3
0
mirror of https://github.com/ergochat/ergo.git synced 2024-11-15 16:39:32 +01:00
ergo/vendor/github.com/tidwall/btree/internal/btree.go

1276 lines
28 KiB
Go
Raw Normal View History

2022-06-16 19:36:02 +02:00
// Copyright 2020 Joshua J Baker. All rights reserved.
// Use of this source code is governed by an MIT-style license that can be
// found in the LICENSE file at https://github.com/tidwall/btree/LICENSE
///////////////////////////////////////////////////////////////////////////////
// BEGIN PARAMS
///////////////////////////////////////////////////////////////////////////////
package btree
import "sync"
// degree is the B-Tree degree, which is equal to maximum number of children
// pre node times two.
// The default is 128, which means each node can have 255 items and 256 child
// nodes.
const degree = 128
// kind is the item type.
// It's important to use the equal symbol, which tells Go to create an alias of
// the type, rather than creating an entirely new type.
type kind = interface{}
// contextKind is the kind of context that can be passed to NewOptions and the
// less function
type contextKind = interface{}
// less returns true if A is less than B.
// The value of context will be whatever was passed to NewOptions through the
// Options.Context field, otherwise nil if the field was not set.
func less(a, b kind, context contextKind) bool {
return context.(func(a, b contextKind) bool)(a, b)
}
// BTree aliases
// These are aliases to the local bTree types and functions, which are exported
// to allow for public use at a package level.
// Rename them if desired, or comment them out to make the library private.
type BTree = bTree
type Options = bOptions
type PathHint = bPathHint
type Iter = bIter
func New(less func(a, b kind) bool) *bTree { return bNew() }
func NewOptions(opts bOptions) *bTree { return bNewOptions(opts) }
// The functions below, which begin with "test*", are required by the
// btree_test.go file. If you choose not use include the btree_test.go file in
// your project then these functions may be omitted.
// testCustomSeed can be used to generate a custom random seed for testing.
// Returning false will use time.Now().UnixNano()
func testCustomSeed() (seed int64, ok bool) {
return 0, false
}
// testMakeItem must return a valid item for testing.
// It's required that the returned item maintains equal order as the
// provided int, such that:
// testMakeItem(0) < testMakeItem(1) < testMakeItem(2) < testMakeItem(10)
func testMakeItem(x int) (item kind) {
return x
}
// testNewBTree must return an operational btree for testing.
func testNewBTree() *bTree {
return bNewOptions(bOptions{
Context: func(a, b contextKind) bool {
if a == nil {
return b != nil
} else if b == nil {
return false
}
return a.(int) < b.(int)
},
})
}
///////////////////////////////////////////////////////////////////////////////
// END PARAMS
///////////////////////////////////////////////////////////////////////////////
// Do not edit code below this line.
const maxItems = degree*2 - 1 // max items per node. max children is +1
const minItems = maxItems / 2
type bTree struct {
mu *sync.RWMutex
cow *cow
root *node
count int
ctx contextKind
locks bool
empty kind
}
type node struct {
cow *cow
count int
items []kind
children *[]*node
}
type cow struct {
_ int // cannot be an empty struct
}
func (tr *bTree) newNode(leaf bool) *node {
n := &node{cow: tr.cow}
if !leaf {
n.children = new([]*node)
}
return n
}
// leaf returns true if the node is a leaf.
func (n *node) leaf() bool {
return n.children == nil
}
// PathHint is a utility type used with the *Hint() functions. Hints provide
// faster operations for clustered keys.
type bPathHint struct {
used [8]bool
path [8]uint8
}
type bOptions struct {
NoLocks bool
Context contextKind
}
// New returns a new BTree
func bNew() *bTree {
return bNewOptions(bOptions{})
}
func bNewOptions(opts bOptions) *bTree {
tr := new(bTree)
tr.cow = new(cow)
tr.mu = new(sync.RWMutex)
tr.ctx = opts.Context
tr.locks = !opts.NoLocks
return tr
}
// Less is a convenience function that performs a comparison of two items
// using the same "less" function provided to New.
