我刚刚看到了一个以通用方式实现的优先级队列可以将满足接口的类型放入队列中。这是围棋的方式吗？还是这会带来任何问题？
// Copyright 2012 Stefan Nilsson
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
// Package prio provides a priority queue.
// The queue can hold elements that implement the two methods of prio.Interface.
package prio
/*
A type that implements prio.Interface can be inserted into a priority queue.
The simplest use case looks like this:
type myInt int
func (x myInt) Less(y prio.Interface) bool { return x < y.(myInt) }
func (x myInt) Index(i int) {}
To use the Remove method you need to keep track of the index of elements
in the heap, e.g. like this:
type myType struct {
value int
index int // index in heap
}
func (x *myType) Less(y prio.Interface) bool { return x.value < y.(*myType).value }
func (x *myType) Index(i int) { x.index = i }
*/
type Interface interface {
// Less returns whether this element should sort before element x.
Less(x Interface) bool
// Index is called by the priority queue when this element is moved to index i.
Index(i int)
}
// Queue represents a priority queue.
// The zero value for Queue is an empty queue ready to use.
type Queue struct {
h []Interface
}
// New returns an initialized priority queue with the given elements.
// A call of the form New(x...) uses the underlying array of x to implement
// the queue and hence might change the elements of x.
// The complexity is O(n), where n = len(x).
func New(x ...Interface) Queue {
q := Queue{x}
heapify(q.h)
return q
}
// Push pushes the element x onto the queue.
// The complexity is O(log(n)) where n = q.Len().
func (q *Queue) Push(x Interface) {
n := len(q.h)
q.h = append(q.h, x)
up(q.h, n) // x.Index(n) is done by up.
}
// Pop removes a minimum element (according to Less) from the queue and returns it.
// The complexity is O(log(n)), where n = q.Len().
func (q *Queue) Pop() Interface {
h := q.h
n := len(h) - 1
x := h[0]
h[0], h[n] = h[n], nil
h = h[:n]
if n > 0 {
down(h, 0) // h[0].Index(0) is done by down.
}
q.h = h
x.Index(-1) // for safety
return x
}
// Peek returns, but does not remove, a minimum element (according to Less) of the queue.
func (q *Queue) Peek() Interface {
return q.h[0]
}
// Remove removes the element at index i from the queue and returns it.
// The complexity is O(log(n)), where n = q.Len().
func (q *Queue) Remove(i int) Interface {
h := q.h
n := len(h) - 1
x := h[i]
h[i], h[n] = h[n], nil
h = h[:n]
if i < n {
down(h, i) // h[i].Index(i) is done by down.
up(h, i)
}
q.h = h
x.Index(-1) // for safety
return x
}
// Len returns the number of elements in the queue.
func (q *Queue) Len() int {
return len(q.h)
}
// Establishes the heap invariant in O(n) time.
func heapify(h []Interface) {
n := len(h)
for i := n - 1; i >= n/2; i-- {
h[i].Index(i)
}
for i := n/2 - 1; i >= 0; i-- { // h[i].Index(i) is done by down.
down(h, i)
}
}
// Moves element at position i towards top of heap to restore invariant.
func up(h []Interface, i int) {
for {
parent := (i - 1) / 2
if i == 0 || h[parent].Less(h[i]) {
h[i].Index(i)
break
}
h[parent], h[i] = h[i], h[parent]
h[i].Index(i)
i = parent
}
}
// Moves element at position i towards bottom of heap to restore invariant.
func down(h []Interface, i int) {
for {
n := len(h)
left := 2*i + 1
if left >= n {
h[i].Index(i)
break
}
j := left
if right := left + 1; right < n && h[right].Less(h[left]) {
j = right
}
if h[i].Less(h[j]) {
h[i].Index(i)
break
}
h[i], h[j] = h[j], h[i]
h[i].Index(i)
i = j
}
}
一般来说，这个包不是最好的选择，但如果你喜欢它并且它满足你的需求，就使用它。我看不出有什么大问题。
与容器/堆相比，这个包的概念是将接口放在节点上，而不是容器上。容器/堆通过使用您的容器提供了更大的灵活性。(你可能已经在一个容器中有了节点，而这个容器甚至可能不是切片。它只需要是可索引的。)另一方面，你可能不关心容器，很乐意让包为你管理它，这是一种常见的情况。这个包的索引管理是容器/堆上的一个很好的功能，尽管即使不需要索引管理，它也会增加方法调用的开销。
总有取舍。容器/堆非常通用。这个包使用一个较小的方法集(2而不是5)，并在上面添加索引管理，但在某些情况下，这只是牺牲了一点通用性和性能。(如果你真的在乎的话，你会想进行基准测试。还有其他差异可能会让指数调用的开销相形见绌。)