wan24-Core 2.29.0

wan24-Core

This core library contains some .NET extensions and boiler plate avoiding helpers. It's designed as core library for a long running process and optimized for that purpose. The code tries to cache agressive whereever it's possible. Types are designed to be thread-safe, if they're likely to be accessed reading/writing from multiple threads (if it's their nature). However, if it's not sure that a types nature is to manage multithreaded access, performance goes before thread safety.

It started as a slender extension method collection and growed up to a larger core utility until today. It's possible that some specialized parts will be splitted into separate libraries in the future, if the core library is getting too big - but the namespace will stay the same.

Some key features:

• Bootstrapping
• Disposable base class for disposable types, which supports asynchronous disposing
  • Dispose attribute for fields/properties which should be disposed automatic when disposing
• CancellationOnDispose cancels a cancellation token when an object is being disposed (or another given cancellation token ws canceled)
• Cancellations combines multiple cancellation tokens into one
• Type helper (type loading)
• Secure byte and char array, which clears its contents when disposing
• Pool rented array as disposable object (which optionally clears its contents when disposing; for byte/char arrays just like the Secure*Array)
• Asynchronous API helper
  • Asynchronous fluent API helper
• Byte array extensions
  • Endian conversion
  • Bit-converter (endian-safe)
  • UTF-8/16/32 (little endian) string decoding
  • Clearing
  • Base64 encoding/decoding
  • Fast XOR, AND and OR using intrinsics
  • Slow compare
• Dictionary extensions
  • Merge with string key prefix
  • Merge a list with the index as key (and an optional key prefix)
• Char array extensions
  • Clearing
  • Base64 decoding
• Array helper extensions
  • Offset/length validation
  • Item index finding
  • Contained items finding
  • AsReadOnly extensions
  • Generic cloning extension
• Array pool extensions
  • Renting a cleared array
• Enumerable extensions
  • Combine enumerables
  • Chunk enumerables
• Enumeration classes
• Reflection extensions
  • Automatic parameter extension when invoking a method (with DI support)
  • Synchronous/asynchronous method invokation
  • Automatic constructor invokation using a given parameter set (with DI support)
  • Nullability detection
  • Get property getter/setter delegates
  • Get cached field/property/method info
• Cache for field/property/method info and custom attributes
• Delegate extensions
  • Delegate list invokation (with or without return values, with DI support)
  • Asynchronous delegate list invokation (with or without return values, with DI support)
• Task extensions
  • Result getting of a generic task
  • Asynchronous task list awaiting
  • Shortcuts for await configurations
  • Shortcuts for starting a function as long running task
  • Shortcuts for starting a function as task with fair execution by the scheduler
• DI helper
  • Service provider adoption
  • DI object factory delegates
  • Asynchronous DI object factory delegates
• Enumeration extensions
  • Get enumeration value display string from DisplayTextAttribute or using ToString (fallback)
  • Determine if all, any or which flags are contained in an enumeration value
  • Remove flags of a mixed enumeration value
  • Get only flags of a mixed enumeration value
  • Value validation
• Number extensions
  • Determine if a type is a number
  • Determine if a number type is unsigned
  • Bit-converter (endian-safe)
  • Determine if a number (or any IComparable) is within a range
• Numeric bitwise extensions
• Collection extensions
  • Add a range of items
• JSON helper
  • Exchangeable JSON encoder/decoder delegates (using System.Text.Json per default)
• JSON extensions
  • Encode an object
  • Decode from a type
  • Decode a string
• Object extensions
  • Type conversion
  • Determine if a value is within a list of values
• String extensions
  • Get UTF-8/16/32 bytes (little endian)
  • Parsing
  • String from/to bytes/bits extensions
  • Determine if string contains only ACSII characters
  • Base64 decoding
• Generic helper
  • Determine if two generic values are equal
  • Determine if a value is null
  • Determine if a value is default
  • Determine if a value is null or default
• DateTime extensions
  • Determine if a time is within a range
  • Determine if a time matches a reference time plus/minus an offset
  • Apply an offset to a time base on a reference time
• TimeSpanHelper
  • Update a timeout
• Queue worker (for actions and/or items)
• Parallel queue worker (for actions and/or items)
• ParallelAsync implementation
  • ForEachAsync with an asynchronous or synchronous input source
  • FilterAsync with an asynchronous or synchronous input source and item filter
  • Filter for synchronous parallel filtering
• Base class for a hosted worker, which implements the IHostedService interface (timed or permanent running)
• EventThrottle for throttling event handler calls
• ProcessThrottle for throttling a processing channel
• OrderedDictionary<tKey, tValue> is used for working with indexed key/value pairs
• Timeout will count down and raise an event, if not reset before reaching the timeout
• ILogger support
  • Logging as global logging helper
  • LoggerBase and DisposableLoggerBase as base classes for a custom logger
  • Logger as ILogger for writing to Logging
  • FileLogger as ILogger for writing to a file
  • ConsoleLogger writes to STDERR
  • DebugLogger writes to the debug console
  • EmailLogger sends an email
• IChangeToken support using ChangeCallback
• Hierarchic configuration using OverrideableConfig
• Cancellation token awaiter
• ObjectPool for pooling objects (DisposableObjectPool for disposable types), and BlockingObjectPool for a strict pool capacity limit
• (Blocking)StreamPool, PooledMemoryStream, PooledTempFileStream and PooledTempStream (hosts written data in memory first)
• ResetEvent for (a)synchronous event waiting
• LazyValue<T>, DisposableLazyValue<T>, AsyncDisposableLazyValue<T> and TimeoutValue<T> for lazy and timeout value serving
• ObjectLockManager<T> for asynchronous and awaitable object locking
• Bitmap for working with bits
• DisposableWrapper<T> for wrapping any (not disposable?) object with the IDisposable and IAsyncDisposable interface using custom dispose actions during runtime
• DisposableAdapter for adopting the IDisposableObject interface from a type which can't extend the DisposableBase type
• Generic object extenions for validating method arguments
• CLI arguments interpreter
• Runtime configuration from CLI arguments
• Fast byte to string and string to byte encoding/decoding (using an URI friendly charset, faster and smaller results than base64 encoding; charset is customizable; encoded data integrity can be validated without decoding; including extensions for numeric type encoding/decoding)
• Collecting periodical statistical values
• Streams
  • StreamBase as base class which implements some disposing logic
  • WrapperStream wraps a base stream and provides LeaveOpen
  • PartialStream wraps a part of a base stream (read-only)
  • NonSeekablePartialStream wraps a part of a non-seekable base stream (read-only)
  • LengthLimitedStream ensures a maximum stream length (only writing)
  • MemoryPoolStream uses an ArrayPool<byte> for storing written data
  • ThrottledStream throttles reading/writing troughput
  • TimeoutStream can timeout async reading/writing methods
  • BlockingBufferStream for writing to / reading from a buffer blocked
  • HubStream for forwarding writing operations to multiple target streams
  • LimitedStream limits reading/writing/seeking capabilities of a stream
  • ZeroStream reads zero bytes and writes to nowhere
  • CountingStream counts red/written bytes
  • PerformanceStream counts red/written bytes and I/O time
  • PauseableStream is a stream which can temporary pause reading/writing
  • EnumerableStream streams an enumerable/enumerator
  • CombinedStream combines multiple streams into one stream (read-only)
  • SynchronizedStream synchronizes IO and seeking
  • RandomStream reads random bytes into the given buffers
  • ChunkedStream for reading/writing stream chunks
  • ExchangeableStream wraps an exchangeable base stream
  • BackupStream writes all red data to another stream
  • ProcessStream uses STDIN/OUT of a process
  • AcidStream for ACID stream IO
  • BackgroundStream for fast writing in background
  • FlushStream for writing to a buffer until a call to flush
  • CutStream for cutting a stream from its position
  • ExactStream for reading exactly the number of given bytes, if possible
  • BackgroundProcessingStream for a background stream buffer filling processor
  • ForceAsyncStream forces all IO operations to be performed asynchronous
  • ForceSyncStream forces all IO operations to be performed synchronous
  • CopyStream does copy a stream to a target stream in a background task
  • DataEventStream blocks reading of a wrapped stream until more input data is available
• Named mutex helper
  • GlobalLock for a synchronous context
  • GlobalLockAsync for an asynchronous context
• Retry helper which supports timeout, delay and cancellation
• Asynchronous event
• Stream extensions
  • Get the number of remaining bytes until the streams end
  • Copy a part of a stream to another stream
  • Generic seek
  • Generic read/write byte
  • Write N zero bytes
  • Write N random bytes
  • Create stream chunks
• Checksum implementation in ChecksumExtensions and ChecksumTransform
• Thread synchronization helper (synchronous / asynchronous)
  • SemaphoreSync uses a SemaphoreSlim for thread synchronization
  • SemaphoreSyncContext is a disposable thread synchronization context
• Networking helper
  • Browsing ethernet adapters filtered
  • Classify LAN, WAN and loopback IP addresses and ethernet adapters
  • IpSubNet implementation for managing IPv4/6 sub-nets (CIDR)
  • Find the next free TCP port for an IP address
• Centralized error handling
• Delayed tasks
• Progress
• Sensitive data handling
• Object storage
• Localization support
• Email sending abstractions
• Commonly used regular expressions and global named expression collection
• Transactions
• TCP, UDP, WebSocket and http(s) port knocking
• Observable collections
  • ChangeTokenCollection
  • ChangeTokenDictionary
  • ConcurrentChangeTokenDictionary
• App JSON configuration
• Customizable object serialization helper
• In-memory cache

