To install Go in Linux you have two options:

I recommend using the second and downloading a distribution. Go's updates maintains backward compatibility and you usually should not be worried about updating your Go binaries frequently.

tar -zxvf go*.*.*.linux-amd64.tar.gz
sudo mv go /usr/local/go

mkdir -p $HOME/go/bin

export PATH=$PATH:/usr/local/go/bin

$ go version
Go version go1.6.2 linux/amd64

To install Go in Windows, you will need administrator privileges. Open your favorite browser and navigate to https://golang.org . Once there click the Download Go button and select Microsoft Windows distribution. A *.msi file will start downloading.

Execute the MSI installer by double clicking it. An installer will appear asking you to accept the End User License Agreement (EULA) and select a target folder for your installation. We will continue with the default path that in my case was C:\Go.

Once the installation is finished you will have to add the binary Go folder, located in C:\Go\bin to your Path. For this, you must go to Control Panel and select System option. Once in System, select the Advanced tab and click the Environment variables button. Here you'll find a window with variables for your current user and system variables. In system variables, you'll find the Path variable. Click it and click the Edit  button to open a text box. You can add your path by adding ;C:\Go/bin at the end of the current line (note the semicolon at the beginning of the path). In recent Windows versions (Windows 10) you will have a manager to add variables easily.

In Mac OS X the installation process is very similar to Linux. Open your favorite browser and navigate to https://golang.org and click the Download Go. From the list of possible distributions that appear, select Apple OS X. This will download a *.pkg file to your download folder.

A window will guide you through the installation process where you have to type your administrator password so that it can put Go binary files in /usr/local/go/bin folder with the proper permissions. Now, open Terminal to test the installation by typing this on it:

$ go version
Go version go1.6.2 darwin/amd64

If you see the installed version, everything was fine. If it doesn't work check that you have followed correctly every step or refer to the documentation at https://golang.org.

Go will always work under the same workspace. This helps the compiler to find packages and libraries that you could be using. This workspace is commonly called GOPATH.

GOPATH has a very important role in your working environment while developing Go software. When you import a library in your code it will search for this library in your $GOPATH/src. The same when you install some Go apps, binaries will be stored in $GOPATH/bin.

At the same, all your source code must be stored in a valid route within $GOPATH/src folder. For example, I store my projects in GitHub and my username is Sayden so, for a project called minimal-mesos-go-framework I will have this folder structure like $GOPATH/src/github.com/sayden/minimal-mesos-go-framework which reflects the URI where this repo is stored at GitHub:

mkdir -p $HOME/go

The $HOME/go path is going to be the destination of our $GOPATH. We have to set an environment variable with our $GOPATH pointing to this folder. To set the environment variable, open again the file $HOME/.bashrc with your favorite text editor and add the following line at the end of it:

export GOPATH=${HOME}/go

Save the file and open a new terminal. To check that everything is working, just write an echo to the $GOPATH variable like this:

echo $GOPATH
/home/mcastro/go

If the output of the preceding command points to your chosen Go path, everything is correct and you can continue to write your first program.

Go language, like most programming languages, has if…else conditional statement for flow control. The Syntax is similar to other languages but you don't need to encapsulate the condition between parenthesis:

ten := 10 
if ten == 20 { 
    println("This shouldn't be printed as 10 isn't equal to 20") 
} else { 
    println("Ten is not equals to 20"); 
} 

The else...if condition works in a similar fashion, you don't need parentheses either and they are declared as programmer would expect:

if "a" == "b" ||  10 == 10 || true == false { 
    println("10 is equal to 10") 
  } else if 11 == 11 &&"go" == "go" { 
  println("This isn't print because previous condition was satisfied"); 
    } else { 
        println("In case no condition is satisfied, print this") 
    } 
} 

The switch statement is also similar to most imperative languages. You take a variable and check possible values for it:

number := 3 
switch(number){ 
    case 1: 
        println("Number is 1") 
    case 2: 
        println("Number is 2") 
    case 3: 
        println("Number is 3") 
} 

The _for_ loop is also similar than in common programming languages but you don't use parentheses either

for i := 0; i<=10; i++ { 
    println(i) 
} 

As you have probably imagined if you have computer science background, we infer an int variable defined as 0 and execute the code between the brackets while the condition (i<=10) is satisfied. Finally, for each execution, we added 1 to the value of i. This code will print the numbers from 0 to 10. You also have a special syntax to iterate over arrays or slices which is range:

for index, value := range my_array { 
    fmt.Printf("Index is %d and value is %d", index, value) 
} 

First, the fmt (format) is a very common Go package that we will use extensively to give shape to the message that we will print in the console.

