问题
I'm after a collection that I can iterate through very fast. I'll also be adding items and removing (specific) items fairly regularly and so ideally would like those operations to be fast too.
I'm developing on the xbox and so am restricted to the compact framework (more or less). It's very important that I keep garbage and object allocations to a minimum, so anything where i can pre-allocate space for my objects would be great.
I'll be storing uints (but can be ints if needed) in the collection. A generic solution would be good though, as i'm sure i'll have a need in the future.
A .net collection would be ideal, failing that something light-weight and open source would be great.
Is there a collection class that would suit my needs? If not, how would I go about creating one?
To elaborate a bit, they are object id's that a class should process each frame. They'll generally be added in ascending order, with gaps. There's no upper limit. Any could be removed though, which would leave gaps.
Iteration order isn't completely important, but it would be very useful (particularly for debugging) if it was in a consistent order.
回答1:
I've got two suggestions to try:
Firstly, what about using Reflector or ILSpy to look inside Generic List<T>? I assume that implementation isn't in the CF or you could use it. Generic List<T> is array backed and uses a doubling algorithm starting at length 4 array, every call to .Add over the Capacity causes it to double and perform an Array.Copy into the new array. It doesn't resize unless you explicitly set Capacity to zero so beware from a memory point of view. Adds are one thing but each Remove will cause the array to be copied and left-shifted after the index that was removed.
The second suggestion is this - what about about creating a custom class which wraps an integer array to handle this which uses a similar double algorithm (or quadrupling) to Generic List<T> (to handle resizing) but starts at say size 256. You could add object integer id's to this out of order as fast as you like and it won't double too often. You could also simulate a remove by designating (int)-1 or uint.MaxValue as a "null index" meaning no Array.Copy on remove. Then, apply some sort of fast sorting per frame to sort the object index array before drawing. If you sort ascending all your -1s will appear at the start (or uint.MaxValues at end) and can be ignored. You can periodically "collect" the index array by resizing and removing the preceeding -1's on a separate thread (beware - use locking ;)
What do you think? Just thinking you will offset some computation once per frame for fast sorting (not expensive on xbox vs. memory allocation/collection on each Remove and some Adds (expensive).
UPDATE - BlockArray - List<T[]> where T is fixed size array
A further comment on this. I was recently experimenting with the most efficient data-structure for dynamically sized lists and discovered by experimentation that array blocks - a class which is backed by a List of T[], where each T[] was an array of fixed size block, e.g. 512, 4096, 8192 as several advantages over a plain List<T>.
I found that an implementation of Add() (where size is unknown) in a List<T[]> vastly outperformed Add() for List<T>, especially when the total size became larger. This is due to List<T>'s doubling algorithm which requests a new array 2x the size of the old, and memcpy's the old array over whenever size is exceeded.
Iterating speed is similar. Pre-allocation (pre-defined capacity) was much slower than List<T> for small block sizes (512), but only slightly slower for large block sizes (8192). Removes are problematic, as require copying/left shifting many blocks.
What's interesting is in a List<T[]> (block list), you can cache/pool the blocks. If small enough, blocks of T[] fit into the small object heap (favouring compaction, quicker allocation), may fit into the L2 cache and a number of blocks could be pre-allocated and pooled
回答2:
How about using Mono's HashSet<T>?
https://github.com/mono/mono/blob/master/mcs/class/System.Core/System.Collections.Generic/HashSet.cs
It's licensed under MIT X11, which is a permissive license.
Iteration order might be a problem depending on how T implements GetHashCode, but that shouldn't be a problem when using uint. The default implementation of GetHashCode on the other hand returns different values on different instances, which might lead to different iteration order.
Iteration order might also depend on the order items are added. i.e. two collections containing the same items might iterate in a different order if the items were added in a different order.
There is also a SortedSet<T> collection, but I don't know what performance characteristics it has.
回答3:
How about a custom class, SparseList<T>:
- a list that allows you to remove items by setting missing values to null (or for SparseValueList, a special value like -1 (like Dr ABT's solution), 0 (default(T)), or int.MinValue, or uint.MaxValue,) and
- then maintaining a list of deleted indices (a Stack<int>). Then when you need to add to the list, it pops a deleted index from the stack and adds the value there. (For multithreading, perhaps ConcurrentQueue<int> would be another idea.)