func (tr *bTree) Less(a, b kind) bool {
return less(a, b, tr.ctx)
}
func (tr *bTree) find(n *node, key kind,
hint *bPathHint, depth int,
) (index int, found bool) {
if hint == nil {
// fast path for no hinting
low := 0
high := len(n.items)
for low < high {
mid := (low + high) / 2
if !tr.Less(key, n.items[mid]) {
low = mid + 1
} else {
high = mid
}
}
if low > 0 && !tr.Less(n.items[low-1], key) {
return low - 1, true
}
return low, false
}
// Try using hint.
// Best case finds the exact match, updates the hint and returns.
// Worst case, updates the low and high bounds to binary search between.
low := 0
high := len(n.items) - 1
if depth < 8 && hint.used[depth] {
index = int(hint.path[depth])
if index >= len(n.items) {
// tail item
if tr.Less(n.items[len(n.items)-1], key) {
index = len(n.items)
goto path_match
}
index = len(n.items) - 1
}
if tr.Less(key, n.items[index]) {
if index == 0 || tr.Less(n.items[index-1], key) {
goto path_match
}
high = index - 1
} else if tr.Less(n.items[index], key) {
low = index + 1
} else {
found = true
goto path_match
}
}
// Do a binary search between low and high
// keep on going until low > high, where the guarantee on low is that
// key >= items[low - 1]
for low <= high {
mid := low + ((high+1)-low)/2
// if key >= n.items[mid], low = mid + 1
// which implies that key >= everything below low
if !tr.Less(key, n.items[mid]) {
low = mid + 1
} else {
high = mid - 1
}
}
// if low > 0, n.items[low - 1] >= key,
// we have from before that key >= n.items[low - 1]
// therefore key = n.items[low - 1],
// and we have found the entry for key.
// Otherwise we must keep searching for the key in index `low`.
if low > 0 && !tr.Less(n.items[low-1], key) {
index = low - 1
found = true
} else {
index = low
found = false
}
path_match:
if depth < 8 {
hint.used[depth] = true
var pathIndex uint8
if n.leaf() && found {
pathIndex = uint8(index + 1)
} else {
pathIndex = uint8(index)
}
if pathIndex != hint.path[depth] {
hint.path[depth] = pathIndex
for i := depth + 1; i < 8; i++ {
hint.used[i] = false
}
}
}
return index, found
}
// SetHint sets or replace a value for a key using a path hint
func (tr *bTree) SetHint(item kind, hint *bPathHint) (prev kind, replaced bool) {
if tr.lock() {
defer tr.unlock()
}
return tr.setHint(item, hint)
}
func (tr *bTree) setHint(item kind, hint *bPathHint) (prev kind, replaced bool) {
if tr.root == nil {
tr.root = tr.newNode(true)
tr.root.items = append([]kind{}, item)
tr.root.count = 1
tr.count = 1
return tr.empty, false
}
prev, replaced, split := tr.nodeSet(&tr.root, item, hint, 0)
if split {
left := tr.cowLoad(&tr.root)
right, median := tr.nodeSplit(left)
tr.root = tr.newNode(false)
*tr.root.children = make([]*node, 0, maxItems+1)
*tr.root.children = append([]*node{}, left, right)
tr.root.items = append([]kind{}, median)
tr.root.updateCount()
return tr.setHint(item, hint)
}
if replaced {
return prev, true
}
tr.count++
return tr.empty, false
}
// Set or replace a value for a key
func (tr *bTree) Set(item kind) (kind, bool) {
return tr.SetHint(item, nil)
}
func (tr *bTree) nodeSplit(n *node) (right *node, median kind) {
i := maxItems / 2
median = n.items[i]
// left node
left := tr.newNode(n.leaf())
left.items = make([]kind, len(n.items[:i]), maxItems/2)
copy(left.items, n.items[:i])
if !n.leaf() {
*left.children = make([]*node, len((*n.children)[:i+1]), maxItems+1)
copy(*left.children, (*n.children)[:i+1])
}
left.updateCount()
// right node
right = tr.newNode(n.leaf())
right.items = make([]kind, len(n.items[i+1:]), maxItems/2)
copy(right.items, n.items[i+1:])
if !n.leaf() {
*right.children = make([]*node, len((*n.children)[i+1:]), maxItems+1)
copy(*right.children, (*n.children)[i+1:])
}
right.updateCount()
*n = *left
return right, median
}
func (n *node) updateCount() {
n.count = len(n.items)
if !n.leaf() {
for i := 0; i < len(*n.children); i++ {
n.count += (*n.children)[i].count
}
}
}
// This operation should not be inlined because it's expensive and rarely
// called outside of heavy copy-on-write situations. Marking it "noinline"
// allows for the parent cowLoad to be inlined.
