linksplatform / collections

LinksPlatform's Platform.Collections Class Library

	C++
	C#

Collections

LinksPlatform's Platform.Collections Class Library.

Namespace: Platform.Collections

Forked from: Konard/LinksPlatform/Platform/Platform.Helpers/Collections

NuGet package: Platform.Collections

Documentation

PDF file with code for e-readers.

Depend on

• System.Numerics.Vectors
• Platform.Interfaces
• Platform.Disposables
• Platform.Ranges
• Platform.Random

Dependent libraries

• Platform.Threading
• Platform.Reflection
• Platform.Communication.Protocol.Lino

collections's Issues

Compare ArrayString and ArraySegment from .NET 5

Collections/csharp/Platform.Collections/Arrays/ArrayString.cs

Line 8 in f35bae2

public class ArrayString<T> : Segment<T>

https://docs.microsoft.com/en-us/dotnet/api/system.arraysegment-1?view=net-5.0

Add extension methods for Arrays and Lists

• GetElementOrDefault
• TryGetElement

Looks like ArrayString is similar to a Segment

Collections/Arrays/ArrayString.cs

Line 3 in a76be8a

public class ArrayString<T>

Collections/Segments/Segment.cs

Line 7 in a76be8a

public class Segment<T> : IEquatable<Segment<T>>, IList<T>

Compare hash functions quality and select the best

Collections/Segments/Segment.cs

Line 31 in a76be8a

public override int GetHashCode()

Collections/Lists/IListExtensions.cs

Line 55 in a76be8a

public static int GenerateHashCode<T>(this IList<T> list)

Add Ensure.NotEmptyAndNotWhiteSpace to EnsureExtensions

And use it here:

https://github.com/linksplatform/Memory/blob/51d78fb523cb55c0784950db95d3e746e6617443/FileMappedResizableDirectMemory.cs#L35

Compare HashSet and BitString intersection performance

Rename the method into Skip

Collections/csharp/Platform.Collections/Lists/IListExtensions.cs

Line 329 in ed094f1

public static T[] SkipFirst<T>(this IList<T> list, int skip)

Use updated PNG logo icon

Collections/Platform.Collections.csproj

Line 14 in 56c3b06

<PackageIconUrl>https://raw.githubusercontent.com/linksplatform/Documentation/0a9177c11892ebb2fe467dc3ea9eaf5a5588f610/doc/Avatar-rainbow.png</PackageIconUrl>

New logo icon:
https://raw.githubusercontent.com/linksplatform/Documentation/18469f4d033ee9a5b7b84caab9c585acab2ac519/doc/Avatar-rainbow-icon-64x64.png

Check actual performance

Collections/Arrays/ArrayPool.cs

Line 6 in e4fbd36

/// TODO: Check actual performance

Rework Walkers in the C++ code to concepts instead of interfaces style

virtuals methods does not support template parameters
This is no good

Collections/cpp/Platform.Collections/Segments/Walkers/AllSegmentsWalkerExtensions.h

Lines 3 to 10 in 188927d

	template <Platform::Collections::System::Array TArray, typename TSegment>
	requires
	std::derived_from<TSegment, std::span<char>> &&
	requires(TArray array, std::string string) {TArray(string.begin(), string.end());}
	static void WalkAll(AllSegmentsWalkerBase<char, TArray, TSegment> walker, std::string string)
	{
	walker.WalkAll(TArray(string.begin(), string.end()));
	}

Collections/cpp/Platform.Collections/Segments/Walkers/AllSegmentsWalkerBase[T].h

Line 14 in 188927d

protected: TSegment CreateSegment(TArray& elements, std::int32_t offset, std::int32_t length)

Collections/cpp/Platform.Collections/Segments/Walkers/DuplicateSegmentsWalkerBase[T, TSegment].h

Lines 4 to 6 in 188927d

	template <typename T, Platform::Collections::System::Array TArray, typename TSegment>
	requires std::derived_from<TSegment, std::span<T>>
	class DuplicateSegmentsWalkerBase<T, TArray, TSegment> : public AllSegmentsWalkerBase<T, TArray, TSegment>

