關於slice分配內存

slice是golang提供的一個很好的符合類型。既支持數據動態擴展，又能隨機訪問，使數據保持很好的局部性。但是slice有一個性能可能導致數據的一致性和預期不一致，就是它會按需爲slice收集內存。參考這段代碼：

=========================================

package main


import "fmt"


func main() {
var data = []string{"one", "", "three"}     //what's the differenct to assign underlying or not
var data2 = [...]string{"one", "", "three"} //what's the differenct to assign underlying or not
intTest := 3
fmt.Printf("intTest %%T after append is %T \n", intTest)
fmt.Printf("intTest %%p after append is %p \n", intTest)
fmt.Printf("data %%v is %v \n", data)
fmt.Printf("data %%q is %q \n", data)
out := data[:0]
for _, s := range data {
if s != "" {
out = append(out, s)
}
}
fmt.Printf("data %%v after append is %v \n", data)
fmt.Printf("data %%q after append is %q \n", data)
fmt.Printf("data %%s after append is %s \n", data)
fmt.Printf("data %%T after append is %T \n", data)
fmt.Printf("data %%p after append is %p \n", data)
fmt.Printf("&data %%p after append is %p \n", &data)
fmt.Printf("data2 %%T after append is %T \n", data2)
fmt.Printf("data2 %%p after append is %p \n", data2)
fmt.Printf("&data2 %%p after append is %p \n", &data2)
fmt.Printf("&data2 %%d after append is %d \n", &data2)


fmt.Printf("out %%v after append is %v \n", out)
fmt.Printf("out %%q after append is %q \n", out)


out2 := append(data[:0], "a", "b", "c", "d")
fmt.Printf("data %%v after append is %v \n", data)
fmt.Printf("data %%q after append is %q \n", data)
fmt.Printf("out2 %%v after append is %v \n", out2)
fmt.Printf("out2 %%q after append is %q \n", out2)


out3 := append(data[:0], "a", "b", "c")
fmt.Printf("data %%v after append is %v \n", data)
fmt.Printf("data %%q after append is %q \n", data)
fmt.Printf("out3 %%v after append is %v \n", out3)
fmt.Printf("out3 %%q after append is %q \n", out3)
}

=========================================

output：

=========================================

1. intTest %T after append is int
2. intTest %p after append is %!p(int=3)
3. data %v is [one  three]
4. data %q is ["one" "" "three"]
5. data %v after append is [one three three]
6. data %q after append is ["one" "three" "three"]
7. data %s after append is [one three three]
8. data %T after append is []string
9. data %p after append is 0xc420062150
10. &data %p after append is 0xc42007a020
11. data2 %T after append is [3]string
12. data2 %p after append is %!p([3]string=[one  three])
13. &data2 %p after append is 0xc420062180
14. &data2 %d after append is &[%!d(string=one) %!d(string=) %!d(string=three)]
15. out %v after append is [one three]
16. out %q after append is ["one" "three"]
17. data %v after append is [one three three]
18. data %q after append is ["one" "three" "three"]
19. out2 %v after append is [a b c d]
20. out2 %q after append is ["a" "b" "c" "d"]
21. data %v after append is [a b c]
22. data %q after append is ["a" "b" "c"]
23. out3 %v after append is [a b c]
24. out3 %q after append is ["a" "b" "c"]

=========================================

從輸出裏我們可以得出兩條結論：

1. 對比第17行（append數據長度超出capacity）和第21行的輸出（append數據長度未超出capacity），slice在容量足夠的時候修改已有的array，在容量不夠的時候才realloc內存基於新的內存建立array

2. 對比第12行和9行，golang是強類型的，%p只對引用內型或者取地址有效

3. 對比第9行和第10行，golang對slice和&slice的值定義了不同的語義