// go:noinline
func (tr *bTree) copy(n *node) *node {
n2 := new(node)
n2.cow = tr.cow
n2.count = n.count
n2.items = make([]kind, len(n.items), cap(n.items))
copy(n2.items, n.items)
if !n.leaf() {
n2.children = new([]*node)
*n2.children = make([]*node, len(*n.children), maxItems+1)
copy(*n2.children, *n.children)
}
return n2
}
// cowLoad loads the provided node and, if needed, performs a copy-on-write.
func (tr *bTree) cowLoad(cn **node) *node {
if (*cn).cow != tr.cow {
*cn = tr.copy(*cn)
}
return *cn
}
func (tr *bTree) nodeSet(cn **node, item kind,
hint *bPathHint, depth int,
) (prev kind, replaced bool, split bool) {
n := tr.cowLoad(cn)
i, found := tr.find(n, item, hint, depth)
if found {
prev = n.items[i]
n.items[i] = item
return prev, true, false
}
if n.leaf() {
if len(n.items) == maxItems {
return tr.empty, false, true
}
n.items = append(n.items, tr.empty)
copy(n.items[i+1:], n.items[i:])
n.items[i] = item
n.count++
return tr.empty, false, false
}
prev, replaced, split = tr.nodeSet(&(*n.children)[i], item, hint, depth+1)
if split {
if len(n.items) == maxItems {
return tr.empty, false, true
}
right, median := tr.nodeSplit((*n.children)[i])
*n.children = append(*n.children, nil)
copy((*n.children)[i+1:], (*n.children)[i:])
(*n.children)[i+1] = right
n.items = append(n.items, tr.empty)
copy(n.items[i+1:], n.items[i:])
n.items[i] = median
return tr.nodeSet(&n, item, hint, depth)
}
if !replaced {
n.count++
}
return prev, replaced, false
}
func (tr *bTree) Scan(iter func(item kind) bool) {
if tr.rlock() {
defer tr.runlock()
}
if tr.root == nil {
return
}
tr.root.scan(iter)
}
func (n *node) scan(iter func(item kind) bool) bool {
if n.leaf() {
for i := 0; i < len(n.items); i++ {
if !iter(n.items[i]) {
return false
}
}
return true
}
for i := 0; i < len(n.items); i++ {
if !(*n.children)[i].scan(iter) {
return false
}
if !iter(n.items[i]) {
return false
}
}
return (*n.children)[len(*n.children)-1].scan(iter)
}
// Get a value for key
func (tr *bTree) Get(key kind) (kind, bool) {
return tr.GetHint(key, nil)
}
// GetHint gets a value for key using a path hint
func (tr *bTree) GetHint(key kind, hint *bPathHint) (kind, bool) {
if tr.rlock() {
defer tr.runlock()
}
if tr.root == nil {
return tr.empty, false
}
n := tr.root
depth := 0
for {
i, found := tr.find(n, key, hint, depth)
if found {
return n.items[i], true
}
if n.children == nil {
return tr.empty, false
}
n = (*n.children)[i]
depth++
}
}
// Len returns the number of items in the tree
func (tr *bTree) Len() int {
return tr.count
}
// Delete a value for a key
func (tr *bTree) Delete(key kind) (kind, bool) {
return tr.DeleteHint(key, nil)
}
// DeleteHint deletes a value for a key using a path hint
func (tr *bTree) DeleteHint(key kind, hint *bPathHint) (kind, bool) {
if tr.lock() {
defer tr.unlock()
}