You need to do this so that you do not need to statically set the type of the enumerated object
Since there is no need to save it, we should stick to the following style:

void WalkAll(IEnumerable auto) { ... }

After fixing #25 add Clear method into IStackExtensions

Proposed code

while(!_stack.IsEmpty)
{
    _ = _stack.Pop();
}

This method will be used here:

https://github.com/linksplatform/Data.Doublets/blob/93ad6046c5c2273623c7ff2b2264d59326b0e827/Sequences/Walkers/SequenceWalkerBase.cs#L16

Implement better bitwise and/or/not algorithms using Vector for BitString

https://stackoverflow.com/questions/55931034/system-numerics-vectort-on-large-data-sets

Use PackageLicenseExpression

Collections/Platform.Collections.csproj

Line 16 in 56c3b06

<PackageLicenseUrl>https://github.com/linksplatform/Collections/blob/master/LICENSE</PackageLicenseUrl>

Try to use System.Runtime.Intrinsics for BitString

Remove blank lines

Collections/Lists/IListExtensions.cs

Line 35 in e4fbd36

Rename to CharSegment

Collections/Segments/CharsSegment.cs

Line 6 in e4fbd36

public unsafe class CharsSegment : Segment<char>

Compare ArrayPool and ConfigurableArrayPool from .NET 5

https://github.com/dotnet/corert/blob/master/src/System.Private.CoreLib/shared/System/Buffers/ConfigurableArrayPool.cs

• Performance should be compared.
• Are memory leaks possible?
• Are both implementations thread safe?

Can be speed up, if BitString will accept array of integers in the constructor

Collections/Platform.Collections/BitStringExtensions.cs

Line 11 in 0e99558

public static void SetRandomBits(this BitString @string)

Rename class to filename

Collections/Lists/IListExtensions.cs

Line 7 in a76be8a

public static class ListExtensions

Add comments on public classes and methods

Platform.Collections

• BitString
• BitStringExtensions
• EnsureExtensions
• ICollectionExtensions
• IDictionaryExtensions
• StringExtensions

Platform.Collections/Arrays

• ArrayFiller[TElement, TReturnConstant]
• ArrayFiller[TElement]
• ArrayPool
• ArrayPool[T]
• ArrayString
• CharArrayExtensions
• GenericArrayExtensions

Platform.Collections/Concurrent

• ConcurrentQueueExtensions
• ConcurrentStackExtensions

Platform.Collections/Lists

• CharIListExtensions
• IListComparer
• IListEqualityComparer
• IListExtensions
• ListFiller

Platform.Collections/Segments

• CharSegment
• Segment

Platform.Collections/Segments/Walkers

• AllSegmentsWalkerBase
• AllSegmentsWalkerBase[T, TSegment]
• AllSegmentsWalkerBase[T]
• AllSegmentsWalkerExtensions
• DictionaryBasedDuplicateSegmentsWalkerBase[T, Segment]
• DictionaryBasedDuplicateSegmentsWalkerBase[T]
• DuplicateSegmentsWalkerBase[T, TSegment]
• DuplicateSegmentsWalkerBase[T]

Platform.Collections/Sets

• ISetExtensions
• SetFiller

Platform.Collections/Stacks

• DefaultStack
• IStack
• IStackExtensions
• IStackFactory
• StackExtensions

Platform.Collections/Trees

• Node

Use BenchmarkDotNet for comparing the performance of different implementations of BitString methods

Check is System.Numeric.Vector is really faster than regular instructions

Implement VectorXor for BitString using System.Numerics.Vector<T>

Make the semantic of Segment as close to ArraySegment as possible

Collections/Platform.Collections/Segments/Segment.cs

Line 12 in f490499

public class Segment<T> : IEquatable<Segment<T>>, IList<T>

https://docs.microsoft.com/en-us/dotnet/api/system.arraysegment-1
https://docs.microsoft.com/en-us/dotnet/api/microsoft.extensions.primitives.stringsegment

Add ToArray without parameters

Collections/Lists/IListExtensions.cs

Line 50 in a1c44f1

public static T[] ToArray<T>(this IList<T> list, Func<T, bool> predicate)

proposed code

public TLink[] ToArray(this IList<T> list)
{
    var array = new TLink[list.Count];
    list.CopyTo(array, 0);
    return array;
}

and use it here:

https://github.com/linksplatform/Data.Doublets/blob/93ad6046c5c2273623c7ff2b2264d59326b0e827/Link.cs#L73

and delete from here:

https://github.com/linksplatform/Data.Doublets/blob/93ad6046c5c2273623c7ff2b2264d59326b0e827/Link.cs#L77

Try to use Vector<ulong> for BitString as an optimization

Collections/Platform.Collections/BitString.cs

Line 19 in e31048d

public class BitString

https://docs.microsoft.com/en-us/dotnet/api/system.numerics.vector-1?view=netstandard-2.1

Think about mechanism of chaining multiple operations with BitString

If we apply multiple operation in single place in the memory this can be faster than applying just one operation for the whole BitString due to memory bandwidth limitations.

