ConsoleAppFramework v5 is Zero Dependency, Zero Overhead, Zero Reflection, Zero Allocation, AOT Safe CLI Framework powered by C# Source Generator; achieves exceptionally high performance and minimal binary size. Leveraging the latest features of .NET 8 and C# 12 (IncrementalGenerator, managed function pointer, params arrays and default values lambda expression, ISpanParsable<T>
, PosixSignalRegistration
, etc.), this library ensures maximum performance while maintaining flexibility and extensibility.
Set
RunStrategy=ColdStart WarmupCount=0
to calculate the cold start benchmark, which is suitable for CLI application.
The magical performance is achieved by statically generating everything and parsing inline. Let's take a look at a minimal example:
using ConsoleAppFramework;
// args: ./cmd --foo 10 --bar 20
ConsoleApp.Run(args, (int foo, int bar) => Console.WriteLine($"Sum: {foo + bar}"));
Unlike typical Source Generators that use attributes as keys for generation, ConsoleAppFramework analyzes the provided lambda expressions or method references and generates the actual code body of the Run method.
internal static partial class ConsoleApp
{
// Generate the Run method itself with arguments and body to match the lambda expression
public static void Run(string[] args, Action<int, int> command)
{
// code body
}
}
Full generated source code
namespace ConsoleAppFramework;
internal static partial class ConsoleApp
{
public static void Run(string[] args, Action<int, int> command)
{
if (TryShowHelpOrVersion(args, 2, -1)) return;
var arg0 = default(int);
var arg0Parsed = false;
var arg1 = default(int);
var arg1Parsed = false;
try
{
for (int i = 0; i < args.Length; i++)
{
var name = args[i];
switch (name)
{
case "--foo":
{
if (!TryIncrementIndex(ref i, args.Length) || !int.TryParse(args[i], out arg0)) { ThrowArgumentParseFailed("foo", args[i]); }
arg0Parsed = true;
break;
}
case "--bar":
{
if (!TryIncrementIndex(ref i, args.Length) || !int.TryParse(args[i], out arg1)) { ThrowArgumentParseFailed("bar", args[i]); }
arg1Parsed = true;
break;
}
default:
if (string.Equals(name, "--foo", StringComparison.OrdinalIgnoreCase))
{
if (!TryIncrementIndex(ref i, args.Length) || !int.TryParse(args[i], out arg0)) { ThrowArgumentParseFailed("foo", args[i]); }
arg0Parsed = true;
break;
}
if (string.Equals(name, "--bar", StringComparison.OrdinalIgnoreCase))
{
if (!TryIncrementIndex(ref i, args.Length) || !int.TryParse(args[i], out arg1)) { ThrowArgumentParseFailed("bar", args[i]); }
arg1Parsed = true;
break;
}
ThrowArgumentNameNotFound(name);
break;
}
}
if (!arg0Parsed) ThrowRequiredArgumentNotParsed("foo");
if (!arg1Parsed) ThrowRequiredArgumentNotParsed("bar");
command(arg0!, arg1!);
}
catch (Exception ex)
{
Environment.ExitCode = 1;
if (ex is ValidationException or ArgumentParseFailedException)
{
LogError(ex.Message);
}
else
{
LogError(ex.ToString());
}
}
}
[MethodImpl(MethodImplOptions.AggressiveInlining)]
static bool TryIncrementIndex(ref int index, int length)
{
if (index < length)
{
index++;
return true;
}
return false;
}
static partial void ShowHelp(int helpId)
{
Log("""
Usage: [options...] [-h|--help] [--version]
Options:
--foo <int> (Required)
--bar <int> (Required)
""");
}
}
As you can see, the code is straightforward and simple, making it easy to imagine the execution cost of the framework portion. That's right, it's zero. This technique was influenced by Rust's macros. Rust has Attribute-like macros and Function-like macros, and ConsoleAppFramework's generation can be considered as Function-like macros.
The ConsoleApp
class, along with everything else, is generated entirely by the Source Generator, resulting in no dependencies, including ConsoleAppFramework itself. This characteristic should contribute to the small assembly size and ease of handling, including support for Native AOT.
Moreover, CLI applications typically involve single-shot execution from a cold start. As a result, common optimization techniques such as dynamic code generation (IL Emit, ExpressionTree.Compile) and caching (ArrayPool) do not work effectively. ConsoleAppFramework generates everything statically in advance, achieving performance equivalent to optimized hand-written code without reflection or boxing.
ConsoleAppFramework offers a rich set of features as a framework. The Source Generator analyzes which modules are being used and generates the minimal code necessary to implement the desired functionality.