return tr.deleteHint(key, hint)
}
func (tr *bTree) deleteHint(key kind, hint *bPathHint) (kind, bool) {
if tr.root == nil {
return tr.empty, false
}
prev, deleted := tr.delete(&tr.root, false, key, hint, 0)
if !deleted {
return tr.empty, false
}
if len(tr.root.items) == 0 && !tr.root.leaf() {
tr.root = (*tr.root.children)[0]
}
tr.count--
if tr.count == 0 {
tr.root = nil
}
return prev, true
}
func (tr *bTree) delete(cn **node, max bool, key kind,
hint *bPathHint, depth int,
) (kind, bool) {
n := tr.cowLoad(cn)
var i int
var found bool
if max {
i, found = len(n.items)-1, true
} else {
i, found = tr.find(n, key, hint, depth)
}
if n.leaf() {
if found {
// found the items at the leaf, remove it and return.
prev := n.items[i]
copy(n.items[i:], n.items[i+1:])
n.items[len(n.items)-1] = tr.empty
n.items = n.items[:len(n.items)-1]
n.count--
return prev, true
}
return tr.empty, false
}
var prev kind
var deleted bool
if found {
if max {
i++
prev, deleted = tr.delete(&(*n.children)[i], true, tr.empty, nil, 0)
} else {
prev = n.items[i]
maxItem, _ := tr.delete(&(*n.children)[i], true, tr.empty, nil, 0)
deleted = true
n.items[i] = maxItem
}
} else {
prev, deleted = tr.delete(&(*n.children)[i], max, key, hint, depth+1)
}
if !deleted {
return tr.empty, false
}
n.count--
if len((*n.children)[i].items) < minItems {
tr.nodeRebalance(n, i)
}
return prev, true
}
// nodeRebalance rebalances the child nodes following a delete operation.
// Provide the index of the child node with the number of items that fell
// below minItems.
func (tr *bTree) nodeRebalance(n *node, i int) {
if i == len(n.items) {
i--
}
// ensure copy-on-write
left := tr.cowLoad(&(*n.children)[i])
right := tr.cowLoad(&(*n.children)[i+1])
if len(left.items)+len(right.items) < maxItems {
// Merges the left and right children nodes together as a single node
// that includes (left,item,right), and places the contents into the
// existing left node. Delete the right node altogether and move the
// following items and child nodes to the left by one slot.
// merge (left,item,right)
left.items = append(left.items, n.items[i])
left.items = append(left.items, right.items...)
if !left.leaf() {
*left.children = append(*left.children, *right.children...)
}
left.count += right.count + 1
// move the items over one slot
copy(n.items[i:], n.items[i+1:])
n.items[len(n.items)-1] = tr.empty
n.items = n.items[:len(n.items)-1]
// move the children over one slot
copy((*n.children)[i+1:], (*n.children)[i+2:])
(*n.children)[len(*n.children)-1] = nil
(*n.children) = (*n.children)[:len(*n.children)-1]
} else if len(left.items) > len(right.items) {
// move left -> right over one slot
// Move the item of the parent node at index into the right-node first
// slot, and move the left-node last item into the previously moved
// parent item slot.