How to get it

This library is available as NuGet package "wan24-Core".

Bootstrapping

The Bootstrapper.Async method calls all static methods having the BootstrapperAttribute. In order to be able to find the methods, it's required to add the BootstrapperAttribute to the assembly.

You may also ad the BootstrapperAttribute to a type and/or the bootstrapper method, in case the assembly contains multiple of them. In the assembly attribute you need to set ScanClasses and/or ScanMethods to true in order to perform a deep scanning during bootstrapping for performance reasons.

The bootstrapper methods may consume parameters which are available from the DI helper. The method may be synchronous or asynchronous. The method can't be defined in a generic class, and it can't be generic itself.

[assembly:Bootstrapper(typeof(YourBootstrapper),nameof(YourBootstrapper.BootstrapperMethod))]

public static class YourBootstrapper
{
    public static async Task BootstrapperMethod()
    {
        // Perform your bootstrapping here
    }
}

// Call the bootstrapper somewhere in your apps initialization code
await Bootstrap.Async();

The BootstrapperAttribute can be initialized with a numeric priority. The bootstrapper will order the found bootstrapping methods by priority, where the one with the highest number will be executed first (assembly and type priorities count, too). At last there's a assembly location, type and method name sorting. Bootstrapper methods will be executed sequential.

If you give a type and a method name to the assembly BootstrapperAttribute, you won't need to add the attribute to the type and the method.

During bootstrapping, the cancellation token which was given to the Bootstrap.Async method, can be injected to a bootstrappers method parameters.

After that bootstrapping was done, the Bootstrap.AsyncBootstrapper will be called. At last the Bootstrap.OnBootstrap event will be raised.

During bootstrapping the Bootstrap.IsBooting property is true. After bootstrapping the Bootstrap.DidBoot property is true.

The bootstrapper will load all referenced assemblies. If you load an assembly later, it'll be bootstrapped automatic and added to the TypeHelper singleton instance.

Type helper

If you use the TypeHelper.AddTypes method, the unknown assemblies of the added types will be added as searchable assemblies automatic.

You may attach to the TypeHelper.OnLoadType event for handling requests more dynamic.

The TypeHelper.GetType method will try Type.GetType first and fall back to the helper, if no type was found.

DI helper

In order to make DI (dependency injection) working, you need to

• set a DiHelper.ServiceProvider and/or
• add DiHelper.(Async)ObjectFactories

The DiHelper.GetDiObjectAsync method will try to resolve the request synchronous, first. But the DiHelper.GetDiObject won't try asynchronous object factories. Created objects will be cached - to avoid that, add the type to the not cached types by using DiHelper.AddNotCachedType.

The Instance property returns a generic service provider, which uses the ServiceProvider and the registered object factories. It also implements the IAsyncServiceProvider interface, which extends IServiceProvider.

Object factories are as generic as possible. They'll receive the requested object type as an argument and may return different instanced based on the particular request. The registration of an object factory requires to define the object type, which the factory method served, as key. This type can be any type, which can be assigned to a requested type, or a generic type definition to match all generic variants. The hierarchic factory selection works like this:

1. Requested type is the registered factory type
2. Requested type is assignable from the registered factory type
3. Requested type is a generic type, and its generic type definition is assignable from the registered factory type

The ScopedDiHelper uses its own object factory collections and service provider and falls back to the DiHelper. Created objects will be disposed, if possible and the ScopedDiHelper is being disposed, and only in case the object wasn't created using the base DiHelper. ScopedDiHelper will also dispose the defined ServiceProvider, if possible.

CAUTION: DiHelper and ScopedDiHelper will cache created objects. For this reason you should never dispose a returned object in your code!

To replace the default DI of a .NET app:

appBuilder.UseServiceProviderFactory(context => new DiHelper.DiServiceProviderFactory(new()
{
    ValidateOnBuild = false,
    ValidateScopes = context.HostingEnvironment.IsDevelopment()
}));

Then you can use the DI helper by injection:

public async Task YourMethod(IAsyncServiceProvider diHelper)
{
    YourService service = await diHelper.GetServiceAsync(typeof(YourService));
    ...
}

Logging

For a global logging, use Logging and set a Logging.Logger - for example the FileLogger:

Logging.Logger = await FileLogger.CreateAsync("/path/to/file.log");
Logging.WriteDebug("Hello world!");

If you need an ILogger instance, you can use Logger:

anyObject.Logger = new Logger();

The Logger will use Logging and allows to define a minimum log level.

WARNING: Never set a Logger instance as Logging.Logger! This will cause an endless loop.

Mixed enumeration value

A mixed enumeration contains X bits enumeration values, and Y bits flags:

[Flags]
public enum MixedEnum : int
{
    None = 0,
    Value1 = 1,
    Value2 = 2,
    Value3 = 3,
    ...
    Flag1 = 1 << 8,
    Flag2 = 1 << 9,
    FLAGS = Flag1 | Flag2 // Required to identify flags
}

The FLAGS value helps these extension methods to handle flag values:

MixedEnum value = MixedEnum.Value1 | MixedEnum.Flag1,
    valueOnly = value.RemoveFlags(),// == MixedEnum.Value1
    flagsOnly = value.OnlyFlags();// == MixedEnum.Flag1

Unsafe code

The library uses unsafe code. If you don't want/need that, you can compile the library with the NO_UNSAFE compiler constant to disable any unsafe operation. Remember to unset the unsafe compiler option, too!

