当我们的代码敲下[]时，便会被go编译器解析为抽象语法树上的切片节点, 被初始化为切片表达式SliceType:
// go/src/cmd/compile/internal/syntax/parser.go
// TypeSpec = identifier [ TypeParams ] [ "=" ] Type .
func (p *parser) typeDecl(group *Group) Decl {
...
if p.tok == _Lbrack {
// d.Name "[" ...
// array/slice type or type parameter list
pos := p.pos()
p.next()
switch p.tok {
...
case _Rbrack:
// d.Name "[" "]" ...
p.next()
d.Type = p.sliceType(pos)
...
}
}
...
}
func (p *parser) sliceType(pos Pos) Expr {
t := new(SliceType)
t.pos = pos
t.Elem = p.type_()
return t
}
// go/src/cmd/compile/internal/syntax/nodes.go
type (
...
// []Elem
SliceType struct {
Elem Expr
expr
}
...
)编译时切片定义为Slice结构体，属性只包含同一类型的元素Elem，编译时通过NewSlice()函数进行创建：
// go/src/cmd/compile/internal/types/type.go
type Slice struct {
Elem *Type // element type
}
func NewSlice(elem *Type) *Type {
if t := elem.cache.slice; t != nil {
if t.Elem() != elem {
base.Fatalf("elem mismatch")
}
if elem.HasTParam() != t.HasTParam() || elem.HasShape() != t.HasShape() {
base.Fatalf("Incorrect HasTParam/HasShape flag for cached slice type")
}
return t
}
t := newType(TSLICE)
t.extra = Slice{Elem: elem}
elem.cache.slice = t
if elem.HasTParam() {
t.SetHasTParam(true)
}
if elem.HasShape() {
t.SetHasShape(true)
}
return t
}切片有两种初始化方式，一种声明即初始化称为字面量初始化，一种称为make初始化，
例如:
litSlic := []int{1,2,3,4} // 字面量初始化
makeSlic := make([]int,0) // make初始化切片字面量的初始化是在生成抽象语法树后进行遍历的walk阶段完成的。通过walkComplit方法，首先会进行类型检查，此时会计算出切片元素的个数length，然后通过slicelit方法完成具体的初始化工作。整个过程会先创建一个数组存储于静态区(static array)，并在堆区创建一个新的切片(auto array)，然后将静态区的数据复制到堆区(copy the static array to the auto array)，对于切片中的元素会按索引位置一个一个的进行赋值。 在程序启动时这一过程会加快切片的初始化。
// go/src/cmd/compile/internal/walk/complit.go
// walkCompLit walks a composite literal node:
// OARRAYLIT, OSLICELIT, OMAPLIT, OSTRUCTLIT (all CompLitExpr), or OPTRLIT (AddrExpr).
func walkCompLit(n ir.Node, init *ir.Nodes) ir.Node {
if isStaticCompositeLiteral(n) && !ssagen.TypeOK(n.Type()) {
n := n.(*ir.CompLitExpr) // not OPTRLIT
// n can be directly represented in the read-only data section.
// Make direct reference to the static data. See issue 12841.
vstat := readonlystaticname(n.Type())
fixedlit(inInitFunction, initKindStatic, n, vstat, init)
return typecheck.Expr(vstat)
}
var_ := typecheck.Temp(n.Type())
anylit(n, var_, init)
return var_
}类型检查时，计算出切片长度的过程为:
// go/src/cmd/compile/internal/typecheck/expr.go
func tcCompLit(n *ir.CompLitExpr) (res ir.Node) {
...
t := n.Type()
base.AssertfAt(t != nil, n.Pos(), "missing type in composite literal")
switch t.Kind() {
...
case types.TSLICE:
length := typecheckarraylit(t.Elem(), -1, n.List, "slice literal")
n.SetOp(ir.OSLICELIT)
n.Len = length
...
}
return n
}切片的具体初始化过程为:
源代码通过注释也写明了整个过程。
// go/src/cmd/compile/internal/walk/complit.go
func anylit(n ir.Node, var_ ir.Node, init *ir.Nodes) {
t := n.Type()
switch n.Op() {
...
case ir.OSLICELIT:
n := n.(*ir.CompLitExpr)
slicelit(inInitFunction, n, var_, init)
...