right.items = append(right.items, tr.empty)
copy(right.items[1:], right.items)
right.items[0] = n.items[i]
right.count++
n.items[i] = left.items[len(left.items)-1]
left.items[len(left.items)-1] = tr.empty
left.items = left.items[:len(left.items)-1]
left.count--
if !left.leaf() {
// move the left-node last child into the right-node first slot
*right.children = append(*right.children, nil)
copy((*right.children)[1:], *right.children)
(*right.children)[0] = (*left.children)[len(*left.children)-1]
(*left.children)[len(*left.children)-1] = nil
(*left.children) = (*left.children)[:len(*left.children)-1]
left.count -= (*right.children)[0].count
right.count += (*right.children)[0].count
}
} else {
// move left <- right over one slot
// Same as above but the other direction
left.items = append(left.items, n.items[i])
left.count++
n.items[i] = right.items[0]
copy(right.items, right.items[1:])
right.items[len(right.items)-1] = tr.empty
right.items = right.items[:len(right.items)-1]
right.count--
if !left.leaf() {
*left.children = append(*left.children, (*right.children)[0])
copy(*right.children, (*right.children)[1:])
(*right.children)[len(*right.children)-1] = nil
*right.children = (*right.children)[:len(*right.children)-1]
left.count += (*left.children)[len(*left.children)-1].count
right.count -= (*left.children)[len(*left.children)-1].count
}
}
}
// Ascend the tree within the range [pivot, last]
// Pass nil for pivot to scan all item in ascending order
// Return false to stop iterating
func (tr *bTree) Ascend(pivot kind, iter func(item kind) bool) {
if tr.rlock() {
defer tr.runlock()
}
if tr.root == nil {
return
}
tr.ascend(tr.root, pivot, nil, 0, iter)
}
// The return value of this function determines whether we should keep iterating
// upon this functions return.
func (tr *bTree) ascend(n *node, pivot kind,
hint *bPathHint, depth int, iter func(item kind) bool,
) bool {
i, found := tr.find(n, pivot, hint, depth)
if !found {
if !n.leaf() {
if !tr.ascend((*n.children)[i], pivot, hint, depth+1, iter) {
return false
}
}
}
// We are either in the case that
// - node is found, we should iterate through it starting at `i`,
// the index it was located at.
// - node is not found, and TODO: fill in.
for ; i < len(n.items); i++ {
if !iter(n.items[i]) {
return false
}
if !n.leaf() {
if !(*n.children)[i+1].scan(iter) {
return false
}
}
}
return true
}
func (tr *bTree) Reverse(iter func(item kind) bool) {
if tr.rlock() {
defer tr.runlock()
}
if tr.root == nil {
return
}
tr.root.reverse(iter)
}
func (n *node) reverse(iter func(item kind) bool) bool {
if n.leaf() {
for i := len(n.items) - 1; i >= 0; i-- {
if !iter(n.items[i]) {
return false
}
}
return true
}
if !(*n.children)[len(*n.children)-1].reverse(iter) {
return false
}
for i := len(n.items) - 1; i >= 0; i-- {
if !iter(n.items[i]) {
return false
}
if !(*n.children)[i].reverse(iter) {
return false
}
}
return true
}
// Descend the tree within the range [pivot, first]
// Pass nil for pivot to scan all item in descending order
// Return false to stop iterating
func (tr *bTree) Descend(pivot kind, iter func(item kind) bool) {
if tr.rlock() {
defer tr.runlock()
}
if tr.root == nil {
return
}
tr.descend(tr.root, pivot, nil, 0, iter)
}
func (tr *bTree) descend(n *node, pivot kind,
hint *bPathHint, depth int, iter func(item kind) bool,
) bool {
i, found := tr.find(n, pivot, hint, depth)
if !found {
if !n.leaf() {
if !tr.descend((*n.children)[i], pivot, hint, depth+1, iter) {
return false
}
}
i--
}
for ; i >= 0; i-- {
if !iter(n.items[i]) {
return false
}
if !n.leaf() {
if !(*n.children)[i].reverse(iter) {
return false
}
}
}
return true
}
// Load is for bulk loading pre-sorted items
func (tr *bTree) Load(item kind) (kind, bool) {
if tr.lock() {
defer tr.unlock()
}
if tr.root == nil {
return tr.setHint(item, nil)
}
n := tr.cowLoad(&tr.root)
for {
n.count++ // optimistically update counts
if n.leaf() {
if len(n.items) < maxItems {
if tr.Less(n.items[len(n.items)-1], item) {
n.items = append(n.items, item)
tr.count++
return tr.empty, false
}
}
break
}
n = tr.cowLoad(&(*n.children)[len(*n.children)-1])
}
// revert the counts
n = tr.root
for {
n.count--
if n.leaf() {
break
}
n = (*n.children)[len(*n.children)-1]
}
return tr.setHint(item, nil)
}
// Min returns the minimum item in tree.
