In the .Net BCL is there a collection data structure similar to list that has a maximum capacity, say configured to 100 items, which when item 101 is added the original first item is popped/removed from the collection thus ensuring the item count never exceeds 100.
I’m using .net 3.5
Thanks in advance
There's not one available, but it should be easy to write a function to accomplish this with an array or collection.
You could inherit from whatever existing collection was most appropriate (Stack, Dequeue, List, CollectionBase, etc.) and implement this feature yourself. Just override or replace the Add() method.
There is no such collection available but one is fairly easy to write. The best way to do this is to create a new collection type that encapsulates an existing collection type.
For example
public class FixedSizeList<T> : IList<T> {
private List<T> _list = new List<T>();
private int _capacity = 100;
public void Add(T value) {
_list.Add(value);
while ( _list.Count > _capacity ) {
_list.RemoveAt(0);
}
}
// Rest omitted for brevity
}
A couple of answers suggested inheritance as a mechanism. This is certainly not a good route especially if you are deriving from one of the generic collections. These collections are not designed to be inherited from and it is very easy to accidentally circumvent any checks you make an capacity as the result of an Add or Remove method.
The primary reason is that these methods are not virtual so they cannot be overridden. You would be forced to either declare an Add method with a different name (thus confusing your users) or re-declare Add with the new syntax. The latter is a very unsafe because as soon as an instance of your class is passed to a reference of the base type, all of your methods will not be called and the list can grow past capacity.
EDIT
As the comment section discussion has indicated, implementing List<T> is not the best approach here. The reason why is that it violates the substitution principle in several cases. The easiest way to display the problem is imagine if my implementation was passed to the following method. This code should pass for any IList<T> implementation but would fail in this case if the list was at capacity.
public void Example<T>(IList<T> list, T value) {
var count = list.Count;
list.Add(value);
var addedValue = list[count];
}
The only collection interface that can be validly implemented for the specified collection is IEnumerable<T>. I've left my implementation up there as an example. But see ShuggyCoUk's answer for an IEnumerable<T> implementation:
You want a circular buffer. This other SO question already talks about that and it might help provide some ideas for you.
What you are describing is a circular buffer. I use these occasionally and recently ported some older code into a genericised C# class (attached). This code is part of SharpNeat V2 development.
This has O(1) performance on add and remove oprations whereas the solution posted that encapsulates a List is O(n). This is because removing the 0th item in a list causes all other items to be shuffled down to fill the gap.
using System;
using System.Collections.Generic;
using System.Text;
namespace SharpNeat.Utility
{
///
/// This is a generic circular buffer of items of type T. A circular buffer must be assigned
/// a capacity at construction time. Items can be enqueued indefintely, but when the buffer's
/// capacity is reached the oldest values in the buffer are overwritten, thus the buffer is best
/// thought of as a circular array or buffer.
///
public class CircularBuffer
{
///
/// Internal array that stores the circular buffer's values.
///
protected T[] _buff;
///
/// The index of the previously enqueued item. -1 if buffer is empty.
///
protected int _headIdx;
///
/// The index of the next item to be dequeued. -1 if buffer is empty.
///
protected int _tailIdx;
#region Constructors
///
/// Constructs a circular buffer with the specified capacity.
///
///
public CircularBuffer(int capacity)
{
_buff = new T[capacity];
_headIdx = _tailIdx = -1;
}
#endregion
#region Properties
///
/// Gets the number of items in the buffer. Returns the buffer's capacity
/// if it is full.
///
public int Length
{
get
{
if(_headIdx == -1)
return 0;
if(_headIdx > _tailIdx)
return (_headIdx - _tailIdx) + 1;
if(_tailIdx > _headIdx)
return (_buff.Length - _tailIdx) + _headIdx + 1;
return 1;
}
}
#endregion
#region Public Methods
///
/// Clear the buffer.
///
public virtual void Clear()
{
_headIdx = _tailIdx = -1;
}
///
/// Enqueue a new item. This overwrites the oldest item in the buffer if the buffer
/// has reached capacity.
///
///
public virtual void Enqueue(T item)
{
if(_headIdx == -1)
{ // buffer is currently empty.
_headIdx = _tailIdx = 0;
_buff[0] = item;
return;
}
// Determine the index to write to.
if(++_headIdx == _buff.Length)
{ // Wrap around.
_headIdx = 0;
}
if(_headIdx == _tailIdx)
{ // Buffer overflow. Increment tailIdx.
if(++_tailIdx == _buff.Length)
{ // Wrap around.
_tailIdx=0;
}
_buff[_headIdx] = item;
return;
}
_buff[_headIdx] = item;
return;
}
///
/// Remove the oldest item from the back end of the buffer and return it.
///
///
public virtual T Dequeue()
{
if(_tailIdx == -1)
{ // buffer is currently empty.
throw new InvalidOperationException("buffer is empty.");
}
T item = _buff[_tailIdx];
if(_tailIdx == _headIdx)
{ // The buffer is now empty.
_headIdx=_tailIdx=-1;
return item;
}
if(++_tailIdx == _buff.Length)
{ // Wrap around.
_tailIdx = 0;
}
return item;
}
///
/// Pop the most recently added item from the front end of the buffer and return it.
///
///
public virtual T Pop()
{
if(_tailIdx == -1)
{ // buffer is currently empty.
throw new InvalidOperationException("buffer is empty.");
}
T item = _buff[_headIdx];
if(_tailIdx == _headIdx)
{ // The buffer is now empty.
_headIdx = _tailIdx =- 1;
return item;
}
if(--_headIdx==-1)
{ // Wrap around.
_headIdx=_buff.Length-1;
}
return item;
}
#endregion
}
}
you can also just override Collection
public class FixedSizeList<T> : Collection<T>
{
public int MaxItems {get;set;}
protected override void InsertItem(int index, T item){
base.InsertItem(index, item);
while (base.Count > MaxItems) {
base.RemoveItem(0);
}
}
}
A really simple rolling window
public class RollingWindow<T> : IEnumerable<T>
{
private readonly T[] data;
private int head;
private int nextInsert = 0;
public RollingWindow(int size)
{
if (size < 1)
throw new Exception();
this.data = new T[size];
this.head = -size;
}
public void Add(T t)
{
data[nextInsert] = t;
nextInsert = (nextInsert + 1) % data.Length;
if (head < 0)
head++;
}
public IEnumerator<T> GetEnumerator()
{
if (head < 0)
{
for (int i = 0; i < nextInsert; i++)
yield return data[i];
}
else
{
for(int i = 0; i < data.Length; i++)
yield return data[(nextInsert + i) % data.Length];
}
}
System.Collections.IEnumerator
System.Collections.IEnumerable.GetEnumerator()
{
return this.GetEnumerator();
}
}