Regarding for, you can use the range keyword to retrieve every item in a collection like my_array and assign them to the value temporal variable. It will also give you an index variable to know the position of the value you're retrieving. It's equivalent to write the following:

for index := 0, index < len(my_array); index++ { 
    value := my_array[index] 
    fmt.Printf("Index is %d and value is %d", index, value) 
} 

If you execute this code, you'll see that the result is the same.

A function is a piece of code with its own variables and flow that doesn't affect anything outside of the opening and close brackets but global package or program variables. Functions in Go has the following composition:

func [function_name] (param1 type, param2 type...) (returned type1, returned type2...) { 
    //Function body 
} 

Following the previous definition, we could have the following example:

func hello(message string) error { 
    fmt.Printf("Hello %s\n", message) 
    return nil 
} 

Functions can call other functions. For example, in our previous hello function, we are receiving a message argument of type string and we are calling a different function fmt.Printf("Hello %s\n", message) with our argument as parameter. Functions can also be used as parameters when calling other functions or be returned.

It is very important to choose a good name for your function so that it is very clear what it is about without writing too many comments over it. This can look a bit trivial but choosing a good name is not so easy. A short name must show what the function does and let the reader imagine what error is it handling or if it's doing any kind of logging. Within your function, you want to do everything that a particular behavior need but also to control expected errors and wrapping them properly.

So, to write a function is more than simply throw a couple of lines that does what you need, that's why it is important to write a unit test, make them small and concise.

An anonymous function is a function without a name. This is useful when you want to return a function from another function that doesn't need a context or when you want to pass a function to a different function. For example, we will create a function that accepts one number and returns a function that accepts a second number that it adds it to the first one. The second function does not have a declarative name (as we have assigned it to a variable) that is why it is said to be anonymous:

func main(){ 
    add := func(m int){ 
         return m+1 
} 
 
    result := add(6) 
 
    //1 + 6 must print 7 
    println(result) 
} 

The add variable points to an anonymous function that adds one to the specified parameter. As you can see, it can be used only for the scope its parent function main and cannot be called from anywhere else.

Anonymous functions are really powerful tools that we will use extensively on design patterns.

Closures are something very similar to anonymous functions but even more powerful. The key difference between them is that an anonymous function has no context within itself and a closure has. Let's rewrite the previous example to add an arbitrary number instead of one:

func main(){ 
    addN := func(m int){ 
        return func(n int){ 
            return m+n 
        }            
    } 
 
    addFive := addN(5) 
    result := addN(6)  
    //5 + 6 must print 7 
    
    println(result) 
}

The addN variable points to a function that returns another function. But the returned function has the context of the m parameter within it. Every call to addN will create a new function with a fixed m value, so we can have main addN functions, each adding a different value.

This ability of closures is very useful to create libraries or deal with functions with unsupported types.

Errors are extensively used in Go, probably thanks to its simplicity. To create an error simply make a call to errors.New(string) with the text you want to create on the error. For example:

err := errors.New("Error example") 

As we have seen before, we can return errors to a function. To handle an error you'll see the following pattern extensively in Go code:

func main(){ 
    err := doesReturnError() 
    if err != nil { 
        panic(err) 
    } 
} 
 
func doesReturnError() error { 
    err := errors.New("this function simply returns an error") 
    return err 
} 

Functions can be declared as variadic. This means that its number of arguments can vary. What this does is to provide an array to the scope of the function that contains the arguments that the function was called with. This is convenient if you don't want to force the user to provide an array when using this function. For example:

func main() { 
    fmt.Printf("%d\n", sum(1,2,3)) 
    fmt.Printf("%d\n", sum(4,5,6,7,8)) 
} 
 
func sum(args ...int) (result int) { 
    for _, v := range args { 
        result += v 
    } 
    return 
} 

In this example, we have a sum function that will return the sum of all its arguments but take a closer look at the main function where we call sum. As you can see now, first we call sum with three arguments and then with five arguments. For sum functions, it doesn't matter how many arguments you pass as it treats its arguments as an array all in all. So on our sum definition, we simply iterate over the array to add each number to the result integer.