Disposable base class

The DisposableBase implements the IDisposable and IAsyncDisposable interfaces. It provides some helpers and events, and also the DisposeAttribute, which can be applied to fields and properties which you wish to dispose automatic when disposing.

When your type derives from the DisposableBase, you'll need to implement the abstract Dispose method:

protected override Dispose(bool disposing)
{
    // Your dispose logic here
}

There are measures to avoid that this method is being called twice.

To implement custom asynchronous disposing:

protected override async Task DisposeCore()
{
    // Your dispose logic here
}

In order to make the DisposeAttribute working, you have to call the protected method DisposeAttributes or DisposeAttributesAsync.

The IsDisposing property value will be true as soon as the disposing process started, and it will never become false again. The IsDisposed property value will be true as soon as the disposing process did finish.

TIP: Use the DisposableBase<T> base type, if you plan to use the DisposeAttribute! This base class will cache the fields/properties once on initialization to get rid of the reflection overhead which DisposableBase requires for this feature.

NOTE: The DisposeAttribute can be applied to byte[] and char[], too, which will simply call the Clear extension method on disposing. Another IEnumerable will be enumerated for disposable items (recursing!).

Queue worker

using QueueWorker worker = new();
await worker.EnqueueAsync((ct) =>
{
    // Do any background action here
});

The QueueWorker class can be extended as you need it.

The ParallelQueueWorker requires a number of threads in the constructor, which defines the degree of parallelism, in which enqueued tasks will be processed.

Queue item worker

using QueueItemWorker<ItemType> worker = new();
await worker.EnqueueAsync(new ItemType());

The QueueItemWorker<T> class can be extended as you need it.

The ParallelItemQueueWorker<T> requires a number of threads in the constructor, which defines the degree of parallelism, in which enqueued items will be processed.

ParallelAsync

Using the .NET parallel implementation it's not possible to invoke asynchronous item handlers. For this you can use the ParallelAsync.ForEachAsync method, which uses a parallel item queue worker in the background for asynchronous processing.

Hosted worker

public class YourHostedWorker : HostedWorkerBase
{
    public YourHostedWorker() : base() { }

    protected override async Task WorkerAsync()
    {
        // Perform the service actions here
    }
}

The hosted worker implements the IHostedService interface and can be extended as you need it.

Timed hosted worker

public class YourHostedWorker : TimedHostedWorkerBase
{
    public YourHostedWorker() : base(interval: 500) { }

    protected override async Task WorkerAsync()
    {
        // Perform the service actions here
    }
}

This example uses a 500ms timer. Based on the defined timer type, the interval will be processed in different ways:

• Default: Next worker run is now plus the interval (used by default)
• Exact: Next worker run is now plus the interval minus the processing duration (used, if the start time of the processing is important)
• ExactCatchingUp: As Exact, but catching up missing processing runs without delay, if a worker run duration exceeds the interval (used, if the number of worker runs is important)

Using the SetTimerAsync method you can change the timer settings at any time. If you give the nextRun parameter, you may set a fixed next run time (which won't effect the given interval, but just force the service to run at a specific time for the next time).

NOTE: The nextRun parameter will also force the service to (re)start!

By setting the property RunOnce to true, the service will stop after running the worker once. In combination with the SetTimerAsync parameter nextRun you can execute the worker at a specific time once.

The hosted worker implements the IHostedService interface and can be extended as you need it.

EventThrottle

public class YourType : DisposableBase
{
    protected readonly YourEventThrottle EventThrottle;

    public YourType() : base() => EventThrottle = new(this);

    // This method will raise the OnEvent
    public void AnyMethod()
    {
        RaiseOnEventThrottled();
    }

    protected override Dispose(bool disposing) => EventThrottle.Dispose();

    // Delegate for OnEvent
    public delegate void YourTypeEvent_Delegate();
    // Event to throttle
    public event YourTypeEvent_Delegate? OnEvent;
    // Raise the OnEvent using the event throttle
    protected void RaiseOnEventThrottled() => EventThrottle.Raise();
    // Finally let the event handlers process the event
    protected void RaiseOnEvent() => OnEvent?.Invoke();

    // Event throttle implementation
    public class YourEventThrottle : EventThrottle
    {
        // Throttle the event handling down to max. one handling per 300ms
        public YourEventThrottle(YourType instance) : base(timeout: 300) => Instance = instance;

        public YourType Instance { get; }

        protected override HandleEvent(DateTime raised, int raisedCount)
        {
            Instance.RaiseOnEvent();
        }
    }
}

If AnyMethod is being called, the event will be forwarded to the event throttle, which decides to throttle or raise the event. If AnyMethod was called three times within 300ms, the first call will be executed in realtime, while the 2nd and the 3rd call will be sqashed and executed once 300ms after the 1st call was processed.

This example assumes you're working with a real event - but you may throttle any event (which may not be a real event) using throttling logic.

ProcessThrottle

public class YourProcessThrottle : ProcessThrottle
{
    // Throttle to processing one object per second
    public YourProcessThrottle() : base(limit: 1, timeout: 1000) { }

    // Processing API using a timeout
    public async Task<int> ProcessAsync(Memory<bool> items, TimeSpan timeout)
        => await ProcessAsync(items.Length, (count) => 
        {
            await Task.Yield();
            Span<bool> toProcess = items.Span[..count];
            items = items[count..];
            // Process toProcess
        }, timeout);

    // Processing API using a cancellation token
    public async Task<int> ProcessAsync(Memory<bool> items, CancellationToken token = default)
        => await ProcessAsync(items.Length, (count) => 
        {
            await Task.Yield();
            Span<bool> toProcess = items.Span[..count];
            items = items[count..];
            // Process toProcess
        }, token);
}

The example will throttle the processing to a maximum of one object per second. Multiple threads may call ProcessAsync concurrent - processing will be organized thread-safe.

The return value of ProcessAsync is the number of objects processed until timeout or canceled.

The processing delegate shouldn't care about the timeout or if canceled and just process the given number of objects.

NOTE: A usage gap will slide the throttling timer. Example:

The timeout was set to 3 objects per 100ms. Now processing goes like this:

• First processed object on 0ms will activate the throttling timeout
• Next processed object on 10ms will increase the object throttling counter
• Next processed object on 110ms will reset the throttling timeout and counter (the usage gap of 100ms does exceed the timeout)
• Next 2 processed objects on 120ms will activate the throttle
• Next object will have to wait until the throttle was released
• The throttle will be released on 210ms, which allows the last object to be processed now

In short words: The throttle timer will not reset in an fixed interval, but the interval starts when processing items.

Change token

Implement by extending (Disposable)ChangeToken (implements IChangeToken and INotifyPropertyChanged):

public class YourObservableType : ChangeToken
{
    private string _Value = string.Empty;

    public YourObservableType() : base() { }

    public string Value
    {
        get => _Value;
        set => SetNewPropertyValue(ref _Value, value, nameof(Value));
    }
}

The HasChanged setter MAY be used. You can also set the _HasChanged field and call InvokeCallbacks any time later. If the default HasChanged setter was used with true, RaisePropertyChanged will be called without a property name. Instead of the HasChanged setter you can also call SetNewPropertyValue from a property setter.