}
}
func slicelit(ctxt initContext, n *ir.CompLitExpr, var_ ir.Node, init *ir.Nodes) {
// make an array type corresponding the number of elements we have
t := types.NewArray(n.Type().Elem(), n.Len)
types.CalcSize(t)
if ctxt == inNonInitFunction {
// put everything into static array
vstat := staticinit.StaticName(t)
fixedlit(ctxt, initKindStatic, n, vstat, init)
fixedlit(ctxt, initKindDynamic, n, vstat, init)
// copy static to slice
var_ = typecheck.AssignExpr(var_)
name, offset, ok := staticinit.StaticLoc(var_)
if !ok || name.Class != ir.PEXTERN {
base.Fatalf("slicelit: %v", var_)
}
staticdata.InitSlice(name, offset, vstat.Linksym(), t.NumElem())
return
}
// recipe for var = []t{...}
// 1. make a static array
// var vstat [...]t
// 2. assign (data statements) the constant part
// vstat = constpart{}
// 3. make an auto pointer to array and allocate heap to it
// var vauto *[...]t = new([...]t)
// 4. copy the static array to the auto array
// *vauto = vstat
// 5. for each dynamic part assign to the array
// vauto[i] = dynamic part
// 6. assign slice of allocated heap to var
// var = vauto[:]
//
// an optimization is done if there is no constant part
// 3. var vauto *[...]t = new([...]t)
// 5. vauto[i] = dynamic part
// 6. var = vauto[:]
// if the literal contains constants,
// make static initialized array (1),(2)
var vstat ir.Node
mode := getdyn(n, true)
if mode&initConst != 0 && !isSmallSliceLit(n) {
if ctxt == inInitFunction {
vstat = readonlystaticname(t)
} else {
vstat = staticinit.StaticName(t)
}
fixedlit(ctxt, initKindStatic, n, vstat, init)
}
// make new auto *array (3 declare)
vauto := typecheck.Temp(types.NewPtr(t))
// set auto to point at new temp or heap (3 assign)
var a ir.Node
if x := n.Prealloc; x != nil {
// temp allocated during order.go for dddarg
if !types.Identical(t, x.Type()) {
panic("dotdotdot base type does not match order's assigned type")
}
a = initStackTemp(init, x, vstat)
} else if n.Esc() == ir.EscNone {
a = initStackTemp(init, typecheck.Temp(t), vstat)
} else {
a = ir.NewUnaryExpr(base.Pos, ir.ONEW, ir.TypeNode(t))
}
appendWalkStmt(init, ir.NewAssignStmt(base.Pos, vauto, a))
if vstat != nil && n.Prealloc == nil && n.Esc() != ir.EscNone {
// If we allocated on the heap with ONEW, copy the static to the
// heap (4). We skip this for stack temporaries, because
// initStackTemp already handled the copy.
a = ir.NewStarExpr(base.Pos, vauto)
appendWalkStmt(init, ir.NewAssignStmt(base.Pos, a, vstat))
}
// put dynamics into array (5)
var index int64
for _, value := range n.List {
if value.Op() == ir.OKEY {
kv := value.(*ir.KeyExpr)
index = typecheck.IndexConst(kv.Key)
if index < 0 {
base.Fatalf("slicelit: invalid index %v", kv.Key)
}
value = kv.Value
}
a := ir.NewIndexExpr(base.Pos, vauto, ir.NewInt(index))
a.SetBounded(true)
index++
// TODO need to check bounds?
switch value.Op() {
case ir.OSLICELIT:
break
case ir.OARRAYLIT, ir.OSTRUCTLIT:
value := value.(*ir.CompLitExpr)
k := initKindDynamic
if vstat == nil {
// Generate both static and dynamic initializations.