An array is a numbered sequence of elements of a single type. You can store 100 different unsigned integers in a unique variable, three strings or 400 bool values. Their size cannot be changed.

You must declare the length of the array on its creation as well as the type. You can also assign some value on creation. For example here you have 100 int values all with 0 as value:

var arr [100]int 

Or an array of size 3 with strings already assigned:

arr := [3]string{"go", "is", "awesome"} 

Here you have an array of 2 bool values that we initialize later:

var arr [2]bool 
arr[0] = true 
arr[1] = false 

Slices are similar to arrays, but their size can be altered on runtime. This is achieved, thanks to the underlying structure of a slice that is an array. So, like arrays, you have to specify the type of the slice and its size. So, use the following line to create a slice:

mySlice := make([]int, 10) 

This command has created an underlying array of ten elements. If we need to change the size of the slice by, for example, adding a new number, we would append the number to the slice:

mySlice := append(mySlice, 5) 

The syntax of append is of the form ([array to append an item to], [item to append]) and returns the new slice, it does not modify the actual slice. This is also true to delete an item. For example, let's delete the first item of the array as following:

mySlice := mySlice[1:] 

Yes, like in arrays. But what about deleting the second item? We use the same syntax:

mySlice = append(mySlice[:1], mySlice[2:]...) 

We take all elements from zero index (included) to the first index (not included) and each element from the second index (included) to the end of the array, effectively deleting the value at the second position in the slice (index 1 as we start counting with 0). As you can see, we use the undetermined arguments syntax as the second parameter.

Maps are like dictionaries--for each word, we have a definition but we can use any type as word or definition and they'll never be ordered alphabetically. We can create maps of string that point to numbers, a string that points to interfaces and structs that point to int and int to function. You cannot use as key: slices, the functions, and maps. Finally, you create maps by using the keyword make and specifying the key type and the value type:

myMap := make(map[string]int) 
myMap["one"] = 1 
myMap["two"] = 2 
fmt.Println(myMap["one"]) 

When parsing JSON content, you can also use them to get a string[interface] map:

myJsonMap := make(map[string]interface{}) 
jsonData := []byte(`{"hello":"world"}`) 
err := json.Unmarshal(jsonData, &myJsonMap) 
if err != nil { 
panic(err) 
} 
fmt.Printf("%s\n", myJsonMap["hello"]) 

The myJsonMap variable is a map that will store the contents of JSON and that we will need to pass its pointer to the Unmarshal function. The jsonData variable declares an array of bytes with the typical content of a JSON object; we are using this as the mock object. Then, we unmarshal the contents of the JSON storing the result of the memory location of myJsonMap variable. After checking that the conversion was ok and the JSON byte array didn't have syntax mistakes, we can access the contents of the map in a JSON-like syntax.

Pointers are hated, loved, and very useful at the same time. To understand what a pointer is can be difficult so let's try with a real world explanation. As we mentioned earlier in this chapter, a pointer is a like a mailbox. Imagine a bunch of mailboxes in a building; all of them have the same size and shape but each refers to a different house within the building. Just because all mailboxes are the same size does not mean that each house will have the same size. We could even have a couple of houses joined, a house that was there but now has a license of commerce, or a house that is completely empty. So the pointers are the mailboxes, all of them of the same size and that refer to a house. The building is our memory and the houses are the types our pointers refer to and the memory they allocate. If you want to receive something in your house, it's far easier to simply send the address of your house (to send the pointer) instead of sending the entire house so that your package is deposited inside. But they have some drawbacks as if you send your address and your house (variable it refers to) disappears after sending, or its type owner change--you'll be in trouble.

How is this useful? Imagine that somehow you have 4 GB of data in a variable and you need to pass it to a different function. Without a pointer, the entire variable is cloned to the scope of the function that is going to use it. So, you'll have 8 GB of memory occupied by using this variable twice that, hopefully, the second function isn't going to use in a different function again to raise this number even more.

You could use a pointer to pass a very small reference to this chunk to the first function so that just the small reference is cloned and you can keep your memory usage low.

While this isn't the most academic nor exact explanation, it gives a good view of what a pointer is without explaining what a stack or a heap is or how they work in x86 architectures.

Pointers in Go are very limited compared with C or C++ pointers. You can't use pointer arithmetic nor can you create a pointer to reference an exact position in the stack.