The RaisePropertyChanged and InvokeCallbacks method SHOULD be called each time a property changed (this will be done when calling SetNewPropertyValue also).

By extending (Disposable)ChangeToken<T> your final type will also implement IObservable<T>.

You may want to use the (Concurrent)ChangeTokenCollection/Dictionary<T> for observing an object list or a (concurrent) key/value dictionary. They implement

• IObservable<T>
• INotifyCollectionChanged
• INotifyPropertyChanged

Observed are all

• IChangeProperty
• IObservable<T>
• INotifyPropertyChanged

item events. You can also use a type which doesn't implement any of these interfaces - then only the collection itself (item adding/removing) is observed.

An instance is pre-configured for use with a ChangeToken. For other objects (which implement IChangeToken and INotifyPropertyChanged) you can modify the

• IgnoreUnnamedPropertyNotifications (default is true)
• InvokeCallbacksOnPropertyChange (default is false)

settings.

The ObserveCollection setting defines, if you'd also like to observe item addings/removals (the collection itself). The default is true. If you set the ObserveItems property to false during the collection object initialization, only the item additions/removals will be observed. Of course you can set both properties to false - in this case the collection won't observe anything.

Hierarchic configuration

Assume this configuration hierarchy:

Level	Description
1	Default values
2	User values (can override default values)
3	Administrator values (can override default/user values)

In code:

public sealed class Config : OverrideableConfig<Config>
{
    public Config() : base()
    {
        SubConfig = new(this, new(this));// User values
        InitProperties();
    }

    private Config(Config parent, Config? sub = null) : base(parent)
    {
        if(sub != null)
        {
            SubConfig = sub;
            sub.ParentConfig = this;
            sub.SubConfig = new(sub);// Administrator values
        }
        InitProperties();
    }

    // A configuration value
    public ConfigOption<string, Config> AnyValue { get; private set; } = null!;

    private void InitProperties()
    {
        AnyValue = ParentConfig == null 
            // The master option has a default value
            ? new(this, nameof(AnyValue), canBeOverridden: true, "default")
            // No default value for a sub-option
            : new(this, nameof(AnyValue));
    }
}

Config config = new(),
    user = config.SubConfig,
    admin = user.SubConfig;

CAUTION: There's no endless-recursion protection for the ParentConfig or the SubConfig properties!

Now users are able to override default values, and administrators are able to override default and/or user values:

// Still the default value
Assert.AreEqual("default", config.AnyValue.FinalValue);

// User overrides the default value
user.AnyValue.Value = "user";
Assert.AreEqual("default", config.AnyValue.Value);
Assert.AreEqual("user", config.AnyValue.FinalValue);

// Administrator overrides the user value
admin.AnyValue.Value = "admin";
Assert.AreEqual("admin", config.AnyValue.FinalValue);

// User can't override the administrator value (but still store his own value 
// in case the administrator would unset his value)
user.AnyValue.Value = "test";
Assert.AreEqual("admin", config.AnyValue.FinalValue);
Assert.AreEqual("test", user.AnyValue.Value);

NOTE: Setting an option value is thread-safe.

It's also possible to flip the hierarchy:

Level	Description
1	Default values
2	Administrator values (can define user visible and optional not overrideable values)
3	User values (can override overrideable values)

Using this hierarchy an administrator could also allow or deny overriding values at any time, for example.

The hierarchy depth isn't limited.

Object locking

The ObjectLockManager<T> helps locking any object during an asynchronous operation:

ObjectLock ol = await ObjectLockManager<AnyType>.Shared.LockAsync(anyObjectKey);
// A 2nd call to ObjectLockManager<AnyType>.Shared.LockAsync would block until the lock was released
await ol.RunTaskAsync(Task.Run(async () => 
{
    // Perform the asynchronous operation here
}));
// ol is disposed already, 'cause the asynchronous operation source task was awaited
// The next ObjectLockManager<AnyType>.Shared.LockAsync call will be processed now, if any
await ol.Task;// To throw any exception during performing the asynchronous operation

If AnyType implements the IObjectKey interface, it can be given to the ObjectLockManager<T> methods as object argument.

NOTE: ObjectLock will dispose itself as soon as RunTaskAsync has been called, and the given task was completed.

CLI arguments interpreter

There a just a few rules:

1. A flag starts with a single dash
2. A key for a value (list) starts with a double dash
3. Keys/values can be quoted using single or double quotes
4. Escape character is the backslash (only applicable in quoted values)
5. A quoted value must be escaped for JSON decoding, a backslash must be double escaped
6. Double quotes in a quoted value must be escaped always

Example:

"-flag" --key 'value1' value2 --key -value3 '--key2' "value"

For appending the value -value3 to the value list of key, the value needs to be added with another --key key identifier, 'cause it starts with a dash and could be misinterpreted as a flag (which would result in a parser error).

A CLI app called with these arguments could interpret them easy using the CliArguments class:

CliArguments cliArgs = new(args);
Assert.IsTrue(cliArgs["flag"]);
Assert.AreEqual(3, cliArgs.All("key").Count);
Assert.AreEqual("value", cliArgs.Single("key2"));

A -- (double dash) may be interpreted as an empty key name or a flag with the name -, based on if a value, which doesn't start with a dash, is following. Examples:

• --: - flag
• -- -: - flag (-- and - are both interpreted as double - flag (double flags will be combined))
• -- value: Empty key with the value value
• -- -key: - and key flags

Keyless arguments will be stored in the KeyLessArguments list - example:

CliArguments ca = CliArguments.Parse("value1 -flag value2 --key value3");
Assert.AreEqual(2, ca.KeyLessArguments.Count);
Assert.AreEqual("value1", ca.KeyLessArguments[0]);
Assert.AreEqual("value2", ca.KeyLessArguments[1]);
Assert.IsTrue(ca["flag"]);
Assert.IsTrue(ca["key", true]);

Fast byte ↔ string encoding/decoding

base64 is supported everywhere, but it's (relative) slow and produces too much overhead, and uses also URI unfriendly characters. In addition it's also not easy to validate base64, or to determine the encoded/decoded value length.

To fix all of these problems, the ByteEncoding class implements a fast encoding, which uses only characters 0-9, a-z, A-Z, dash and underscore and produces less overhead than base64. The encoded/decoded value length can be calculated in advance, and it's fast and easy to detect errors in the encoded data without having to decode it, first.

// In case you want to use a prepared output buffer
int encodedLen = anyByteArray.GetEncodedLength();

// Encoding
char[] encoded = anyByteArray.Encode();

// In case you want to use a prepared output buffer
int decodedLen = encoded.GetDecodedlength();

// Decoding
byte[] decoded = encoded.Decode();

Using extensions numeric values can be en-/decoded on the fly, too. The special EncodeNumberCompact extension methods determine the smallest value matching numeric type before encoding (use DecodeCompactNumber with the original numeric type as generic argument for decoding).

NOTE: Encoding an empty array results in an empty string. Encoding 0 results in an empty string, too. Nothing encodes to nothing and decodes to nothing, too.