Translate the code from C# to C++

10000 RUB reward can be claimed by anyone who completes this task.

Check for memory leaks

Collections/Arrays/ArrayPool.cs

Line 7 in e4fbd36

/// TODO: Check for memory leaks

Create ArrayExtensions with Clone method

Proposed code:

public static T[] Clone<T>(this T[] array)
{
    var copy = new T[array.Length];
    Array.Copy(array, 0, copy, 0, array.Length);
    return copy;
}

Use IListExtensions.EqualTo instead

Collections/Segments/Segment.cs

Line 46 in a76be8a

for (var i = 0; i < Length; i++)

Add TrimSingle to remove just one symbol for each end withing specified characters range

Collections/StringExtensions.cs

Line 6 in 539bb67

public static class StringExtensions

For example s.TrimSingle('"'). Should be used to fix linksplatform/IO#10.

Check if MethodImpl attribute really make a difference

Collections/Concurrent/ConcurrentQueueExtensions.cs

Line 9 in a76be8a

[MethodImpl(MethodImplOptions.AggressiveInlining)]

what the

syntax errors

Use Range.NonNegativeInt64 property here

Collections/Platform.Collections/Arrays/ArrayPool[T].cs

Line 54 in 84cedfa

Ensure.Always.ArgumentInRange(size, (0L, long.MaxValue));

Based on linksplatform/Ranges#24

Use week references here to avoid memory leaks

Collections/csharp/Platform.Collections/Arrays/ArrayPool[T].cs

Line 25 in 036ccac

private readonly Dictionary<long, Stack<T[]>> _pool = new Dictionary<long, Stack<T[]>>(ArrayPool.DefaultSizesAmount);

bug in method

the case when the old array size is the same as the new size is not handled.

Make all readonly fields public

Collections/csharp/Platform.Collections/Lists/ListFiller.cs

Line 8 in 29dcff5

protected readonly List<TElement> _list;

Rename the method into SkipAndAdd

Collections/csharp/Platform.Collections/Lists/IListExtensions.cs

Line 169 in ed094f1

public static void AddSkipFirst<T>(this IList<T> list, IList<T> elements, int skip)

Should be a version of ArrayFiller without return constant

Collections/Arrays/ArrayFiller.cs

Line 6 in a76be8a

public class ArrayFiller<TElement, TReturnConstant>

Rethink and test Length property implementation

Collections/BitString.cs

Line 128 in 56c3b06

public long Length

Add PeekOrDefault method to StackExtensions and IStackExtensions

Add IsEmpty property

Collections/Stacks/IStack.cs

Line 3 in a1c44f1

public interface IStack<TElement>

Peek and Pop cannot work correctly when IsEmpty is equal to true.

Move logic from Platform.Data.Doublets

Based on linksplatform/Data.Doublets#91

Think about ability to manual refreshing of borders (minimum and maximum non-zero words)

Collections/BitString.cs

Line 21 in 56c3b06

public class BitString

That means let the user decide when to refresh borders (no automatic refreshing).

Add a benchmark to compare performance of CharSegment and Segment

Rename class to filename

Collections/ISetExtensions.cs

Line 5 in a76be8a

public static class SetExtensions

Compare what is faster to store BitsSetIn16Bits or to calculate it on demand

Collections/BitString.cs

Line 24 in a3b1fae

private static readonly byte[][] BitSetsIn16Bits;

Add PopOrDefault method to IStackExtensions

Use netcoreapp3.1 instead of netcoreapp3.0

Collections/csharp/Platform.Collections.Tests/Platform.Collections.Tests.csproj

Line 4 in 5b1923f

<TargetFrameworks>net472;netcoreapp2.1;netcoreapp3.0</TargetFrameworks>