Pointers in Go can be declared like this:

number := 5 

Here number := 5 code represents our 4 GB variable and pointer_to_number contains the reference (represented by an ampersand) to this variable. It's the direction to the variable (the one that you put in the mailbox of this house/type/variable). Let's print the variable pointer_to_number , which is a simple variable:

println(pointer_to_number) 
0x005651FA 

What's that number? Well, the direction to our variable in memory. And how can I print the actual value of the house? Well, with an asterisk (*) we tell the compiler to take the value that the pointer is referencing, which is our 4 GB variable.

 println(*pointer_to_number) 
5 

A struct is an object in Go. It has some similarities with classes in OOP as they have fields. Structs can implement interfaces and declare methods. But, for example, in Go, there's not inheritance. Lack of inheritance looks limiting but in fact, composition over inheritance was a requirement of the language.

To declare a structure, you have to prefix its name with the keyword type and suffix with the keyword struct and then you declare any field or method between brackets, for example:

type Person struct { 
    Name string 
    Surname string 
    Hobbies []string 
    id string 
} 

In this piece of code, we have declared a Person structure with three public fields (Name, Age , and Hobbies) and one private field (id, if you recall the Visibility section in this chapter, lowercase fields in Go refers to private fields are just visible within the same package). With this struct, we can now create as many instances of Person as we want. Now we will write a function called GetFullName that will give the composition of the name and the surname of the struct it belongs to:

func (person *Person) GetFullName() string { 
    return fmt.Sprintf("%s %s", person.Name, person.Surname) 
} 
 
func main() { 
    p := Person{ 
        Name: "Mario", 
        Surname: "Castro", 
        Hobbies: []string{"cycling", "electronics", "planes"}, 
        id: "sa3-223-asd", 
    } 
 
    fmt.Printf("%s likes %s, %s and %s\n", p.GetFullName(), p.Hobbies[0], p.Hobbies[1], p.Hobbies[2]) 
} 

Methods are defined similarly to functions but in a slightly different way. There is a(p *Person) that refers to a pointer to the created instance of the struct (recall the Pointers section in this chapter). It's like using the keyword this in Java or self in Python when referring to the pointing object.

Maybe you are thinking why does (p *Person) have the pointer operator to reflect that p is actually a pointer and not a value? This is because you can also pass Person by value by removing the pointer signature, in which case a copy of the value of Person is passed to the function. This has some implications, for example, any change that you make in p if you pass it by value won't be reflected in source p. But what about our GetFullName() method?

func (person Person) GetFullName() string { 
    return fmt.Sprintf("%s %s", person.Name, person.Surname) 
} 

Its console output has no effect in appearance but a full copy was passed before evaluating the function. But if we modify person here, the source p won't be affected and the new person value will be available only on the scope of this function.

On the main function, we create an instance of our structure called p. As you can see, we have used implicit notation to create the variable (the := symbol). To set the fields, you have to refer to the name of the field, colon, the value, and the comma (don't forget the comma at the end!). To access the fields of the instantiated structure, we just refer to them by their name like p.Name or p.Surname. You use the same syntax to access the methods of the structure like p.GetFullName().

The output of this program is:

$ go run main.go 
Mario Castro likes cycling, electronics and planes

Structures can also contain another structure (composition) and implement interface methods apart from their own but, what's an interface method?

An interface is something really simple but powerful. It's usually defined as a contract between the objects that implement it but this explanation isn't clear enough in my honest opinion for newcomers to the interface world.

A water-pipe is a contract too; whatever you pass through it must be a liquid. Anyone can use the pipe, and the pipe will transport whatever liquid you put in it (without knowing the content). The water-pipe is the interface that enforces that the users must pass liquids (and not something else).

Let's think about another example: a train. The railroads of a train are like an interface. A train must construct (implement) its width with a specified value so that it can enter the railroad but the railroad never knows exactly what it's carrying (passengers or cargo). So for example, an interface of the railroad will have the following aspect:

type RailroadWideChecker interface { 
    CheckRailsWidth() int 
} 

The RailroadWideChecker is the type our trains must implement to provide information about their width. The trains will verify that the train isn't too wide or too narrow to use its railroads:

type Railroad struct { 
    Width int 
} 
 
func (r *Railroad) IsCorrectSizeTrain(r RailRoadWideChecker) bool { 
    return r.CheckRailsWidth() != r.Width 
} 