If required, the used encoding character map can be customized. You may use any 64 ASCII characters (0..127) long map with unique items.

String parser

Using the Parse extension method for a string, you can parse placeholders into a string and modify the output using (customizable) parser functions:

Dictionary<string, string> data = new()
{
    {"name", "value"}
};
Assert.AreEqual("value", "%{name}".Parse(data));

You may setup the default parser data in StringExtensions.ParserEnvironment. The given parser data will override defaults.

You can execute as many parser functions on the output as required, separated using ::

%{input:func1:func2(param1,param2,...):func3():...}

The first optional segment is always a parser data variable name (if not used, the sequence starts with a : to indicate a function call). A function may or may not have parameters. The result of a function will be provided for the next function. Available functions:

Function	Syntax	Usage
sub	%{input:sub([offset/length](,[length]))}	extracts a sub-string
left	%{input:left([length])}	takes X characters from the left
right	%{input:right([length])}	takes X characters from the right
trim	%{input:trim}	removes white-spaces from the value
discard	%{input:discard}	no parameters, discards the current output
escape_html	%{input:escape_html}	escapes the value for use within HTML
escape_json	%{input:escape_json}	escapes the value for use within double quotes (double quotes will be trimmed from the JSON result!)
escape_uri	%{input:escape_uri}	escapes the value for use within an URI
set	%{input:set([name])} sets the current output as parser variable with the given name
var	%{:var([name])} gets a parser data variable value
item	%{:item([index],[item/name](,[item](,...)))} gets an item from a list (if using a variable name, its value will be splitted using pipe)
prepend	%{input:prepend([string])}	prepends a string
append	%{input:append([string])}	appends a string
insert	%{input:insert([index],[string])}	inserts a string at an index
remove	%{input:remove([offset/length](,[length]))}	removes a part (from the left)
concat	%{:concat([string],[string](,[string](...))}	concatenates strings
join	%{:join([separator],[string],[string](,...))}	joins strings
math	%{:math([operator],[value1],[value2](,...))}	performs math
rx	%{:rx([group_index]],[name/pattern])}	exchanges the parser regular expression and content group index for the next parser operations (the next round)
format	%{input:format([format])}	to format a numeric value
str_format	%{input:str_format(([value1](,...))} to format the string value
insert_item	%{input:insert_item([index],[items_name])}	to insert an item (items will be splitted by pipe)
remove_item	%{input:remove_item([index])}	to remove an item (items will be splitted by pipe)
sort	%{input:sort((desc))}	to sort items
foreach	%{input:foreach([name])}	to parse a parser data value for each item (will be stored in _item)
if	%{input:if([name](,[name]))}	to parse a parser data value, if the value is 1 (else parse the second given parser data value)
split	%{input:split(prefix)}	to split items by pipe and set them as parser data using the prefix and appending the zero based item index
range	%{:range([start],[count])}	to create a numeric range
dummy	%{:dummy(...)}	does nothing (may be used as comment)

Available math operators:

Operator	Function
+	Summarize
-	Substract
*	Multiply
/	Divide
%	Modulo
a	Average
i	Minimum
x	Maximum
r	Round (2nd value is the number of decimals)
f	Floor
c	Ceiling
p	Y power of X (double conversion will be applied)
=	Equality (0 is not equal, 1 if equal)
<	Lower than (0 is not lower, 1 if lower)
>	Greater than (0 is not greater, 1 if greater)
s	Change the sign

Numbers are written in invariant culture float style. decimal will be used as number format.

To create a custom parser function:

StringExtensions["func_name"] = (context) => 
{
    // Work with the StringParserContext and return the value to use or set context.Error for error handling
    return context.Value;
};

Example:

StringExtensions["upper"] = (context) => context.Value.ToUpper();

Dictionary<string, string> data = new()
{
    {"name", "value"}
};
Assert.AreEqual("VALUE", "%{name:upper}".Parse(data));

CAUTION: A placeholder must produce the same result, if it occurs repeated! A repeated placeholder won't be parsed more than once, but being replaced with the result of the first parsed placeholder.

Example:

Dictionary<string, string> data = new()
{
    {"name", "value"}
};
string tmpl = "%{name}%{name:len:set(name):discard}%{name}";
Assert.AreEqual("valuevalue", tmpl.Parse(data));

From the logic value5 would be expected. To get value5, finally, you'll have to modify the template:

%{name}%{name:len:set(name):discard}%{name:dummy}

TIP: Almost all function parameters may be parser data variable names, too, if they have a $ prefix. To support that, use the TryGetData method of the StringParserContext, if a parameter value starts with $.

TIP: To ensure having all required parameters, use the EnsureValidParameterCount of the StringParserContext. The method allows you to define a number of allowed parameter counts (including zero) and produces a common error message, if the function call syntax is wrong.

TIP: A custom parser function may change the parser regular expression and content group by changing Rx and RxGroup.

The string parser works recursive. To avoid an endless recursion, the default parsing round count limit is 3. The current parsing round is accessable trough the parser data _round. If a parser function parses a template, the called parser will work in the current parsing round context and respect the limit, too. Youmay set another default limit in StringExtensions.ParserMaxRounds.

The default behavior for errors is to throw an exception. If error throwing was disabled, in case of an error a placeholder will stay in clear text, and a function will return the unaltered value.

You may modify the placeholder declaration by setting another regular expression to StringExtensions.RxParser. Group $1 must contain the whole placeholder, while group $2 is required to contain the inner placeholder contents (like variable name, function calls, parameters, etc.). There's no way to customize the inner placeholder content syntax at present. You may also give a custom regular expression to the Parse extension method, if you want an isolated parsing. You can modify the inner content group index by setting StringExtensions.RxParserGroup or giving rxGroup to the Parse methods.

CAUTION: Be careful with customized parser functions: A mistake could let a manipulated string harm your computer!

Retry helper

RetryInfo<object> result = await RetryHelper.TryActionAsync(
    async (currentTry, cancellation) => 
    {
        // Perform any critical action which may throw or timeout and return a value (or not)
    },
    maxNumberOfTries: 3,
    timeout: TimeSpan.FromSeconds(30),
    delay: TimeSpan.FromSeconds(3)
    );

// This will throw an exception, if failed, or return the action delegate return value, if succeed
object returnValue = result.ThrowIfFailed();

TryAction* will try to execute an action for a maximum of N times, optional having a total timeout, and optional performing a delay after a failed try. The given action delegate may also return a value, which you can then find in the RetryInfo<T>.Result property, if Succeed is true.

The RetryInfo<T> object contains some runtime informations:

• Start, done time and total runtime
• Number of tries processed (a timeout or cancellation may throw before the action is being called)
• Catched exceptions during tries
• If succeed, cancelled or timeout
• The action delegate return value (if any)

NOTE: There's also a synchronous TryAction method, which supports timeout and cancellation also.