// Returns nil if the tree has no items.
func (tr *bTree) Min() (kind, bool) {
if tr.rlock() {
defer tr.runlock()
}
if tr.root == nil {
return tr.empty, false
}
n := tr.root
for {
if n.leaf() {
return n.items[0], true
}
n = (*n.children)[0]
}
}
// Max returns the maximum item in tree.
// Returns nil if the tree has no items.
func (tr *bTree) Max() (kind, bool) {
if tr.rlock() {
defer tr.runlock()
}
if tr.root == nil {
return tr.empty, false
}
n := tr.root
for {
if n.leaf() {
return n.items[len(n.items)-1], true
}
n = (*n.children)[len(*n.children)-1]
}
}
// PopMin removes the minimum item in tree and returns it.
// Returns nil if the tree has no items.
func (tr *bTree) PopMin() (kind, bool) {
if tr.lock() {
defer tr.unlock()
}
if tr.root == nil {
return tr.empty, false
}
n := tr.cowLoad(&tr.root)
var item kind
for {
n.count-- // optimistically update counts
if n.leaf() {
item = n.items[0]
if len(n.items) == minItems {
break
}
copy(n.items[:], n.items[1:])
n.items[len(n.items)-1] = tr.empty
n.items = n.items[:len(n.items)-1]
tr.count--
if tr.count == 0 {
tr.root = nil
}
return item, true
}
n = tr.cowLoad(&(*n.children)[0])
}
// revert the counts
n = tr.root
for {
n.count++
if n.leaf() {
break
}
n = (*n.children)[0]
}
return tr.deleteHint(item, nil)
}
// PopMax removes the minimum item in tree and returns it.
// Returns nil if the tree has no items.
func (tr *bTree) PopMax() (kind, bool) {
if tr.lock() {
defer tr.unlock()
}
if tr.root == nil {
return tr.empty, false
}
n := tr.cowLoad(&tr.root)
var item kind
for {
n.count-- // optimistically update counts
if n.leaf() {
item = n.items[len(n.items)-1]
if len(n.items) == minItems {
break
}
n.items[len(n.items)-1] = tr.empty
n.items = n.items[:len(n.items)-1]
tr.count--
if tr.count == 0 {
tr.root = nil
}
return item, true
}
n = tr.cowLoad(&(*n.children)[len(*n.children)-1])
}
// revert the counts
n = tr.root
for {
n.count++
if n.leaf() {
break
}
n = (*n.children)[len(*n.children)-1]
}
return tr.deleteHint(item, nil)
}
// GetAt returns the value at index.
// Return nil if the tree is empty or the index is out of bounds.
func (tr *bTree) GetAt(index int) (kind, bool) {
if tr.rlock() {
defer tr.runlock()
}
if tr.root == nil || index < 0 || index >= tr.count {
return tr.empty, false
}
n := tr.root
for {
if n.leaf() {
return n.items[index], true
}
i := 0
for ; i < len(n.items); i++ {
if index < (*n.children)[i].count {
break
} else if index == (*n.children)[i].count {
return n.items[i], true
}
index -= (*n.children)[i].count + 1
}
n = (*n.children)[i]
}
}
// DeleteAt deletes the item at index.