The Railroad is implemented by an imaginary station object that contains the information about the width of the railroads in this station and that has a method to check whether a train fits the needs of the railroad with the IsCorrectSizeTrain method. The IsCorrectSizeTrain method receives an interface object which is a pointer to a train that implements this interface and returns a validation between the width of the train and the railroad:

Type Train struct { 
    TrainWidth int 
} 
 
func (p *Train) CheckRailsWidth() int { 
    return p.TrainWidth 
} 

Now we have created a passenger's train. It has a field to contain its width and implements our CheckRailsWidth interface method. This structure is considered to fulfill the needs of a RailRoadWideChecker interface (because it has an implementation of the methods that the interfaces ask for).

So now, we'll create a railroad of 10 units wide and two trains--one of 10 units wide that fit the railroad size and another of 15 units that cannot use the railroad.

func main(){ 
    railroad := Railroad{Width:10} 
 
    passengerTrain := Train{TrainWidth: 10} 
    cargoTrain := Train {TrainWidth: 15} 
 
    canPassengerTrainPass := railroad.IsCorrectSizeTrain(passengerTrain) 
    canCargoTrainPass := railroad.IsCorrectSizeTrain(cargoTrain) 
 
    fmt.Printf("Can passenger train pass? %b\n", canPassengerTrainPass) 
    fmt.Printf("Can cargo train pass? %b\n", canCargoTrainPass) 
} 

Let's dissect this main function. First, we created a railroad object of 10 units called railroad. Then two trains, of 10 and 15 units' width for passengers and cargo respectively. Then, we pass both objects to the railroad method that accepts interfaces of the RailroadWideChecker interface. The railroad itself does not know the width of each train separately (we'll have a huge list of trains) but it has an interface that trains must implement so that it can ask for each width and returns a value telling you if a train can or cannot use of the railroads. Finally, the output of the call to printf function is the following:

Can passenger train pass? true
Can cargo train pass? false

As I mentioned earlier, interfaces are so widely used during this book that it doesn't matter if it still looks confusing for the reader as they'll be plenty of examples during the book.

Testing is very important in every programming language. Go creators knew it and decided to provide all libraries and packages needed for the test in the core package. You don't need any third-party library for testing or code coverage.

The package that allows for testing Go apps is called, conveniently, testing. We will create a small app that sums two numbers that we provide through the command line:

func main() { 
    //Atoi converts a string to an int 
    a, _ := strconv.Atoi(os.Args[1]) 
    b, _ := strconv.Atoi(os.Args[2]) 
 
    result := sum(a,b) 
    fmt.Printf("The sum of %d and %d is %d\n", a, b, result) 
} 
 
func sum(a, b int) int { 
    return a + b 
} 

Let's execute our program in the terminal to get the sum:

$ go run main.go 3 4
The sum of 3 and 4 is 7

By the way, we're using the strconv package to convert strings to other types, in this case, to int. The method Atoi receives a string and returns an int and an error that, for simplicity, we are ignoring here (by using the underscore).

Ok, so let's write a test that checks the correct result of the sum. We're creating a new file called main_test.go. By convention, test files are named like the files they're testing plus the _test suffix:

func TestSum(t *testing.T) { 
    a := 5 
    b := 6 
    expected := 11 
 
    res := sum(a, b) 
    if res != expected { 
        t.Errorf("Our sum function doens't work, %d+%d isn't %d\n", a, b, res) 
    } 
} 

Testing in Go is used by writing methods started with the prefix Test, a test name, and the injection of the testing.T pointer called t. Contrary to other languages, there are no asserts nor special syntax for testing in Go. You can use Go syntax to check for errors and you call t with information about the error in case it fails. If the code reaches the end of the Test function without arising errors, the function has passed the tests.

To run a test in Go, you must use the go test -v command (-v is to receive verbose output from the test) keyword, as following:

$ go test -v
=== RUN   TestSum
--- PASS: TestSum (0.00s)
PASS
ok   github.com/go-design-patterns/introduction/ex_xx_testing 0.001s

Our tests were correct. Let's see what happens if we break things on purpose and we change the expected value of the test from 11 to 10:

$ go test
--- FAIL: TestSum (0.00s)
    main_test.go:12: Our sum function doens't work, 5+6 isn't 10
FAIL
exit status 1
FAIL  github.com/sayden/go-design-patterns/introduction/ex_xx_testing 0.002s

The test has failed (as we expected). The testing package provides the information you set on the test. Let's make it work again and check test coverage. Change the value of the variable expected from 10 to 11 again and run the command go test -cover to see code coverage:

$ go test -cover
PASS
coverage: 20.0% of statements
ok  github.com/sayden/go-design-patterns/introduction/ex_xx_testing 0.001s

The -cover options give us information about the code coverage for a given package. Unfortunately, it doesn't provide information about overall application coverage.