- SIGINT/SIGTERM(Ctrl+C) handling with gracefully shutdown via
CancellationToken
- Filter(middleware) pipeline to intercept before/after execution
- Exit code management
- Support for async commands
- Registration of multiple commands
- Registration of nested commands
- Setting option aliases and descriptions from code document comment
System.ComponentModel.DataAnnotations
attribute-based Validation- Dependency Injection for command registration by type and public methods
Microsoft.Extensions
(Logging, Configuration, etc...) integration- High performance value parsing via
ISpanParsable<T>
- Parsing of params arrays
- Parsing of JSON arguments
- Help(
-h|--help
) option builder - Default show version(
--version
) option
As you can see from the generated output, the help display is also fast. In typical frameworks, the help string is constructed after the help invocation. However, in ConsoleAppFramework, the help is embedded as string constants, achieving the absolute maximum performance that cannot be surpassed!
This library is distributed via NuGet, minimal requirement is .NET 8 and C# 12.
PM> Install-Package ConsoleAppFramework
ConsoleAppFramework is an analyzer (Source Generator) and does not have any dll references. When referenced, the entry point class ConsoleAppFramework.ConsoleApp
is generated internally.
The first argument of Run
or RunAsync
can be string[] args
, and the second argument can be any lambda expression, method, or function reference. Based on the content of the second argument, the corresponding function is automatically generated.
using ConsoleAppFramework;
ConsoleApp.Run(args, (string name) => Console.WriteLine($"Hello {name}"));
You can execute command like sampletool --name "foo"
.
- The return value can be
void
,int
,Task
, orTask<int>
- If an
int
is returned, that value will be set toEnvironment.ExitCode
- If an
- By default, option argument names are converted to
--lower-kebab-case
- For example,
jsonValue
becomes--json-value
- Option argument names are case-insensitive, but lower-case matches faster
- For example,
When passing a method, you can write it as follows:
ConsoleApp.Run(args, Sum);
void Sum(int x, int y) => Console.Write(x + y);
Additionally, for static functions, you can pass them as function pointers. In that case, the managed function pointer arguments will be generated, resulting in maximum performance.
unsafe
{
ConsoleApp.Run(args, &Sum);
}
static void Sum(int x, int y) => Console.Write(x + y);
public static unsafe void Run(string[] args, delegate* managed<int, int, void> command)
Unfortunately, currently static lambdas cannot be assigned to function pointers, so defining a named function is necessary.
When defining an asynchronous method using a lambda expression, the async
keyword is required.
// --foo, --bar
await ConsoleApp.RunAsync(args, async (int foo, int bar, CancellationToken cancellationToken) =>
{
await Task.Delay(TimeSpan.FromSeconds(5), cancellationToken);
Console.WriteLine($"Sum: {foo + bar}");
});
You can use either the Run
or RunAsync
method for invocation. It is optional to use CancellationToken
as an argument. This becomes a special parameter and is excluded from the command options. Internally, it uses PosixSignalRegistration
to handle SIGINT
, SIGTERM
, and SIGKILL
. When these signals are invoked (e.g., Ctrl+C), the CancellationToken is set to CancellationRequested. If CancellationToken
is not used as an argument, these signals will not be handled, and the program will terminate immediately. For more details, refer to the CancellationToken and Gracefully Shutdown section.
By default, if -h
or --help
is provided, or if no arguments are passed, the help display will be invoked.
ConsoleApp.Run(args, (string message) => Console.Write($"Hello, {message}"));
Usage: [options...] [-h|--help] [--version]
Options:
--message <string> (Required)
In ConsoleAppFramework, instead of using attributes, you can provide descriptions and aliases for functions by writing Document Comments. This avoids the common issue in frameworks where arguments become cluttered with attributes, making the code difficult to read. With this approach, a natural writing style is achieved.
ConsoleApp.Run(args, Commands.Hello);
static class Commands
{
/// <summary>
/// Display Hello.
/// </summary>
/// <param name="message">-m, Message to show.</param>
public static void Hello(string message) => Console.Write($"Hello, {message}");
}
Usage: [options...] [-h|--help] [--version]
Display Hello.
Options:
-m|--message <string> Message to show. (Required)
To add aliases to parameters, list the aliases separated by |
before the comma in the comment. For example, if you write a comment like -a|-b|--abcde, Description.
, then -a
, -b
, and --abcde
will be treated as aliases, and Description.
will be the description.
Unfortunately, due to current C# specifications, lambda expressions and local functions do not support document comments, so a class is required.