// Return nil if the tree is empty or the index is out of bounds.
func (tr *bTree) DeleteAt(index int) (kind, bool) {
if tr.lock() {
defer tr.unlock()
}
if tr.root == nil || index < 0 || index >= tr.count {
return tr.empty, false
}
var pathbuf [8]uint8 // track the path
path := pathbuf[:0]
var item kind
n := tr.cowLoad(&tr.root)
outer:
for {
n.count-- // optimistically update counts
if n.leaf() {
// the index is the item position
item = n.items[index]
if len(n.items) == minItems {
path = append(path, uint8(index))
break outer
}
copy(n.items[index:], n.items[index+1:])
n.items[len(n.items)-1] = tr.empty
n.items = n.items[:len(n.items)-1]
tr.count--
if tr.count == 0 {
tr.root = nil
}
return item, true
}
i := 0
for ; i < len(n.items); i++ {
if index < (*n.children)[i].count {
break
} else if index == (*n.children)[i].count {
item = n.items[i]
path = append(path, uint8(i))
break outer
}
index -= (*n.children)[i].count + 1
}
path = append(path, uint8(i))
n = tr.cowLoad(&(*n.children)[i])
}
// revert the counts
var hint bPathHint
n = tr.root
for i := 0; i < len(path); i++ {
if i < len(hint.path) {
hint.path[i] = uint8(path[i])
hint.used[i] = true
}
n.count++
if !n.leaf() {
n = (*n.children)[uint8(path[i])]
}
}
return tr.deleteHint(item, &hint)
}
// Height returns the height of the tree.
// Returns zero if tree has no items.
func (tr *bTree) Height() int {
if tr.rlock() {
defer tr.runlock()
}
var height int
if tr.root != nil {
n := tr.root
for {
height++
if n.leaf() {
break
}
n = (*n.children)[0]
}
}
return height
}
// Walk iterates over all items in tree, in order.
// The items param will contain one or more items.
func (tr *bTree) Walk(iter func(item []kind) bool) {
if tr.rlock() {
defer tr.runlock()
}
if tr.root != nil {
tr.root.walk(iter)
}
}
func (n *node) walk(iter func(item []kind) bool) bool {
if n.leaf() {
if !iter(n.items) {
return false
}
} else {
for i := 0; i < len(n.items); i++ {
(*n.children)[i].walk(iter)
if !iter(n.items[i : i+1]) {
return false
}
}
(*n.children)[len(n.items)].walk(iter)
}
return true
}
// Copy the tree. This is a copy-on-write operation and is very fast because
// it only performs a shadowed copy.
func (tr *bTree) Copy() *bTree {
if tr.lock() {
defer tr.unlock()
}
tr.cow = new(cow)
tr2 := new(bTree)
*tr2 = *tr
tr2.mu = new(sync.RWMutex)
tr2.cow = new(cow)
return tr2
}
func (tr *bTree) lock() bool {
if tr.locks {
tr.mu.Lock()
}
return tr.locks
}
func (tr *bTree) unlock() {
tr.mu.Unlock()
}
func (tr *bTree) rlock() bool {
if tr.locks {
tr.mu.RLock()
}
return tr.locks
}
func (tr *bTree) runlock() {
tr.mu.RUnlock()
}
// Iter represents an iterator
type bIter struct {
tr *bTree
locked bool
seeked bool
atstart bool
atend bool
stack []iterStackItem
item kind
}
type iterStackItem struct {
n *node
i int
}
// Iter returns a read-only iterator.
// The Release method must be called finished with iterator.
func (tr *bTree) Iter() bIter {
var iter bIter
iter.tr = tr
iter.locked = tr.rlock()
return iter
}
// Seek to item greater-or-equal-to key.
// Returns false if there was no item found.
func (iter *bIter) Seek(key kind) bool {
if iter.tr == nil {
return false
}
iter.seeked = true
iter.stack = iter.stack[:0]
if iter.tr.root == nil {
return false
}
n := iter.tr.root
for {
i, found := iter.tr.find(n, key, nil, 0)
iter.stack = append(iter.stack, iterStackItem{n, i})
if found {
return true
}
if n.leaf() {
if i == len(n.items) {
iter.stack = iter.stack[:0]
return false
}
return true
}
n = (*n.children)[i]
}
}
// First moves iterator to first item in tree.