Asynchronous events

// Example type using an asynchronous event
public class YourType
{
    public readonly AsyncEvent<YourType, EventArgs> OnYourEvent;

    public YourType() => OnYourEvent = new(this);

    public async Task RaiseOnYourEventAsync()
        => await ((IAsyncEvent<YourType, EventArgs>)OnYourEvent).RaiseEventAsync();
}

// An example asynchronous event listener
async Task eventListener(YourType sender, EventArgs e, CancellationToken ct)
{
    ...
}

// Attach to the event and raise it
YourType obj = new();
Assert.IsFalse(obj.OnYourEvent);
obj.OnYourEvent.Listen(eventListener);
Assert.IsTrue(obj.OnYourEvent);
await obj.RaiseOnYourEventAsync();

// Detach the event listener
obj.OnYourEvent.Detach(eventListener);
Assert.IsFalse(obj.OnYourEvent);

An AsyncEvent<tSender, tArgs> instance will only export public event informations and functions like adding/removing event handlers, and if event handlers are present. For raising the event, you need to use the RaiseEventAsync methods which are available from the IAsyncEvent<tSender, tArgs> interface.

Timeout, cancellation, synchronous and asynchronous event handlers are supported. The AsyncEvent<tSender, tArgs> is designed to be thread-safe, while multiple threads are allowed to raise the event in parallel.

Checksum

ChecksumExtensions and ChecksumTransform allow generating a checksum:

byte[] data = ...,
    moreData = ...,
    checksum = data.CreateChecksum();
moreData.UpdateChecksum(checksum);

The default checksum length is 8 bytes and needs to be a power of two, if being customized. If you need a numeric value from the checksum bytes:

ulong numericChecksum = checksum.AsSpan().ToULong();

The algorithm uses XOR to modify the checksum bytes, which are zero by default. If the input data is only zero, the checksum will stay at zero. If you use the same input data for a 2nd time, the checksum will be equal to the one from the 1st time.

The ChecksumTransform is a HashAlgorithm and can be used as every .NET implemented hash algorithm (even it's not a hash, but only a checksum!):

byte[] checksum = ChecksumTransform.HashData(data);

You may register the checksum algorithm as "Checksum" using the Register method:

ChecksumTransform.Register();

IP sub-nets

The IpSubNet structure helps working with IPv4/6 sub-nets. It stores the network address and the bit-mask, for being able to provide

• the broadcast address or
• any IP address within the sub-net IP range and
• the number of usable IP addresses in the sub-net

and being able to

• determine if an IP address is within a sub-net
• determine if two sub-nets intersect
• a sub-net matches within another sub-net
• enumerate sub-net IP addresses
• compare sub-net lengths
• extend or shrink sub-nets
• combine two sub-nets
• serialize sub-net information platform independent
• determine if a sub-net is LAN (private), WAN (public) or loopback
• validate the correctness of a sub-net

on the fly, and many things more.

To construct the structure, you'll need one of these informations:

• Network CIDR notated ("192.168.0.0/24" for example)
• Network as IPAddress (all zero bytes will count the mask bits)
• Network as integer and the number of mask bits
• Network as IPAddress and the number of mask bits
• Network and mask as IPAddress
• Serialized sub-net data

Some basic examples:

// Create from CIDR notation
IpSubNet net = new("192.168.0.0/24");

// Validate CIDR notation
if(IpSubNet.TryParse("::/128", out IpSubNet subNet))
{
    // Valid CIDR notated sub-net
}

// Determine the network kind
Assert.IsTrue(net.IsLan);
Assert.IsFalse(net.IsWan);
Assert.IsFalse(net.IsLoopback);

// Get any IP address within a sub-net
Assert.AreEqual(IPAddress.Parse("192.168.0.1"), net[1]);

// Get the broadcast IP address
Assert.AreEqual(IPAddress.Parse("192.168.0.255"), net.BroadcastIPAddress);

// Determine if an IP address is within a sub-net
Assert.IsTrue(IPAddress.Parse("192.168.0.1") == net);
Assert.IsTrue(IPAddress.Parse("192.168.1.1") != net);

// Extend/shrink a sub-net
Assert.AreEqual("192.168.0.0/23", (net << 1).ToString());// Expand by one bit
Assert.AreEqual("192.168.0.0/25", (net >> 1).ToString());// Shrink by one bit

// Combine two sub-nets
IpSubNet combined = net + new IpSubNet("192.168.254.0/24");
Assert.AreEqual("192.168.0.0/16", combined.ToString());

// Merge two compatible (!) sub-nets
IpSubNet merged = net | new IpSubNet("192.168.0.0/8");
Assert.AreEqual("192.168.0.0/8", merged.ToString());

// Determine if two sub-nets intersect, or one fits into another
IpSubNet largerNet = new("192.168.0.0/16"),
    smallerNet = new("192.168.0.0/30"),
    otherNet = new("10.0.0.0/8");
Assert.AreEqual(net & largetNet, net);// net fits into largerNet
Assert.AreEqual(net & smallerNet, smallerNet);// net intersects smallerNet
Assert.AreEqual(net & otherNet, IpSubNet.ZeroV4);// no intersection between net and otherNet

// Serialization
byte[] serialized = net;// Serialize
IpSubNet net2 = serialized;// Deserialize
Assert.AreEqual(net, net2);

Centralized error handling

By setting ErrorHandling.ErrorHandler to your custom error handler, you can handle errors centralized. The error handling

1. will write to the debug console
2. will write to the logging
3. invoke the ErrorHandling.ErrorHandler (if any)
4. raise the ErrorHandling.OnError event

You may set ErrorHandling.ErrorCollectingHandler as error handler, or call that method from your custom error handler to collect errors in ErrorHandling.Errors.

Unhandled exceptions of the current app domain will be handled by this error handling. To handle a catched exception within your code, you can call the ErrorHandling.Handle method.

By setting ErrorHandling.DebugOnError to true (which is the default), an attached debugger will break before ErrorHandling.Handle handles an exception, finally.

Your custom error handler may

• store environment informations in a DBMS
• send an email
• do whatever is required to handle any error later

The ErrorHandling uses an ErrorInfo object, which can be implicit casted from/to an Exception. For your custom error handling you may want to host additional error informations, which you may give as tag to the constructor, or you create a custom error information type, which extends ErrorInfo. You can define an additional error message, if you use the constructor which accepts a string as first argument.

CAUTION: An unhandled exception during error handling could cause an endless loop. For this reason any uncatched error handling exception MUST be ignored - they'll be written to STDERR instead.

NOTE: The default error handling won't act as fist chance error handler. You'll need to call ErrorHandling.Handle from your code in order to handle a catched exception manually.

NOTE: You can specify an error source ID, which may be one of the pre- defined IDs from the ErrorHandling constants, or a custom value. If you use custom values, please only use bits 17..31, since the bits 1..16 are reserved for pre-defined error source IDs. Example for defining a custom error source ID:

public const int CUSTOM_ERROR_SOURCE = 1 << 16;

You can count from one as usual, but shift the ID 16 bits to the left, which enables you to define up to 32,768 different positive custom error sources. You may also use all the Int32 negative values for +2,147,483,648 custom error source IDs.

Delayed tasks

You'll need to add the DelayService.Instance to your apps hosted services, then you can use the delay like this:

await new Delay(TimeSpan.FromSeconds(3)).Task;

The line above will wait for 3 seconds and then continue in the current processing, while the delay could be used from other threads, too, if you did communicate the delays GUID (delays will organize themselfes in the DelayTable).

WARNING: The delays are not exact!

NOTE: Delay will be disposed automatted.

To cancel a delay, call the Cancel(Async) method of the Delay instance.

If the Delay was disposed or cancelled, awaiting the Task will throw an ObjectDisposedException or OperationCancelledException.