In addition to -h|--help
, there is another special built-in option: --version
. In default, it displays the AssemblyInformationalVersion
or AssemblyVersion
. You can configure version string by ConsoleApp.Version
, for example ConsoleApp.Version = "2001.9.3f14-preview2";
.
If you want to register multiple commands or perform complex operations (such as adding filters), instead of using Run/RunAsync
, obtain the ConsoleAppBuilder
using ConsoleApp.Create()
. Call Add
, Add<T>
, or UseFilter<T>
multiple times on the ConsoleAppBuilder
to register commands and filters, and finally execute the application using Run
or RunAsync
.
var app = ConsoleApp.Create();
app.Add("", (string msg) => Console.WriteLine(msg));
app.Add("echo", (string msg) => Console.WriteLine(msg));
app.Add("sum", (int x, int y) => Console.WriteLine(x + y));
// --msg
// echo --msg
// sum --x --y
app.Run(args);
The first argument of Add
is the command name. If you specify an empty string ""
, it becomes the root command. Unlike parameters, command names are case-sensitive and cannot have multiple names.
With Add<T>
, you can add multiple commands at once using a class-based approach, where public methods are treated as commands. If you want to write document comments for multiple commands, this approach allows for cleaner code, so it is recommended. Additionally, as mentioned later, you can also write clean code for Dependency Injection (DI) using constructor injection.
var app = ConsoleApp.Create();
app.Add<MyCommands>();
app.Run(args);
public class MyCommands
{
/// <summary>Root command test.</summary>
/// <param name="msg">-m, Message to show.</param>
[Command("")]
public void Root(string msg) => Console.WriteLine(msg);
/// <summary>Display message.</summary>
/// <param name="msg">Message to show.</param>
public void Echo(string msg) => Console.WriteLine(msg);
/// <summary>Sum parameters.</summary>
/// <param name="x">left value.</param>
/// <param name="y">right value.</param>
public void Sum(int x, int y) => Console.WriteLine(x + y);
}
When you check the registered commands with --help
, it will look like this. Note that you can register multiple Add<T>
and also add commands using Add
.
Usage: [command] [options...] [-h|--help] [--version]
Root command test.
Options:
-m|--msg <string> Message to show. (Required)
Commands:
echo Display message.
sum Sum parameters.
By default, the command name is derived from the method name converted to lower-kebab-case
. However, you can change the name to any desired value using the [Command(string commandName)]
attribute.
If the class implements IDisposable
or IAsyncDisposable
, the Dispose or DisposeAsync method will be called after the command execution.
You can create a deep command hierarchy by adding commands with paths separated by space(
) when registering them. This allows you to add commands at nested levels.
var app = ConsoleApp.Create();
app.Add("foo", () => { });
app.Add("foo bar", () => { });
app.Add("foo bar barbaz", () => { });
app.Add("foo baz", () => { });
// Commands:
// foo
// foo bar
// foo bar barbaz
// foo baz
app.Run(args);
Add<T>
can also add commands to a hierarchy by passing a string commandPath
argument.
var app = ConsoleApp.Create();
app.Add<MyCommands>("foo");
// Commands:
// foo Root command test.
// foo echo Display message.
// foo sum Sum parameters.
app.Run(args);
In ConsoleAppFramework
, the number and types of registered commands are statically determined at compile time. For example, let's register the following four commands:
app.Add("foo", () => { });
app.Add("foo bar", (int x, int y) => { });
app.Add("foo bar barbaz", (DateTime dateTime) => { });
app.Add("foo baz", async (string foo = "test", CancellationToken cancellationToken = default) => { });
The Source Generator generates four fields and holds them with specific types.
partial struct ConsoleAppBuilder
{
Action command0 = default!;
Action<int, int> command1 = default!;
Action<global::System.DateTime> command2 = default!;
Func<string, global::System.Threading.CancellationToken, Task> command3 = default!;
partial void AddCore(string commandName, Delegate command)
{
switch (commandName)
{
case "foo":
this.command0 = Unsafe.As<Action>(command);
break;
case "foo bar":
this.command1 = Unsafe.As<Action<int, int>>(command);
break;
case "foo bar barbaz":
this.command2 = Unsafe.As<Action<global::System.DateTime>>(command);
break;
case "foo baz":
this.command3 = Unsafe.As<Func<string, global::System.Threading.CancellationToken, Task>>(command);
break;
default:
break;
}
}
}
This ensures the fastest execution speed without any additional unnecessary allocations such as arrays and without any boxing since it holds static delegate types.