// See issue #31987.
k = initKindLocalCode
}
fixedlit(ctxt, k, value, a, init)
continue
}
if vstat != nil && ir.IsConstNode(value) { // already set by copy from static value
continue
}
// build list of vauto[c] = expr
ir.SetPos(value)
as := ir.NewAssignStmt(base.Pos, a, value)
appendWalkStmt(init, orderStmtInPlace(typecheck.Stmt(as), map[string][]*ir.Name{}))
}
// make slice out of heap (6)
a = ir.NewAssignStmt(base.Pos, var_, ir.NewSliceExpr(base.Pos, ir.OSLICE, vauto, nil, nil, nil))
appendWalkStmt(init, orderStmtInPlace(typecheck.Stmt(a), map[string][]*ir.Name{}))
}当使用make初始化一个切片时，会被编译器解析为一个OMAKESLICE操作:
// go/src/cmd/compile/internal/walk/expr.go
func walkExpr1(n ir.Node, init *ir.Nodes) ir.Node {
switch n.Op() {
...
case ir.OMAKESLICE:
n := n.(*ir.MakeExpr)
return walkMakeSlice(n, init)
...
}如果make初始化一个较大的切片则会逃逸到堆中，如果分配了一个较小的切片则直接在栈中分配。
walkMakeSlice函数中，如果未指定切片的容量Cap，则初始容量等于切片的长度。n.Esc() == ir.EscNone，则会先在内存中创建一个同样容量大小的数组NewArray(), 然后按切片长度将数组中的值arr[:l]赋予切片。makeslice和makeslice64在堆中完成对切片的初始化。// go/src/cmd/compile/internal/walk/builtin.go
func walkMakeSlice(n *ir.MakeExpr, init *ir.Nodes) ir.Node {
l := n.Len
r := n.Cap
if r == nil {
r = safeExpr(l, init)
l = r
}
...
if n.Esc() == ir.EscNone {
if why := escape.HeapAllocReason(n); why != "" {
base.Fatalf("%v has EscNone, but %v", n, why)
}
// var arr [r]T
// n = arr[:l]
i := typecheck.IndexConst(r)
if i < 0 {
base.Fatalf("walkExpr: invalid index %v", r)
}
...
t = types.NewArray(t.Elem(), i) // [r]T
var_ := typecheck.Temp(t)
appendWalkStmt(init, ir.NewAssignStmt(base.Pos, var_, nil)) // zero temp
r := ir.NewSliceExpr(base.Pos, ir.OSLICE, var_, nil, l, nil) // arr[:l]
// The conv is necessary in case n.Type is named.
return walkExpr(typecheck.Expr(typecheck.Conv(r, n.Type())), init)
}
// n escapes; set up a call to makeslice.
// When len and cap can fit into int, use makeslice instead of
// makeslice64, which is faster and shorter on 32 bit platforms.
len, cap := l, r
fnname := "makeslice64"
argtype := types.Types[types.TINT64]
// Type checking guarantees that TIDEAL len/cap are positive and fit in an int.
// The case of len or cap overflow when converting TUINT or TUINTPTR to TINT
// will be handled by the negative range checks in makeslice during runtime.
if (len.Type().IsKind(types.TIDEAL) || len.Type().Size() <= types.Types[types.TUINT].Size()) &&
(cap.Type().IsKind(types.TIDEAL) || cap.Type().Size() <= types.Types[types.TUINT].Size()) {
fnname = "makeslice"
argtype = types.Types[types.TINT]
}
fn := typecheck.LookupRuntime(fnname)
ptr := mkcall1(fn, types.Types[types.TUNSAFEPTR], init, reflectdata.TypePtr(t.Elem()), typecheck.Conv(len, argtype), typecheck.Conv(cap, argtype))
ptr.MarkNonNil()
len = typecheck.Conv(len, types.Types[types.TINT])
cap = typecheck.Conv(cap, types.Types[types.TINT])
sh := ir.NewSliceHeaderExpr(base.Pos, t, ptr, len, cap)
return walkExpr(typecheck.Expr(sh), init)
}切片在栈中初始化还是在堆中初始化，存在一个临界值进行判断。临界值maxImplicitStackVarSize默认为64kb。从下面的源代码可以看到，显式变量声明explicit variable declarations 和隐式变量implicit variables逃逸的临界值并不一样。
var变量声明以及:=赋值操作时，内存逃逸的临界值为10M, 小于该值的对象会分配在栈中。64kb，小于该值的对象会分配在栈中。p := new(T)
p := &T{}
s := make([]T, n)
s := []byte("...") // go/src/cmd/compile/internal/ir/cfg.go
var (
// maximum size variable which we will allocate on the stack.
// This limit is for explicit variable declarations like "var x T" or "x := ...".