TDD is the acronym for Test Driven Development. It consists of writing the tests first before writing the function (instead of what we did just before when we wrote the sum function first and then we wrote the test function).

TDD changes the way to write code and structure code so that it can be tested (a lot of code you can find in GitHub, even code that you have probably written in the past is probably very difficult, if not impossible, to test).

So, how does it work? Let's explain this with a real life example--imagine that you are in summer and you want to be refreshed somehow. You can build a pool, fill it with cold water, and jump into it. But in TDD terms, the steps will be:

So let's repeat the previous example but with a multiplication. First, we will write the declaration of the function that we're going to test:

func multiply(a, b int) int { 
    return 0 
} 

Now let's write the test that will check the correctness of the previous function:

import "testing" 
 
func TestMultiply(t *testing.T) { 
    a := 5 
    b := 6 
    expected := 30 
 
    res := multiply(a, b) 
    if res != expected { 
        t.Errorf("Our multiply function doens't work, %d*%d isn't %d\n", a, b, res) 
    } 
} 

And we test it through the command line:

$ go test
--- FAIL: TestMultiply (0.00s)
main_test.go:12: Our multiply function doens't work, 5+6 isn't 0
FAIL
exit status 1
FAIL    github.com/sayden/go-designpatterns/introduction/ex_xx_testing/multiply    

0.002s

Nice. Like in our pool example where the water wasn't there yet, our function returns an incorrect value too. So now we have a function declaration (but isn't defined yet) and the test that fails. Now we have to make the test pass by writing the function and executing the test to check:

func multiply(a, b int) int { 
    return a*b 
} 

And we execute again our testing suite. After writing our code correctly, the test should pass so we can continue to the refractoring process:

$ go test
PASS
ok      github.com/sayden/go-design-patterns/introduction/ex_xx_testing/multiply    
0.001s

Great! We have developed the multiply function following TDD. Now we must refactor our code but we cannot make it more simple or readable so the loop can be considered closed.

During this book, we will write many tests that define the functionality that we want to achieve in our patterns. TDD promotes encapsulation and abstraction (just like design patterns do).

Have you realized that we have imported the package encoding/json? Why is it prefixed with the word encoding? If you take a look at Go's source code to the src/encoding folder you'll find many interesting packages for encoding/decoding such as, XML, HEX, binary, or even CSV.

Now something a bit more complicated:

type MyObject struct { 
    Number int 
    `json:"number"` 
    Word string 
} 
 
func main(){ 
    object := MyObject{5, "Packt"} 
    oJson, _ := json.Marshal(object) 
    fmt.Printf("%s\n", oJson) 
} 
$ {"Number":5,"Word":"Packt"}

Conveniently, it also works pretty well with structures but what if I want to not use uppercase in the JSON data? You can define the output/input name of the JSON in the structure declaration:

type MyObject struct { 
    Number int 
    Word string 
} 
 
func main(){ 
    object := MyObject{5, "Packt"} 
    oJson, _ := json.Marshal(object) 
    fmt.Printf("%s\n", oJson) 
} 
$ {"number":5,"string":"Packt"}

We have not only lowercased the names of the keys, but we have even changed the name of the Word key to string.

Enough of marshalling, we will receive JSON data as an array of bytes, but the process is very similar with some changes:

type MyObject struct { 
Number int`json:"number"` 
Word string`json:"string"` 
} 
 
func main(){ 
    jsonBytes := []byte(`{"number":5, "string":"Packt"}`) 
    var object MyObject 
    err := json.Unmarshal(jsonBytes, &object) 
    if err != nil { 
        panic(err) 
    } 
    fmt.Printf("Number is %d, Word is %s\n", object.Number, object.Word) 
} 