Command routing also generates a switch of nested string constants.
partial void RunCore(string[] args)
{
if (args.Length == 0)
{
ShowHelp(-1);
return;
}
switch (args[0])
{
case "foo":
if (args.Length == 1)
{
RunCommand0(args, args.AsSpan(1), command0);
return;
}
switch (args[1])
{
case "bar":
if (args.Length == 2)
{
RunCommand1(args, args.AsSpan(2), command1);
return;
}
switch (args[2])
{
case "barbaz":
RunCommand2(args, args.AsSpan(3), command2);
break;
default:
RunCommand1(args, args.AsSpan(2), command1);
break;
}
break;
case "baz":
RunCommand3(args, args.AsSpan(2), command3);
break;
default:
RunCommand0(args, args.AsSpan(1), command0);
break;
}
break;
default:
ShowHelp(-1);
break;
}
}
The C# compiler performs complex generation for string constant switches, making them extremely fast, and it would be difficult to achieve faster routing than this.
The method parameter names and types determine how to parse and bind values from the command-line arguments. When using lambda expressions, optional values and params
arrays supported from C# 12 are also supported.
ConsoleApp.Run(args, (
[Argument]DateTime dateTime, // Argument
[Argument]Guid guidvalue, //
int intVar, // required
bool boolFlag, // flag
MyEnum enumValue, // enum
int[] array, // array
MyClass obj, // object
string optional = "abcde", // optional
double? nullableValue = null, // nullable
params string[] paramsArray // params
) => { });
When using ConsoleApp.Run
, you can check the syntax of the command line in the tooltip to see how it is generated.
For the rules on converting parameter names to option names, aliases, and how to set documentation, refer to the Option aliases section.
Parameters marked with the [Argument]
attribute receive values in order without parameter names. This attribute can only be set on sequential parameters from the beginning.
To convert from string arguments to various types, basic primitive types (string
, char
, sbyte
, byte
, short
, int
, long
, uint
, ushort
, ulong
, decimal
, float
, double
) use TryParse
. For types that implement ISpanParsable<T>
(DateTime
, DateTimeOffset
, Guid
, BigInteger
, Complex
, Half
, Int128
, etc.), IParsable.TryParse or ISpanParsable.TryParse is used.
For enum
, it is parsed using Enum.TryParse(ignoreCase: true)
.
bool
is treated as a flag and is always optional. It becomes true
when the parameter name is passed.
Array parsing has three special patterns.
For a regular T[]
, if the value starts with [
, it is parsed using JsonSerialzier.Deserialize
. Otherwise, it is parsed as comma-separated values. For example, [1,2,3]
or 1,2,3
are allowed as values. To set an empty array, pass []
.
For params T[]
, all subsequent arguments become the values of the array. For example, if there is an input like --paramsArray foo bar baz
, it will be bound to a value like ["foo", "bar", "baz"]
.
If none of the above cases apply, JsonSerializer.Deserialize<T>
is used to perform binding as JSON. However, CancellationToken
and ConsoleAppContext
are treated as special types and excluded from binding. Also, parameters with the [FromServices]
attribute are not subject to binding.
If you want to change the deserialization options, you can set JsonSerializerOptions
to ConsoleApp.JsonSerializerOptions
.
To perform custom binding to existing types that do not support ISpanParsable<T>
, you can create and set up a custom parser. For example, if you want to pass System.Numerics.Vector3
as a comma-separated string like 1.3,4.12,5.947
and parse it, you can create an Attribute
with AttributeTargets.Parameter
that implements IArgumentParser<T>
's static bool TryParse(ReadOnlySpan<char> s, out Vector3 result)
as follows:
[AttributeUsage(AttributeTargets.Parameter)]
public class Vector3ParserAttribute : Attribute, IArgumentParser<Vector3>
{
public static bool TryParse(ReadOnlySpan<char> s, out Vector3 result)
{
Span<Range> ranges = stackalloc Range[3];
var splitCount = s.Split(ranges, ',');
if (splitCount != 3)
{
result = default;
return false;
}
float x;
float y;
float z;
if (float.TryParse(s[ranges[0]], out x) && float.TryParse(s[ranges[1]], out y) && float.TryParse(s[ranges[2]], out z))
{
result = new Vector3(x, y, z);
return true;
}
result = default;
return false;
}
}
By setting this attribute on a parameter, the custom parser will be called when parsing the args.
ConsoleApp.Run(args, ([Vector3Parser] Vector3 position) => Console.WriteLine(position));
While there are some standards for command-line arguments, such as UNIX tools and POSIX, there is no absolute specification. The Command-line syntax overview for System.CommandLine provides an explanation of the specifications adopted by System.CommandLine. However, ConsoleAppFramework, while referring to these specifications to some extent, does not necessarily aim to fully comply with them.