- the enumerator can skip over deleted items (to support foreach)
- items can be removed from the collection during enumeration! (I have to do this a lot, so this is nice.)
- indexer provides raw access to list that contains nulls. So if you use a for(;;) loop, be prepared to filter out nulls.
- call Compact() if/when you want to remove all the nulls
- If one never calls compact during a game, I am worried about iterating through a huge number of nulls. So for an experimental alternative to compact, see SparseListCleaningEnumerator: auto-shrink the list every time it is enumerated, at least for single-threaded situations: when MoveNext moves away from an item, it peeks the stack to see if the index is lower and if so it assigns the item to the lower index, removing it from the current position, which will shrink the list. Balancing might take many iterations and involve multiple moves before optimization, unless the stack is replaced with a sortedlist, or the stack is sorted occasionally. If the last value is null, this won't work, because the index will be buried in the free index stack (replacing the stack with something sorted would avoid this).
I implemented this (not yet tested) but I'm storing actual references to my game entities instead of id's, so you'll need to adapt this for int's or Nullable somehow. (Ok to make sure I answer your question's int/uint requirement I also added a SparseValueList<T> that's slightly different, using default(T) instead of null. This means you can't use 0 in the list.) You could perhaps take out versioning if you don't think you need it -- most games may not.
/// <summary>
/// Specifying null as value has unspecified results.
/// CopyTo may contain nulls.
/// </summary>
/// <typeparam name="T"></typeparam>
public class SparseList<T> : IList<T>
where T : class
{
int version = 0;
List<T> list = new List<T>();
Stack<int> freeIndices = new Stack<int>();
public int Capacity { get { return list.Capacity; } set { list.Capacity = value; } }
public void Compact()
{
var sortedIndices = freeIndices.ToList();
foreach (var i in sortedIndices.OrderBy(x => x).Reverse())
{
list.RemoveAt(i);
}
freeIndices.Clear();
list.Capacity = list.Count;
version++; // breaks open enumerators
}
public int IndexOf(T item)
{
return list.IndexOf(item);
}
/// <summary>
/// Slow (forces a compact), not recommended
/// </summary>
/// <param name="index"></param>
/// <param name="item"></param>
public void Insert(int index, T item)
{
// One idea: remove index from freeIndices if it's in there. Stack doesn't support this though.
Compact(); // breaks the freeIndices list, so apply it before insert
list.Insert(index, item);
version++; // breaks open enumerators
}
public void RemoveAt(int index)
{
if (index == Count - 1) { list.RemoveAt(index); }
else { list[index] = null; freeIndices.Push(index); }
//version++; // Don't increment version for removals
}
public T this[int index]
{
get
{
return list[index];
}
set
{
if (value == null) throw new ArgumentNullException();
list[index] = value;
}
}
public void Add(T item)
{
if (item == null) throw new ArgumentNullException();
if (freeIndices.Count == 0) { list.Add(item); return; }
list[freeIndices.Pop()] = item;
//version++; // Don't increment version for additions? It could result in missing the new value, but shouldn't break open enumerators
}
public void Clear()
{
list.Clear();
freeIndices.Clear();
version++;
}
public bool Contains(T item)
{
if (item == null) return false;
return list.Contains(item);
}
/// <summary>
/// Result may contain nulls
/// </summary>
/// <param name="array"></param>
/// <param name="arrayIndex"></param>
public void CopyTo(T[] array, int arrayIndex)
{
list.CopyTo(array, arrayIndex);
}
//public void CopyNonNullTo(T[] array, int arrayIndex)
//{
//}
/// <summary>
/// Use this for iterating via for loop.
/// </summary>
public int Count { get { return list.Count; } }
/// <summary>
/// Don't use this for for loops! Use Count.