A delay is similar to Timeout, but it doesn't use its own timer and is a bit more easy to use for some specialized tasks.

Progress

A ProcessingProgress can be a

• counting progress with a total and a current count
• a progress collection with counting sub-progresses

A progress collection receives events of sub-progresses and forwards their events. The collection is self-managing - done sub-progresses will be removed and disposed automatically.

To display a progress with automatic updates, you can attach to the events of the progress (collection):

• OnProgress: The progress changed (will be forwarded until the root)
• OnAllProgress: The overall progress changed (will be raised after OnProgress, but won't be forwarded)
• OnStatus: A progress status message was updated (won't be forwarded)
• OnDone: A progress was done (will be forwarded until the root)

A progress can be canceled using the Cancel method. The OnDone event will be raised, IsDone will be false, but IsCanceled will be true. Canceling a collection means canceling all sub-progresses, too.

You can use the AllProgress property to get the current progress in %.

Example counting progress:

using ProcessingProgress progress = new()
{
    Total = 50
};
for(int i = 0; i < 50; i++)
{
    // Do some work
    progress.Update();// Increase the current count by one
}

Example progress collection:

using ProcessingProgress progressCollection = new();
ProcessingProgress progress = new()
{
    Total = 50
};
progressCollection.AddSubProgress(progress);
for(int i = 0; i < 50; i++)
{
    // Do some work
    progress.Update();// Increase the current count by one
}
// Now progress was disposed and removed from progressCollection, because it was done

NOTE: Total may be changed until a progress was completed or canceled.

Sensitive data handling

Using the SensitiveDataAttribute you can mark properties which host sensitive information. This could be used for the logging, for example: As you don't want sensitive data to appear in your logfiles, you may want to filter them out during logging. This could look like this:

public class YourObject
{
    public string LoggedData { get; set; }

    [SensitiveData]
    public string HiddenData { get; set; }
}

Logging.WriteDebug($"Object: {yourObjectInstance.ToDictionary().ToJson()}");

The ToDictionary object extension will filter the sensitive information from the given object, so that the ToJson extension will process on a sanitized data structure, which doesn't contain sensitive data.

You may use the attribute in other places, too, and handle values from such marked properties accordingly. It's also possible to extend the attribute with a value sanization method:

public class HidePasswordAttribute : SensitiveDataAttribute
{
    public HidePasswordAttribute() : base() { }

    public override bool CanSanitizeValue => true;

    public override object? CreateSanitizedValue(object obj, string propertyName, object? value)
    {
        if(value is not string pwd) return "(no password string value)";
        if(pwd.Length < 12) return "(password value too short)";
        if(pwd.Length > byte.MaxValue) return "(password value too long)";
        return "(valid password value hidden)";
    }
}

As soon as CanSanitizeValue delivers true, supporting code should call the CreateSanitizedValue method to create a replacement for the the actual value in an output.

Object storage

An object storage stores objects in memory and in any backend. If a number of in-memory objects was reached, least accessed objects will be removed from memory. On request an object can be re-created from the backend, and will then be stored in-memory again.

The object doesn't have to be stored in a backend. They may also be objects which require a lot of resources for their initialization, but will be accessed frequently and should be cached for that reason, for example.

There's only one requirement for an object to be object-storable: It needs to export a non-nullable unique object key by implementing the interface IStoredObject<T>.

All in all the object storage is a kind of memory cache for a single object type. The configured in-memory limit is only a soft-limit, 'cause the storage won't limit the number of used objects - but the number of unused, cached objects.

The implementing storage can control

• synchronous/asynchronous object creation
• object disposing

and override any other base object storage operation, if required.

Implemented operations:

• GetObject(Async): Get an object by its key (the returned wrapper needs to be disposed!)
• Release: Release object usage (will be called from the returned wrapper of GetObject(Async), when it's being disposed)
• Remove: Remove the object from the storage (if it's being deleted permanently, for example)

Localization

A basic localization support without built-in plural handling is available:

Translation.Locales["en-US"] = new(new Dictionary<string, string>()
{
    {"Hello", "Hello"},
    ...
});
Translation.Locales["de-DE"] = new(new Dictionary<string, string>()
{
    {"Hello", "Hallo"},
    ...
});
Translation.Current = Translation.Locales["de-DE"];

This initializes English and German translations, where English is always the main locale. To translate a text:

using static wan24.TranslationHelper.Ext;

string translated = _("Hello");

TIP: If you'd like to enable a keyword extractor to find texts, which will be stored as variable and translated from there later, when the locale is known, you can use the variable = __("Text"); syntax (when using static wan24.TranslationHelper;). The double score method returns the given string value 1:1 and is only being used as parser hint.

Or for a specific locale:

using static wan24.Translation;

string translated = Localize("de-DE", "Hello");

To implement plural support, you can extend the TranslationTerms type:

public sealed class YourTerms : TranslationTerms// Implements IReadOnlyDictionary<string, string>
{
    public YourTerms(IReadOnlyDictionary<string, string> terms) : base(terms) { }

    public override bool PluralSupport => true;

    public override string GetTerm(in string key, in int count)
    {
        // Return the translated plural term
    }
}

NOTE: Translation supports string parser usage.

To combine multiple translations for a single locale into one, you can use the CombinedTranslationTerms type.

You may also use localized filenames. For this you'll need to store files as

• filename.ext (fallback, if a known locale or an existing file isn't required)
• filename.en-EN.ext (localized file)
• ...

Then you can localize a filename:

using static wan24.Translation;

string fn = LocalizedFileName("de-DE", "/path/to/filename.ext");
Assert.AreEqual("/path/to/filename.de-DE.ext", fn);

IStringLocalizer interface

Using the .NET IStringLocalizer interface you can use the wan24-Core localization like this:

// After setting a Translation.Current as described above
builder.Services.AddSingleton<IStringLocalizerFactory, StringLocalizerFactory>();
builder.Services.AddTransient(typeof(IStringSerializer<>), typeof(GenericTranslation<>));

The StringLocalizerFactory and GenericTranslation<> will fall back to Translation.Current.

Informations for translators

A string to translate may contain placeholders like %{N}, where N is any number. These placeholders address variables and may occur in any order in the translation, as long as the original N value is being used (the placeholders must not be re-numbered in the translation). Also placeholders with a name instead of a numeric value are possible and should be used 1:1 within the translation (but may be reordered, if the grammatics require it).

The escape sequence \n or \r\n is a line break which must be used for a line break in the translation, too.

The escape sequence \t is a tabulator which must be used for a tabulator in the translation, too.

The escape sequence \" is a double-quote. Double-quotes should be escaped that way.

If a filename is localized using a locale code like en-US, the translation must use its new locale code instead - example: filename.en-US.ext becomes to filename.de-DE.ext in a German translation.