For example, specifications that change behavior based on -x
and -X
or allow bundling -f -d -x
as -fdx
are not easy to understand and also take time to parse. The poor performance of System.CommandLine may be influenced by its adherence to complex grammar. Therefore, ConsoleAppFramework prioritizes performance and clear rules. It uses lower-kebab-case as the basis while allowing case-insensitive matching. It does not support ambiguous grammar that cannot be processed in a single pass or takes time to parse.
System.CommandLine seems to be aiming for a new direction in .NET 9 and .NET 10, but from a performance perspective, it will never surpass ConsoleAppFramework.
In ConsoleAppFramework, when you pass a CancellationToken
as an argument, it can be used to check for interruption commands (SIGINT/SIGTERM/SIGKILL - Ctrl+C) rather than being treated as a parameter. For handling this, ConsoleAppFramework performs special code generation when a CancellationToken
is included in the parameters.
using var posixSignalHandler = PosixSignalHandler.Register(ConsoleApp.Timeout);
var arg0 = posixSignalHandler.Token;
await Task.Run(() => command(arg0!)).WaitAsync(posixSignalHandler.TimeoutToken);
If a CancellationToken is not passed, the application is immediately forced to terminate when an interruption command (Ctrl+C) is received. However, if a CancellationToken is present, it internally uses PosixSignalRegistration
to hook SIGINT/SIGTERM/SIGKILL and sets the CancellationToken to a canceled state. Additionally, it prevents forced termination to allow for a graceful shutdown.
If the CancellationToken is handled correctly, the application can perform proper termination processing based on the application's handling. However, if the CancellationToken is mishandled, the application may not terminate even when an interruption command is received. To avoid this, a timeout timer starts after the interruption command, and the application is forcibly terminated again after the specified time.
The default timeout is 5 seconds, but it can be changed using ConsoleApp.Timeout
. For example, setting it to ConsoleApp.Timeout = Timeout.InfiniteTimeSpan;
disables the forced termination caused by the timeout.
The hooking behavior using PosixSignalRegistration
is determined by the presence of a CancellationToken
(or always takes effect if a filter is set). Therefore, even for synchronous methods, it is possible to change the behavior by including a CancellationToken
as an argument.
If the method returns int
or Task<int>
, ConsoleAppFramework
will set the return value to the exit code. Due to the nature of code generation, when writing lambda expressions, you need to explicitly specify either int
or Task<int>
.
// return Random ExitCode...
ConsoleApp.Run(args, int () => Random.Shared.Next());
// return StatusCode
await ConsoleApp.RunAsync(args, async Task<int> (string url, CancellationToken cancellationToken) =>
{
using var client = new HttpClient();
var response = await client.GetAsync(url, cancellationToken);
return (int)response.StatusCode;
});
If the method throws an unhandled exception, ConsoleAppFramework always set 1
to the exit code. Also, in that case, output Exception.ToString
to ConsoleApp.LogError
(the default is Console.WriteLine
). If you want to modify this code, please create a custom filter. For more details, refer to the Filter section.
ConsoleAppFramework
performs validation when the parameters are marked with attributes for validation from System.ComponentModel.DataAnnotations
(more precisely, attributes that implement ValidationAttribute
). The validation occurs after parameter binding and before command execution. If the validation fails, it throws a ValidationException
.
ConsoleApp.Run(args, ([EmailAddress] string firstArg, [Range(0, 2)] int secondArg) => { });
For example, if you pass arguments like args = "--first-arg invalid.email --second-arg 10".Split(' ');
, you will see validation failure messages such as:
The firstArg field is not a valid e-mail address.
The field secondArg must be between 0 and 2.
By default, the ExitCode is set to 1 in this case.
Filters are provided as a mechanism to hook into the execution before and after. To use filters, define an internal class
that implements ConsoleAppFilter
.
internal class NopFilter(ConsoleAppFilter next) : ConsoleAppFilter(next) // ctor needs `ConsoleAppFilter next` and call base(next)
{
// implement InvokeAsync as filter body
public override async Task InvokeAsync(ConsoleAppContext context, CancellationToken cancellationToken)
{
try
{
/* on before */
await Next.InvokeAsync(context, cancellationToken); // invoke next filter or command body
/* on after */
}
catch
{
/* on error */
throw;
}
finally
{
/* on finally */
}
}
}
Filters can be attached multiple times to "global", "class", or "method" using UseFilter<T>
or [ConsoleAppFilter<T>]
. The order of filters is global → class → method, and the execution order is determined by the definition order from top to bottom.
