GoSudaGo Interfaces and Struct Embedding: Differences from Class Inheritance and Composition-Based Design
Overview
Go interfaces make it easy for several structs to have functions with the same arguments and return values, but this differs from the way Java's extends keyword extends and overrides the behavior of the inherited functions. Unless you properly understand Go's compositional code reuse, you will easily confuse it with inheritance, but it is difficult to achieve a perfect theoretical understanding from the start. Let's look at it together through scenarios where mistakes are likely to happen.
Common Mistakes
Beginners may make the following mistakes.
1package main
2import (
3 "fmt"
4 "strings"
5)
6
7type Fruits interface {
8 GetBrix() float64
9 GetName() string
10 SetLabel()
11 GetLabel(string) string
12 PrintAll()
13}
14
15type Apple struct {
16 Label string
17 Name string
18 Brix float64
19}
20
21type Watermelon struct {
22 Label string
23 Name string
24 Brix float64
25}
26
27func (a *Apple) PrintAll() {
28 fmt.Printf("Fruit: %s, Label: %s, Brix: %v\n", a.Name, a.Label, a.Brix)
29}
30
31const (
32 NO_LABEL = "EMPTY LABEL"
33)
34
35func (a *Apple) SetLabel(lbl string) {
36 a.Brix = 14.5;
37 a.Name = "apple";
38 lbl_lower := strings.ToLower(lbl)
39 if strings.Contains(lbl_lower, a.Name) {
40 fmt.Println("Succeed: Label was ", lbl)
41 a.Label = lbl;
42 } else {
43 fmt.Println("Failed: Label was ", lbl)
44 a.Label = NO_LABEL;
45 }
46}
47
48func (w *Watermelon) SetLabel(lbl string) {
49 w.Brix = 10;
50 w.Name = "watermelon";
51 lbl_lower := strings.ToLower(lbl)
52 if strings.Contains(lbl_lower, w.Name) {
53 w.Label = lbl;
54 } else {
55 w.Label = NO_LABEL;
56 }
57}
58
59func main() {
60 fmt.Println("Inheritance test #1")
61 apple := new(Apple)
62 watermelon := apple
63 apple.SetLabel("Apple_1")
64 fmt.Println("Apple, before copied to Watermelon")
65 apple.PrintAll()
66 watermelon.SetLabel("WaterMelon_2")
67 fmt.Println("Apple, after copied to Watermelon")
68 apple.PrintAll()
69 fmt.Println("Watermelon, which inherited Apple's Method")
70 watermelon.PrintAll()
71}
If you assume that Go follows traditional inheritance, this code may seem to have no problems. However, its output is as follows.
1Inheritance test #1
2Succeed: Label was Apple_1
3Apple, before copied to Watermelon
4Fruit: apple, Label: Apple_1, Brix: 14.5
5Failed: Label was WaterMelon_2
6Apple, after copied to Watermelon
7Fruit: apple, Label: EMPTY LABEL, Brix: 14.5
8Watermelon, which inherited Apple's Method
9Fruit: apple, Label: EMPTY LABEL, Brix: 14.5
Here, the behavior of Go becomes clear, only.
1watermelon := apple
This code does not convert Apple into the Watermelon class at all. Rather, watermelon is merely a pointer to apple.
Let me emphasize again: Go does not follow the traditional concept of inheritance.
If you write code with this misunderstanding, you will encounter critical errors such as meaningless pointer creation and unexpected copying of functions for other structs.
So, what does exemplary code look like?