Enumeration classes

Using the EnumerationBase<T> base type you can implement enumeration classes like this:

public sealed class YourEnum : EnumerationBase<YourEnum>
{
    public static readonly YourEnum Value1 = new(1, nameof(Value1));
    public static readonly YourEnum Value2 = new(2, nameof(Value2));
    ...

    private YourEnum(int value, string name) : base(value, name) { }
}

Your implementation needs to fit some restrictions:

1. Values and names are unique
2. Names must match their readonly-field name
3. Your type must be sealed and use private construction
4. Your type must extend EnumerationBase<T> (not EnumerationBase directly)

Transactions

You can choose between sequential transactions (Transaction) and parallel action executing transactions (ParallelTransaction):

Sequential

using Transaction transaction = new();
object? returnValue = transaction.Execute(
    ()=> /* Perform the action here and return a value (optional) */, 
    (transaction, returnValue) => /* Rollback for the action */
    );
returnValue = await transaction.ExecuteAsync(
    async (cancelToken)=> /* Perform the action here and return a value (optional) */, 
    async (transaction, returnValue. cancelToken) => /* Rollback for the action */
    );
// Commit the actions (if disposing, uncommitted actions will be rolled back!)
transaction.Commit();

Parallel

ParallelTransaction transaction = new();
await using(transaction)
{
    // The Execute methods will synchronize enqueueing the action asynchronous
    (int index, Task task) = await transaction.ExecuteAsync(
        async (cancelToken) => /* Perform the action here */, 
        async (transaction, returnValue, cancelToken) => /* Rollback for the action */
        );
    // index has the action index which allows to retrieve the return value or the exception later
    (index, Task<object?> resultTask) = await transaction.ExecuteAsync(
        async (cancelToken) => /* Perform the action here and return a value */, 
        async (transaction, returnValue, cancelToken) => /* Rollback for the action */
        );
    object? returnValue = await resultTask;// Get the return value from the action task
    await transaction.WaitDoneAsync();// Will throw if any action failed
    // Commit the actions (if disposing, uncommitted actions will be rolled back!)
    transaction.Commit();
}

You can cancel pending actions using the CancelAsync method. A canceled transaction needs to be rolled back before reuse. If an action failed, the transaction can be canceled by setting CancelOnError to true (which is the default). Committing an undone or canceled transaction will throw.

Using the OnError/Done events you may become informed on error, or if all pending actions are done (OnDone may be called multiple times).

Nested transactions

Using the Append(Async) methods of a transaction, you may nest in any other transaction (which won't be disposed, if the hosting transaction is disposing!).

App JSON configuration

Using the AppConfig you can easily implement a JSON configuration for your app:

await AppConfig.LoadAsync();

This will load the config.json, apply configured settings (and also the configuration from CLI arguments) and bootstrap your app.

For implementing a customized configuration, you can extend AppConfig and change the loading at your apps startup slightly:

await AppConfig.LoadAsync<YourAppConfig>();

NOTE: Use validation attributes on configuration properties! They'll be used to validate the configured values during loading the JSON structure.

AppConfigBase can be used as an app configuration object base class, which provides support for the AppConfigAttribute, but doesn't include the settings from AppConfig. You only need to implement the Apply(Async) methods.

The AppConfigAttribute should be used for every property which can store an IAppConfig value and should be applied automatic. You can set a Priority and specify that the configuration should only be applied AfterBootstrap. To apply such sub-configurations, call the ApplyProperties(Async) methods from your Apply(Async) method implementations.

When using the AppConfig, this is the app configuration process:

1. configure Logging
2. configure Settings and ENV
3. apply DefaultCliArguments to ENV.CliArguments
4. configure static [CliConfig] properties from Properties
5. apply custom [AppConfig] properties before bootstrapping
6. apply CLI configuration arguments using CliConfig.Apply
7. bootstrap using Bootstrap.Async
8. apply custom [AppConfig] properties after bootstrapping

This means:

• you can define factory settings before applying a JSON configuration
• the JSON configuration defines the app setup defaults, which override the factory defaults
• app setup defaults can be overridden with CLI arguments, if required

You may disable CLI argument configuration and bootstrapping using the ApplyCliArguments and Bootstrap properties (which can't be overridden by the JSON configuration, but by extending AppConfig!).

In-memory cache

The InMemoryCache can be used for caching any kind of object in memory:

InMemoryCacheOptions options = new()
{
    SoftCountLimit = 1_000
    // SoftCountLimit, SoftSizeLimit, AgeLimit or IdleLimit is required
};
using InMemoryCache cache = new(options);
await cache.StartAsync();

// Adding a new item to the cache
cache.Add("key", item);
await cache.AddAsync("key", item);

// Get (or add) an item from the cache
item = cache.Get("key")?.Item;
item = (await cache.GetAsync("key"))?.Item;

// Remove an item from the cache
cache.TryRemove("key");

// Clear the cache while the cache is still serving items
cache.Clear(disposeItems: true);
await cache.ClearAsync(disposeItems: true);

NOTE: This in-memory cache implementation isn't compatible with the .NET built-in approaches. It implements IHostedService and needs to be started before it can be used! The asynchronous methods should be used where possible, if the chached item type implements IAsyncDisposable - otherwise the synchronous methods are fine, too.

The cache capacity can be limited by

• number of cache entries (hard limit)
• cached item size (hard limit)

while auto-removing the cache is being done by a background job by

• cache entry timeout (removals won't be timed exactly!)
• cache entry age
• cache entry idle state
• number of cache entries (soft limit)
• cached item size (soft limit)
• max. memory usage in bytes
• optional custom management strategies

in a fixed interval. An item may also be removed when disposing, if it's an IDisposableObject and the cache entry was configured to observe the items disposing (not recommended if avoidable).

NOTE: The in-memory cache isn't too strict with limits, they might overflow slightly.

Hard limits are being applied when adding new items, while soft limits are being applied by the tidy process. The cache size can be manually reduced at any time by

• a max. number of cache entries
• a max. item size
• a max. cache entry age
• a max. cache entry idle time
• an optional custom cache entry reducing strategy

The GetAsync method allows a delegate for creating a new cached item, which will be added to the cache in case there was no cache entry for the requested key already.

Cached items which are being auto-removed from the cache will be disposed, if possible. To avoid disposing an item while it's still in use, you can wrap the item with an AutoDisposer<T>:

InMemoryCacheOptions options = new()
{
    SoftCountLimit = 1_000
    // SoftCountLimit, SoftSizeLimit, AgeLimit or IdleLimit is required
};
using InMemoryCache<AutoDisposer<ItemType>> cache = new(options);
await cache.StartAsync();

// Add a new item to the cache
cache.Add("key", new(item));
await cache.AddAsync("key", new(item));

// Get (or add) an item from the cache
using AutoDisposer<ItemType>.Context? itemContext = 
    await cache.GetItemContext("key");
using AutoDisposer<ItemType>.Context? itemContext = 
    await cache.GetItemContextAsync("key");
// Now itemContext.Object can be used while it's not being disposed for sure

// Remove an item from the cache
if(cache.TryRemove("key")?.Item is AutoDisposer<ItemType> removedItemDisposer)
    removedItemDisposer.ShouldDispose = true;
if(cache.TryRemove("key")?.Item is AutoDisposer<ItemType> removedItemDisposer)
    await removedItemDisposer.SetShouldDisposeAsync();

WARNING: If the cache is being disposed, all cached items will be disposed, too, no matter if they're still in use or not!

CAUTION: If the InMemoryCacheOptions.MaxItemSize value was set, and an oversized disposable item is being added, an OutOfMemoryException will be thrown.

IInMemoryCacheItem interface

By implementing the IInMemoryCacheItem interface a cacheable item can export its unique key and its size in the cache. While the key should never change and is used for adding a new item only, the size property getter may return variable values during runtime, which will be respected during a cache cleanup.

Returning cache entry options is optional.