var app = ConsoleApp.Create();
// global filters
app.UseFilter<NopFilter>(); //order 1
app.UseFilter<NopFilter>(); //order 2
app.Add<MyCommand>();
app.Run(args);
// per class filters
[ConsoleAppFilter<NopFilter>] // order 3
[ConsoleAppFilter<NopFilter>] // order 4
public class MyCommand
{
// per method filters
[ConsoleAppFilter<NopFilter>] // order 5
[ConsoleAppFilter<NopFilter>] // order 6
public void Echo(string msg) => Console.WriteLine(msg);
}
Filters allow various processes to be shared. For example, the process of measuring execution time can be written as follows:
internal class LogRunningTimeFilter(ConsoleAppFilter next) : ConsoleAppFilter(next)
{
public override async Task InvokeAsync(ConsoleAppContext context, CancellationToken cancellationToken)
{
var startTime = Stopwatch.GetTimestamp();
ConsoleApp.Log($"Execute command at {DateTime.UtcNow.ToLocalTime()}"); // LocalTime for human readable time
try
{
await Next.InvokeAsync(context, cancellationToken);
ConsoleApp.Log($"Command execute successfully at {DateTime.UtcNow.ToLocalTime()}, Elapsed: " + (Stopwatch.GetElapsedTime(startTime)));
}
catch
{
ConsoleApp.Log($"Command execute failed at {DateTime.UtcNow.ToLocalTime()}, Elapsed: " + (Stopwatch.GetElapsedTime(startTime)));
throw;
}
}
}
In case of an exception, the ExitCode
is usually 1
, and the stack trace is also displayed. However, by applying an exception handling filter, the behavior can be changed.
internal class ChangeExitCodeFilter(ConsoleAppFilter next) : ConsoleAppFilter(next)
{
public override async Task InvokeAsync(ConsoleAppContext context, CancellationToken cancellationToken)
{
try
{
await Next.InvokeAsync(context, cancellationToken);
}
catch (Exception ex)
{
if (ex is OperationCanceledException) return;
Environment.ExitCode = 9999; // change custom exit code
ConsoleApp.LogError(ex.Message); // .ToString() shows stacktrace, .Message can avoid showing stacktrace to user.
}
}
}
Filters are executed after the command name routing is completed. If you want to prohibit multiple executions for each command name, you can use ConsoleAppContext.CommandName
as the key.
internal class PreventMultipleSameCommandInvokeFilter(ConsoleAppFilter next) : ConsoleAppFilter(next)
{
public override async Task InvokeAsync(ConsoleAppContext context, CancellationToken cancellationToken)
{
var basePath = Assembly.GetEntryAssembly()?.Location.Replace(Path.DirectorySeparatorChar, '_');
var mutexKey = $"{basePath}$$${context.CommandName}"; // lock per command-name
using var mutex = new Mutex(true, mutexKey, out var createdNew);
if (!createdNew)
{
throw new Exception($"already running command:{context.CommandName} in another process.");
}
await Next.InvokeAsync(context, cancellationToken);
}
}
If you want to pass values between filters or to commands, you can use ConsoleAppContext.State
. For example, if you want to perform authentication processing and pass around the ID, you can write code like the following. Since ConsoleAppContext
is an immutable record, you need to pass the rewritten context to Next using the with
syntax.
internal class AuthenticationFilter(ConsoleAppFilter next) : ConsoleAppFilter(next)
{
public override async Task InvokeAsync(ConsoleAppContext context, CancellationToken cancellationToken)
{
var requestId = Guid.NewGuid();
var userId = await GetUserIdAsync();
// setup new state to context
var authedContext = context with { State = new ApplicationContext(requestId, userId) };
await Next.InvokeAsync(authedContext, cancellationToken);
}
// get user-id from DB/auth saas/others
async Task<int> GetUserIdAsync()
{
await Task.Delay(TimeSpan.FromSeconds(1));
return 1999;
}
}
record class ApplicationContext(Guid RequiestId, int UserId);
Commands can accept ConsoleAppContext
as an argument. This allows using the values processed by filters.
var app = ConsoleApp.Create();
app.UseFilter<AuthenticationFilter>();
app.Add("", (int x, int y, ConsoleAppContext context) =>
{
var appContext = (ApplicationContext)context.State!;
var requestId = appContext.RequiestId;
var userId = appContext.UserId;
Console.WriteLine($"Request:{requestId} User:{userId} Sum:{x + y}");
});
app.Run(args);
ConsoleAppContext
also has a ConsoleAppContext.Arguments
property that allows you to obtain the (string[] args
) passed to Run/RunAsync.