Appropriate Examples in Go
1package main
2import (
3 "fmt"
4 "strings"
5)
6
7type Fruits interface {
8 GetBrix() float64
9 GetName() string
10 SetLabel()
11 GetLabel(string) string
12 PrintAll()
13}
14
15type BaseFruit struct {
16 Name string
17 Brix float64
18}
19
20type Apple struct {
21 Label string
22 Fruit BaseFruit
23}
24
25type Watermelon struct {
26 Label string
27 Fruit BaseFruit
28
29}
30
31func (b *BaseFruit) PrintAll() {
32 fmt.Printf("Fruit: %s, Brix: %v\n", b.Name, b.Brix)
33}
34
35
36const (
37 NO_LABEL = "EMPTY LABEL"
38)
39
40func (a *Apple) SetLabel(lbl string) {
41 a.Fruit.Brix = 14.5;
42 a.Fruit.Name = "apple";
43 lbl_lower := strings.ToLower(lbl)
44 if strings.Contains(lbl_lower, a.Fruit.Name) {
45 fmt.Println("Succeed: Label was ", lbl)
46 a.Label = lbl;
47 } else {
48 fmt.Println("Failed: Label was ", lbl)
49 a.Label = NO_LABEL;
50 }
51 fmt.Printf("Fruit %s label set to %s\n", a.Fruit.Name, a.Label);
52 a.Fruit.PrintAll()
53}
54
55func (w *Watermelon) SetLabel(lbl string) {
56 w.Fruit.Brix = 10;
57 w.Fruit.Name = "Watermelon";
58 lbl_lower := strings.ToLower(lbl)
59 if strings.Contains(lbl_lower, w.Fruit.Name) {
60 w.Label = lbl;
61 } else {
62 w.Label = NO_LABEL;
63 }
64 fmt.Printf("Fruit %s label set to %s\n", w.Fruit.Name, w.Label);
65 w.Fruit.PrintAll()
66}
67
68func main() {
69 apple := new(Apple)
70 watermelon := new(Watermelon)
71 apple.SetLabel("Apple_1")
72 watermelon.SetLabel("WaterMelon_2")
73}
However, it is still possible in Go to make things look like inheritance. This is the example of anonymous embedding. It is possible by declaring the inner struct as an unnamed struct. In this case, the fields of the lower struct can be accessed directly even without explicitly specifying them. Using this pattern of promoting the fields of the lower struct to the upper struct can, depending on the case, improve readability. However, when the lower struct needs to be shown explicitly, it is recommended not to use it.
1package main
2import (
3 "fmt"
4 "strings"
5)
6
7type Fruits interface {
8 GetBrix() float64
9 GetName() string
10 SetLabel()
11 GetLabel(string) string
12 PrintAll()
13}
14
15type BaseFruit struct {
16 Name string
17 Brix float64
18}
19
20type Apple struct {
21 Label string
22 BaseFruit
23}
24
25type Watermelon struct {
26 Label string
27 BaseFruit
28
29}
30
31func (b *BaseFruit) PrintAll() {
32 fmt.Printf("Fruit: %s, Brix: %v\n", b.Name, b.Brix)
33}
34
35
36const (
37 NO_LABEL = "EMPTY LABEL"
38)
39
40func (a *Apple) SetLabel(lbl string) {
41 a.Brix = 14.5;
42 a.Name = "apple";
43 lbl_lower := strings.ToLower(lbl)
44 if strings.Contains(lbl_lower, a.Name) {
45 fmt.Println("Succeed: Label was ", lbl)
46 a.Label = lbl;
47 } else {
48 fmt.Println("Failed: Label was ", lbl)
49 a.Label = NO_LABEL;
50 }
51 fmt.Printf("Fruit %s label set to %s\n", a.Name, a.Label);
52 a.PrintAll()
53}
54
55func (w *Watermelon) SetLabel(lbl string) {
56 w.Brix = 10;
57 w.Name = "Watermelon";
58 lbl_lower := strings.ToLower(lbl)
59 if strings.Contains(lbl_lower, w.Name) {
60 w.Label = lbl;
61 } else {
62 w.Label = NO_LABEL;
63 }
64 fmt.Printf("Fruit %s label set to %s\n", w.Name, w.Label);
65 w.PrintAll()
66}
67
68func main() {
69 apple := new(Apple)
70 watermelon := new(Watermelon)
71 apple.SetLabel("Apple_1")
72 watermelon.SetLabel("WaterMelon_2")
73}
This example has the following difference.
1w.PrintAll() // Automatic promotion call through the unnamed struct, not w.Friut.PrintAll()
The important points of both examples are as follows.
- Keep main simple, and organize functions by functionality
- Use different objects for different structs
- Use an inner struct when sharing is necessary
What are the advantages of this programming philosophy?
Advantages
- Clear distinction between methods that need to be shared and those that do not
- Separation of responsibility across individual structs and methods
- Code structurally separated according to the specification of required functionality
At first, Go may be unfamiliar because it differs from traditional OOP, but once you become accustomed to it, explicit programming becomes possible.
Summary
- Isolate areas of responsibility
- Divide finely at the level of structs
- Methods must not be understood like abstract classes in Java
- Program explicitly and concretely Go should be treated as simpler and clearer than the traditional OOP model, and handled individually. Rather than programming in an extensible manner, write code by separating it step by step and structurally.