// Note: the flag smallframes can update this value.
MaxStackVarSize = int64(10 * 1024 * 1024)
// maximum size of implicit variables that we will allocate on the stack.
// p := new(T) allocating T on the stack
// p := &T{} allocating T on the stack
// s := make([]T, n) allocating [n]T on the stack
// s := []byte("...") allocating [n]byte on the stack
// Note: the flag smallframes can update this value.
MaxImplicitStackVarSize = int64(64 * 1024)
// MaxSmallArraySize is the maximum size of an array which is considered small.
// Small arrays will be initialized directly with a sequence of constant stores.
// Large arrays will be initialized by copying from a static temp.
// 256 bytes was chosen to minimize generated code + statictmp size.
MaxSmallArraySize = int64(256)
)切片的make初始化就属于s := make([]T, n)操作，当切片元素分配的内存大小大于64kb时, 切片会逃逸到堆中进行初始化。此时会调用运行时函数makeslice来完成这一个过程:
// go/src/runtime/slice.go
func makeslice(et *_type, len, cap int) unsafe.Pointer {
mem, overflow := math.MulUintptr(et.size, uintptr(cap))
if overflow || mem > maxAlloc || len < 0 || len > cap {
// NOTE: Produce a 'len out of range' error instead of a
// 'cap out of range' error when someone does make([]T, bignumber).
// 'cap out of range' is true too, but since the cap is only being
// supplied implicitly, saying len is clearer.
// See golang.org/issue/4085.
mem, overflow := math.MulUintptr(et.size, uintptr(len))
if overflow || mem > maxAlloc || len < 0 {
panicmakeslicelen()
}
panicmakeslicecap()
}
return mallocgc(mem, et, true)
}根据切片的运行时结构定义，运行时切片结构底层维护着切片的长度len、容量cap以及指向数组数据的指针array:
// go/src/runtime/slice.go
type slice struct {
array unsafe.Pointer
len int
cap int
}
// 或者
// go/src/reflect/value.go
// SliceHeader is the runtime representation of a slice.
type SliceHeader struct {
Data uintptr
Len int
Cap int
}从切片的运行时结构已经知道，切片底层数据是一个数组，切片本身只是持有一个指向改数组数据的指针。因此，当我们对切片进行截取操作时，新的切片仍然指向原切片的底层数据，当对原切片数据进行更新时，意味着新切片相同索引位置的数据也发生了变化:
slic := []int{1, 2, 3, 4, 5}
slic1 := slic[:2]
fmt.Printf("slic1: %v\n", slic1)
slic[0] = 0
fmt.Printf("slic: %v\n", slic)
fmt.Printf("slic1: %v\n", slic1)
// slic1: [1 2]
// slic: [0 2 3 4 5]
// slic1: [0 2]切片截取后，虽然底层数据没有发生变化，但指向的数据范围发生了变化，表现为截取后的切片长度、容量会相应发生变化:
slic := []int{1, 2, 3, 4, 5}
slic1 := slic[:2]
slic2 := slic[2:]
fmt.Printf("len(slic): %v\n", len(slic))
fmt.Printf("cap(slic): %v\n", cap(slic))
fmt.Printf("len(slic1): %v\n", len(slic1))
fmt.Printf("cap(slic1): %v\n", cap(slic1))
fmt.Printf("len(slic2): %v\n", len(slic2))
fmt.Printf("cap(slic2): %v\n", cap(slic2))
// len(slic): 5
// cap(slic): 5
// len(slic1): 2
// cap(slic1): 5
// len(slic2): 3
// cap(slic2): 3
所以，切片截取变化的是底层data指针、长度以及容量，data指针指向的数组数据本身没有变化。切片的赋值拷贝就等价于于全切片，底层data指针仍然指向相同的数组地址，长度和容量保持不变:
slic := []int{1, 2, 3, 4, 5}
s := slic // 等价于 s := slic[:]当切片作为参数传递时，即使切片中包含大量的数据，也只是切片数据地址的拷贝，拷贝的成本是较低的。
当我们想要完整拷贝一个切片时，可以使用内置的copy函数，效果类似于"深拷贝"。
slic := []int{1, 2, 3, 4, 5}
var slic1 []int
copy(slic1, slic)
fmt.Printf("slic: %p\n", slic)
fmt.Printf("slic1: %p\n", slic1)
// slic: 0xc0000aa030
// slic1: 0x0
完整复制后，新的切片指向了新的内存地址。切片的复制在运行时会调用slicecopy()函数，通过memmove移动数据到新的内存地址:
// go/src/runtime/slice.go
func slicecopy(toPtr unsafe.Pointer, toLen int, fromPtr unsafe.Pointer, fromLen int, width uintptr) int {
if fromLen == 0 || toLen == 0 {
return 0
}
n := fromLen
if toLen < n {
n = toLen
}
...