ConsoleAppFilter
is defined as internal
for each project by the Source Generator. Therefore, an additional library is provided for referencing common filter definitions across projects.
PM> Install-Package ConsoleAppFramework.Abstractions
This library includes the following classes:
IArgumentParser<T>
ConsoleAppContext
ConsoleAppFilter
ConsoleAppFilterAttribute<T>
Internally, when referencing ConsoleAppFramework.Abstractions
, the USE_EXTERNAL_CONSOLEAPP_ABSTRACTIONS
compilation symbol is added. This disables the above classes generated by the Source Generator, and prioritizes using the classes within the library.
In general frameworks, filters are dynamically added at runtime, resulting in a variable number of filters. Therefore, they need to be allocated using a dynamic array. In ConsoleAppFramework, the number of filters is statically determined at compile time, eliminating the need for any additional allocations such as arrays or lambda expression captures. The allocation amount is equal to the number of filter classes being used plus 1 (for wrapping the command method), resulting in the shortest execution path.
app.UseFilter<NopFilter>();
app.UseFilter<NopFilter>();
app.UseFilter<NopFilter>();
app.UseFilter<NopFilter>();
app.UseFilter<NopFilter>();
// The above code will generate the following code:
sealed class Command0Invoker(string[] args, Action command) : ConsoleAppFilter(null!)
{
public ConsoleAppFilter BuildFilter()
{
var filter0 = new NopFilter(this);
var filter1 = new NopFilter(filter0);
var filter2 = new NopFilter(filter1);
var filter3 = new NopFilter(filter2);
var filter4 = new NopFilter(filter3);
return filter4;
}
public override Task InvokeAsync(ConsoleAppContext context, CancellationToken cancellationToken)
{
return RunCommand0Async(context.Arguments, args, command, context, cancellationToken);
}
}
When an async Task
completes synchronously, it returns the equivalent of Task.CompletedTask
, so ValueTask
is not necessary.
The execution processing of ConsoleAppFramework
fully supports DI
. When you want to use a logger, read a configuration, or share processing with an ASP.NET project, using Microsoft.Extensions.DependencyInjection
or other DI libraries can make processing convenient.
Lambda expressions passed to Run, class constructors, methods, and filter constructors can inject services obtained from IServiceProvider
. Let's look at a minimal example. Setting any System.IServiceProvider
to ConsoleApp.ServiceProvider
enables DI throughout the system.
// Microsoft.Extensions.DependencyInjection
var services = new ServiceCollection();
services.AddTransient<MyService>();
using var serviceProvider = services.BuildServiceProvider();
// Any DI library can be used as long as it can create an IServiceProvider
ConsoleApp.ServiceProvider = serviceProvider;
// When passing to a lambda expression/method, using [FromServices] indicates that it is passed via DI, not as a parameter
ConsoleApp.Run(args, ([FromServices]MyService service, int x, int y) => Console.WriteLine(x + y));
When passing to a lambda expression or method, the [FromServices]
attribute is used to distinguish it from command parameters. When passing a class, Constructor Injection can be used, resulting in a simpler appearance.
Let's try injecting a logger and enabling output to a file. The libraries used are Microsoft.Extensions.Logging and Cysharp/ZLogger (a high-performance logger built on top of MS.E.Logging).
// Package Import: ZLogger
var services = new ServiceCollection();
services.AddLogging(x =>
{
x.ClearProviders();
x.SetMinimumLevel(LogLevel.Trace);
x.AddZLoggerConsole();
x.AddZLoggerFile("log.txt");
});
using var serviceProvider = services.BuildServiceProvider(); // using for logger flush(important!)
ConsoleApp.ServiceProvider = serviceProvider;
var app = ConsoleApp.Create();
app.Add<MyCommand>();
app.Run(args);
// inject logger to constructor
public class MyCommand(ILogger<MyCommand> logger)
{
[Command("")]
public void Echo(string msg)
{
logger.ZLogInformation($"Message is {msg}");
}
}
ConsoleApp
has replaceable default logging methods ConsoleApp.Log
and ConsoleApp.LogError
used for Help display and exception handling. If using ILogger<T>
, it's better to replace these as well.
using var serviceProvider = services.BuildServiceProvider(); // using for cleanup(important)
ConsoleApp.ServiceProvider = serviceProvider;
// setup ConsoleApp system logger
var logger = serviceProvider.GetRequiredService<ILogger<Program>>();
ConsoleApp.Log = msg => logger.LogInformation(msg);
ConsoleApp.LogError = msg => logger.LogError(msg);
DI can also be effectively used when reading application configuration from appsettings.json
. For example, suppose you have the following JSON file.