/// </summary>
public int NonNullCount
{
get { return list.Count - freeIndices.Count; }
}
public bool IsReadOnly
{
get { return false; }
}
public bool Remove(T item)
{
int i = list.IndexOf(item);
if (i < 0) return false;
if (i == list.Count - 1)
{
// Could throw . Could add check in
list.RemoveAt(i);
}
else
{
list[i] = null;
freeIndices.Push(i);
}
//version++; // Don't increment version for removals
return true;
}
public IEnumerator<T> GetEnumerator()
{
return new SparseListEnumerator(this);
}
private class SparseListEnumerator : IEnumerator<T>, IRemovingEnumerator
{
SparseList<T> list;
int version;
int index = -1;
public SparseListEnumerator(SparseList<T> list)
{
this.list = list;
this.version = list.version;
//while (Current == null && MoveNext()) ;
}
public T Current
{
get {
if (index >= list.Count) return null; // Supports removing last items of collection without throwing on Enumerator access
return list[index];
}
}
public void Dispose()
{
list = null;
}
object IEnumerator.Current
{
get { return Current; }
}
public bool MoveNext()
{
do
{
if (version != list.version) { throw new InvalidOperationException("Collection modified"); }
index++;
return index < list.Count;
} while (Current == null);
}
public void Reset()
{
index = -1;
version = list.version;
}
/// <summary>
/// Accessing Current after RemoveCurrent may throw a NullReferenceException or return null.
/// </summary>
public void RemoveCurrent()
{
list.RemoveAt(index);
}
}
private class SparseListCleaningEnumerator : IEnumerator<T>, IRemovingEnumerator
{
SparseList<T> list;
int version;
int index = -1;
public SparseListCleaningEnumerator(SparseList<T> list)
{
this.list = list;
this.version = list.version;
//while (Current == null && MoveNext()) ;
}
public T Current
{
get
{
if (index >= list.Count) return null; // Supports removing last items of collection without throwing on Enumerator access
return list[index];
}
}
public void Dispose()
{
list = null;
}
object IEnumerator.Current
{
get { return Current; }
}
public bool MoveNext()
{
do
{
if (version != list.version) { throw new InvalidOperationException("Collection modified"); }
if (index > 0
&& Current != null // only works for values that are set, otherwise the index is buried in the free index stack somewhere
)
{
int freeIndex = list.freeIndices.Peek();
if (freeIndex < index)
{
list.freeIndices.Pop();
list[freeIndex] = list[index];
list.RemoveAt(index);
}
}
index++;
return index < list.Count;
} while (Current == null);
}
public void Reset()
{
index = -1;
version = list.version;
}
/// <summary>
/// Accessing Current after RemoveCurrent may throw a NullReferenceException or return null.
/// </summary>
public void RemoveCurrent()
{
list.RemoveAt(index);
}
}
IEnumerator IEnumerable.GetEnumerator()
{
return GetEnumerator();
}
}
/// <summary>
/// Like SparseList but Supports value types, using default(T) in place of null.
/// This means values of default(T) are not permitted as values in the collection.
/// CopyTo may contain default(T).
/// TODO: Use EqualityComparer<T>.Default instead of default(T).Equals()
/// </summary>
/// <typeparam name="T"></typeparam>
public class SparseValueList<T> : IList<T>
{
int version = 0;
List<T> list = new List<T>();
Stack<int> freeIndices = new Stack<int>();
public int Capacity { get { return list.Capacity; } set { list.Capacity = value; } }
public void Compact()
{
var sortedIndices = freeIndices.ToList();
foreach (var i in sortedIndices.OrderBy(x => x).Reverse())
{
list.RemoveAt(i);
}
freeIndices.Clear();
list.Capacity = list.Count;
version++; // breaks open enumerators
}
public int IndexOf(T item)
{
return list.IndexOf(item);
}
/// <summary>
/// Slow (forces a compact), not recommended
/// </summary>
/// <param name="index"></param>
/// <param name="item"></param>
public void Insert(int index, T item)
{
// One idea: remove index from freeIndices if it's in there. Stack doesn't support this though.
