Arguments - flag
File I/O - line, CSV, JSON
OS - path, env, exec, pipe, tempfile
Testing
Benchmarking
Multithreading
Regular expressions

KorfLab/learning-go