// Returns false if the tree is empty.
func (iter *bIter) First() bool {
if iter.tr == nil {
return false
}
iter.atend = false
iter.atstart = false
iter.seeked = true
iter.stack = iter.stack[:0]
if iter.tr.root == nil {
return false
}
n := iter.tr.root
for {
iter.stack = append(iter.stack, iterStackItem{n, 0})
if n.leaf() {
break
}
n = (*n.children)[0]
}
s := &iter.stack[len(iter.stack)-1]
iter.item = s.n.items[s.i]
return true
}
// Last moves iterator to last item in tree.
// Returns false if the tree is empty.
func (iter *bIter) Last() bool {
if iter.tr == nil {
return false
}
iter.seeked = true
iter.stack = iter.stack[:0]
if iter.tr.root == nil {
return false
}
n := iter.tr.root
for {
iter.stack = append(iter.stack, iterStackItem{n, len(n.items)})
if n.leaf() {
iter.stack[len(iter.stack)-1].i--
break
}
n = (*n.children)[len(n.items)]
}
s := &iter.stack[len(iter.stack)-1]
iter.item = s.n.items[s.i]
return true
}
// First moves iterator to first item in tree.
// Returns false if the tree is empty.
func (iter *bIter) Release() {
if iter.tr == nil {
return
}
if iter.locked {
iter.tr.runlock()
iter.locked = false
}
iter.stack = nil
iter.tr = nil
}
// Next moves iterator to the next item in iterator.
// Returns false if the tree is empty or the iterator is at the end of
// the tree.
func (iter *bIter) Next() bool {
if iter.tr == nil {
return false
}
if !iter.seeked {
return iter.First()
}
if len(iter.stack) == 0 {
if iter.atstart {
return iter.First() && iter.Next()
}
return false
}
s := &iter.stack[len(iter.stack)-1]
s.i++
if s.n.leaf() {
if s.i == len(s.n.items) {
for {
iter.stack = iter.stack[:len(iter.stack)-1]
if len(iter.stack) == 0 {
iter.atend = true
return false
}
s = &iter.stack[len(iter.stack)-1]
if s.i < len(s.n.items) {
break
}
}
}
} else {
n := (*s.n.children)[s.i]
for {
iter.stack = append(iter.stack, iterStackItem{n, 0})
if n.leaf() {
break
}
n = (*n.children)[0]
}
}
s = &iter.stack[len(iter.stack)-1]
iter.item = s.n.items[s.i]
return true
}
// Prev moves iterator to the previous item in iterator.
// Returns false if the tree is empty or the iterator is at the beginning of
// the tree.
func (iter *bIter) Prev() bool {
if iter.tr == nil {
return false
}
if !iter.seeked {
return false
}
if len(iter.stack) == 0 {
if iter.atend {
return iter.Last() && iter.Prev()
}
return false
}
s := &iter.stack[len(iter.stack)-1]
if s.n.leaf() {
s.i--
if s.i == -1 {
for {
iter.stack = iter.stack[:len(iter.stack)-1]
if len(iter.stack) == 0 {
iter.atstart = true
return false
}
s = &iter.stack[len(iter.stack)-1]
s.i--
if s.i > -1 {
break
}
}
}
} else {
n := (*s.n.children)[s.i]
for {
iter.stack = append(iter.stack, iterStackItem{n, len(n.items)})
if n.leaf() {
iter.stack[len(iter.stack)-1].i--
break
}
n = (*n.children)[len(n.items)]
}
}
s = &iter.stack[len(iter.stack)-1]
iter.item = s.n.items[s.i]
return true
}
// Item returns the current iterator item.
func (iter *bIter) Item() kind {
return iter.item
}