Compact(); // breaks the freeIndices list, so apply it before insert
list.Insert(index, item);
version++; // breaks open enumerators
}
public void RemoveAt(int index)
{
if (index == Count - 1) { list.RemoveAt(index); }
else { list[index] = default(T); freeIndices.Push(index); }
//version++; // Don't increment version for removals
}
public T this[int index]
{
get
{
return list[index];
}
set
{
if (default(T).Equals(value)) throw new ArgumentNullException();
list[index] = value;
}
}
public void Add(T item)
{
if (default(T).Equals(item)) throw new ArgumentNullException();
if (freeIndices.Count == 0) { list.Add(item); return; }
list[freeIndices.Pop()] = item;
//version++; // Don't increment version for additions? It could result in missing the new value, but shouldn't break open enumerators
}
public void Clear()
{
list.Clear();
freeIndices.Clear();
version++;
}
public bool Contains(T item)
{
if (default(T).Equals(item)) return false;
return list.Contains(item);
}
/// <summary>
/// Result may contain default(T)'s
/// </summary>
/// <param name="array"></param>
/// <param name="arrayIndex"></param>
public void CopyTo(T[] array, int arrayIndex)
{
list.CopyTo(array, arrayIndex);
}
//public void CopyNonNullTo(T[] array, int arrayIndex)
//{
//}
/// <summary>
/// Use this for iterating via for loop.
/// </summary>
public int Count { get { return list.Count; } }
/// <summary>
/// Don't use this for for loops! Use Count.
/// </summary>
public int NonNullCount
{
get { return list.Count - freeIndices.Count; }
}
public bool IsReadOnly
{
get { return false; }
}
public bool Remove(T item)
{
int i = list.IndexOf(item);
if (i < 0) return false;
if (i == list.Count - 1)
{
// Could throw . Could add check in
list.RemoveAt(i);
}
else
{
list[i] = default(T);
freeIndices.Push(i);
}
//version++; // Don't increment version for removals
return true;
}
public IEnumerator<T> GetEnumerator()
{
return new SparseValueListEnumerator(this);
}
private class SparseValueListEnumerator : IEnumerator<T>, IRemovingEnumerator
{
SparseValueList<T> list;
int version;
int index = -1;
public SparseValueListEnumerator(SparseValueList<T> list)
{
this.list = list;
this.version = list.version;
//while (Current == default(T) && MoveNext()) ;
}
public T Current
{
get
{
if (index >= list.Count) return default(T); // Supports removing last items of collection without throwing on Enumerator access
return list[index];
}
}
public void Dispose()
{
list = null;
}
object IEnumerator.Current
{
get { return Current; }
}
public bool MoveNext()
{
do
{
if (version != list.version) { throw new InvalidOperationException("Collection modified"); }
index++;
return index < list.Count;
} while (default(T).Equals(Current));
}
public void Reset()
{
index = -1;
version = list.version;
}
/// <summary>
/// Accessing Current after RemoveCurrent may throw a NullReferenceException or return default(T).
/// </summary>
public void RemoveCurrent()
{
list.RemoveAt(index);
}
}
private class SparseValueListCleaningEnumerator : IEnumerator<T>, IRemovingEnumerator
{
SparseValueList<T> list;
int version;
int index = -1;
public SparseValueListCleaningEnumerator(SparseValueList<T> list)
{
this.list = list;
this.version = list.version;
while (default(T).Equals(Current) && MoveNext()) ;
}
public T Current
{
get
{
if (index >= list.Count) return default(T); // Supports removing last items of collection without throwing on Enumerator access
return list[index];
}
}
public void Dispose()
{
list = null;
}
object IEnumerator.Current
{
get { return Current; }
}
public bool MoveNext()
{
do
{
if (version != list.version) { throw new InvalidOperationException("Collection modified"); }
if (index > 0
&& (!default(T).Equals(Current)) // only works for values that are set, otherwise the index might be buried in the stack somewhere
)
{
int freeIndex = list.freeIndices.Peek();
if (freeIndex < index)
{
list.freeIndices.Pop();
list[freeIndex] = list[index];
list.RemoveAt(index);
}
}
index++;
return index < list.Count;
} while (default(T).Equals(Current));
}
public void Reset()
{
index = -1;
version = list.version;
}
/// <summary>
/// Accessing Current after RemoveCurrent may throw a NullReferenceException or return default(T).
/// </summary>
public void RemoveCurrent()
{
list.RemoveAt(index);
}
}
IEnumerator IEnumerable.GetEnumerator()
{
return GetEnumerator();
}
}
来源：https://stackoverflow.com/questions/8761238/collection-with-very-fast-iterating-and-good-addition-and-remove-speeds