if size == 1 { // common case worth about 2x to do here
// TODO: is this still worth it with new memmove impl?
*(*byte)(toPtr) = *(*byte)(fromPtr) // known to be a byte pointer
} else {
memmove(toPtr, fromPtr, size)
}
return n
}切片元素个数可以动态变化，切片初始化后会确定一个初始化容量，当容量不足时会在运行时通过growslice进行扩容:
func growslice(et *_type, old slice, cap int) slice {
...
newcap := old.cap
doublecap := newcap + newcap
if cap > doublecap {
newcap = cap
} else {
const threshold = 256
if old.cap < threshold {
newcap = doublecap
} else {
// Check 0 < newcap to detect overflow
// and prevent an infinite loop.
for 0 < newcap && newcap < cap {
// Transition from growing 2x for small slices
// to growing 1.25x for large slices. This formula
// gives a smooth-ish transition between the two.
newcap += (newcap + 3*threshold) / 4
}
// Set newcap to the requested cap when
// the newcap calculation overflowed.
if newcap <= 0 {
newcap = cap
}
}
}
...
memmove(p, old.array, lenmem)
return slice{p, old.len, newcap}
}从growslice的代码可以看出：
cap)大于二倍旧容量(old.cap)时，最终容量(newcap)是新申请的容量；cap)小于二倍旧容量(old.cap)时，newcap += (newcap + 3*threshold) / 4来确定最终容量。实际的表现为:memmove()函数将旧的数组移动到新的地址，因此扩容后新的切片一般和原来的地址不同。示例:
var slic []int
oldCap := cap(slic)
for i := 0; i < 2048; i++ {
slic = append(slic, i)
newCap := cap(slic)
grow := float32(newCap) / float32(oldCap)
if newCap != oldCap {
fmt.Printf("len(slic):%v cap(slic):%v grow:%v %p\n", len(slic), cap(slic), grow, slic)
}
oldCap = newCap
}
// len(slic):1 cap(slic):1 grow:+Inf 0xc0000140c0
// len(slic):2 cap(slic):2 grow:2 0xc0000140e0
// len(slic):3 cap(slic):4 grow:2 0xc000020100
// len(slic):5 cap(slic):8 grow:2 0xc00001e340
// len(slic):9 cap(slic):16 grow:2 0xc000026080
// len(slic):17 cap(slic):32 grow:2 0xc00007e000
// len(slic):33 cap(slic):64 grow:2 0xc000100000
// len(slic):65 cap(slic):128 grow:2 0xc000102000
// len(slic):129 cap(slic):256 grow:2 0xc000104000
// len(slic):257 cap(slic):512 grow:2 0xc000106000
// len(slic):513 cap(slic):1024 grow:2 0xc000108000
// len(slic):1025 cap(slic):1280 grow:1.25 0xc00010a000
// len(slic):1281 cap(slic):1696 grow:1.325 0xc000114000
// len(slic):1697 cap(slic):2304 grow:1.3584906 0xc00011e000切片在编译时定义为Slice结构体，并通过NewSlice()函数进行创建；
type Slice struct {
Elem *Type // element type
}切片的运行时定义为slice结构体, 底层维护着指向数组数据的指针，切片长度以及容量；
type slice struct {
array unsafe.Pointer
len int
cap int
}makeslice函数进行内存分配，当内存占用大于64kb时会逃逸到堆中；copy复制切片时，会在运行时会调用slicecopy()函数，通过memmove移动数据到了新的内存地址；growslice函数完成的，一般表现为：memmove()函数将旧的数组移动到新的地址，因此扩容后地址会发生变化。