{
"Position": {
"Title": "Editor",
"Name": "Joe Smith"
},
"MyKey": "My appsettings.json Value",
"AllowedHosts": "*"
}
Using Microsoft.Extensions.Configuration.Json
, reading, binding, and registering with DI can be done as follows.
// Package Import: Microsoft.Extensions.Configuration.Json
var configuration = new ConfigurationBuilder()
.SetBasePath(Directory.GetCurrentDirectory())
.AddJsonFile("appsettings.json")
.Build();
// Bind to services( Package Import: Microsoft.Extensions.Options.ConfigurationExtensions )
var services = new ServiceCollection();
services.Configure<PositionOptions>(configuration.GetSection("Position"));
using var serviceProvider = services.BuildServiceProvider();
ConsoleApp.ServiceProvider = serviceProvider;
var app = ConsoleApp.Create();
app.Add<MyCommand>();
app.Run(args);
// inject options
public class MyCommand(IOptions<PositionOptions> options)
{
[Command("")]
public void Echo(string msg)
{
ConsoleApp.Log($"Binded Option: {options.Value.Title} {options.Value.Name}");
}
}
public class PositionOptions
{
public string Title { get; set; } = "";
public string Name { get; set; } = "";
}
If you have other applications such as ASP.NET in the entire project and want to use common DI and configuration set up using Microsoft.Extensions.Hosting
, you can share them by setting the IServiceProvider
of IHost
after building.
// Package Import: Microsoft.Extensions.Hosting
var builder = Host.CreateApplicationBuilder(); // don't pass args.
using var host = builder.Build(); // use using for host lifetime
ConsoleApp.ServiceProvider = host.Services; // use host ServiceProvider
ConsoleApp.Run(args, ([FromServices] ILogger<Program> logger) => logger.LogInformation("Hello World!"));
ConsoleAppFramework has its own lifetime management (see the CancellationToken(Gracefully Shutdown) and Timeout section), so Host's Start/Stop is not necessary. However, be sure to use the Host itself.
As it is, the DI scope is not set, but by using a global filter, you can add a scope for each command execution. ConsoleAppFilter
can also inject services via constructor injection, so let's get the IServiceProvider
.
var app = ConsoleApp.Create();
app.UseFilter<ServiceProviderScopeFilter>();
internal class ServiceProviderScopeFilter(IServiceProvider serviceProvider, ConsoleAppFilter next) : ConsoleAppFilter(next)
{
public override async Task InvokeAsync(ConsoleAppContext context, CancellationToken cancellationToken)
{
// create Microsoft.Extensions.DependencyInjection scope
await using var scope = serviceProvider.CreateAsyncScope();
await Next.InvokeAsync(context, cancellationToken);
}
}
However, since the construction of the filters is performed before execution, automatic injection using scopes is only effective for the command body itself.
There are multiple ways to run a CLI application in .NET:
run
is convenient when you want to execute the csproj
directly, such as for starting command tools in CI. build
and publish
are quite similar, so it's possible to discuss them in general terms, but it's a bit difficult to talk about the precise differences. For more details, it's a good idea to check out build
vs publish
-- can they be friends? · Issue #26247 · dotnet/sdk.
Also, to run with Native AOT, please refer to the Native AOT deployment overview. In any case, ConsoleAppFramework thoroughly implements a dependency-free and reflection-free approach, so it shouldn't be an obstacle to execution.
v4 was running on top of Microsoft.Extensions.Hosting
, so build a Host in the same way and set up a ServiceProvider.
using var host = Host.CreateDefaultBuilder().Build(); // use using for host lifetime
ConsoleApp.ServiceProvider = host.ServiceProvider;
var app = ConsoleApp.Create(args); app.Run();
->var app = ConsoleApp.Create(); app.Run(args);
app.AddCommand/AddSubCommand
->app.Add(string commandName)
app.AddRootCommand
->app.Add("")
app.AddCommands<T>
->app.Add<T>
app.AddSubCommands<T>
->app.Add<T>(string commandPath)
app.AddAllCommandType
->NotSupported
(useAdd<T>
manually)[Option(int index)]
->[Argument]
[Option(string shortName, string description)]
->Xml Document Comment
ConsoleAppFilter.Order
->NotSupported
(global -> class -> method declrative order)ConsoleAppOptions.GlobalFilters
->app.UseFilter<T>
ConsoleAppBase
-> injectConsoleAppContext
,CancellationToken
to method
This library is under the MIT License.