The big difference here is that you have to allocate the space for the structure first (with a zero value) and the pass the reference to the method Unmarshal so that it tries to fill it. When you use Unmarshal, the first parameter is the array of bytes that contains the JSON information while the second parameter is the reference (that's why we are using an ampersand) to the structure we want to fill. Finally, let's use a generic map[string]interface{} method to hold the content of a JSON:

type MyObject struct { 
    Number int     `json:"number"` 
    Word string    `json:"string"` 
} 
 
func main(){ 
    jsonBytes := []byte(`{"number":5, "string":"Packt"}`) 
    var dangerousObject map[string]interface{} 
    err := json.Unmarshal(jsonBytes, &dangerousObject) 
    if err != nil { 
        panic(err) 
    } 
 
    fmt.Printf("Number is %d, ", dangerousObject["number"]) 
    fmt.Printf("Word is %s\n", dangerousObject["string"]) 
    fmt.Printf("Error reference is %v\n",  
dangerousObject["nothing"])
} 
$ Number is %!d(float64=5), Word is Packt 
Error reference is <nil> 

What happened in the result? This is why we described the object as dangerous. You can point to a nil location when using this mode if you call a non-existing key in the JSON. Not only this, like in the example, it could also interpret a value as a float64 when it is simply a byte, wasting a lot of memory.

So remember to just use map[string]interface{} when you need dirty quick access to JSON data that is fairly simple and you have under control the type of scenarios described previously.

A linter analyzes source code to detect errors or improvements. The golint linter is available on https://github.com/golang/lint for installation (it doesn't come bundled with the compiler). It is very easy to use and is integrated some IDEs to be run when you save a source code file (Atom or Sublime Text, for example). Do you remember the implicit/explicit code that we run when talking about variables? Let's lint it:

//Explicitly declaring a "string" variable 
var explicit string = "Hello, I'm a explicitly declared variable" 
    
//Implicitly declaring a "string". 
Type inferred inferred := ", I'm an inferred variable " 

$ golint main.go

The main.go:10:21: command should omit the type string from the declaration of the explicitString variable; it will be inferred from the right-hand side.

It is telling us that Go compiler will actually infer this type of a variable from the code and you don't need to declare its type. What about the Train type on the interface section?

Type Train struct { 
    TrainWidth int 
} 

$ golint main.go

The main.go:5:6: type exported Train type should have a comment or remain not exported.

In this case, it's pointing us that a public type such as Train type must be commented so that users can read the generated documentation to know its behavior.

The gofmt tool comes bundled with the compiler that already has access to it. Its purpose is to provide a set of indentation, formatting, spacing and few other rules to achieve good-looking Go code. For example, let's take the code of Hello World and make it a bit weirder by inserting spaces everywhere:

package main 

func  main(){ 
    println("Hello World!") 
} 

$ gofmt main.go 
package main 
 
func main() { 
        println("Hello World!") 
} 

The gofmt command prints it correctly again. What is more, we can use the -w flag to overwrite the original file:

$ gofmt -w main.go

And now we'll have our file properly corrected.

Go documentation is pretty extended and verbose. You can find detailed information about any topic you want to achieve. The godoc tool also helps you access this documentation directly from the command line. For example, we can query the package encoding/json:

$godoc cmd/encoding/json
[...]
FUNCTIONS
func Compact(dst *bytes.Buffer, src []byte) error
Compact appends to dst the JSON-encoded src with insignificant space
characters elided.
func HTMLEscape(dst *bytes.Buffer, src []byte)
[...]

You can also use grep, a bash utility for Linux and Mac, to find specific information about some functionality. For example, we'll use grep to look for text that mentions anything about parsing JSON files:

$ godoc cmd/encoding/json | grep parse

The Unmarshal command parses the JSON encoded data and stores the result in the object being parsed.

One of the things that the golint command warns about is to use the beginning of a comment with the same name of the function it describes. This way, if you don't remember the name of the function that parses JSON, you can use godoc with grep and search for parse so the beginning of the line will always be the function name like in the example preceding the Unmarshal command.

The goimport tool is a must have in Go. Sometimes you remember your packages so well that you don't need to search much to remember their API but it's more difficult to remember the project they belong to when doing the import. The goimport command helps you by searching your $GOPATH for occurrences of a package that you could be using to provide you with the project import line automatically. This is very useful if you configure your IDE to run goimport on save so that all used packages in the source file are imported automatically if you used them. It also works the other way around--if you delete the function you were using from a package and the package isn't being used anymore, it will remove the import line.