🐚Shell Structure

Explaining the inner workings of all the kernel shells

Kernel shells can be built by implementing two different interfaces and base classes. Why two? Because the shell handler relies on:

BaseShell and IShell: To hold shell type and initialization code
BaseShellInfo and IShellInfo: To hold shell commands and the base shell

Shell Handler

The shell handler, ShellManager, uses the available shell list, which holds the BaseShellInfo abstract class, to manipulate with that shell. That class can be get, depending on the needed type, with the ShellManager.GetShellInfo() function in the ︎Nitrocid.Shell namespace.

The shell handler also contains two properties: CurrentShellType and LastShellType. The former property holds the current shell type, which can be used with the shell management functions. The latter property holds the last shell type, which is usually the shell that you exited. However, there are three cases:

If there are no shells in the shell stack, it returns the primary Shell
If there is only one shell in the stack, it returns the current shell as the last one
If there are two or more shells in the stack, it returns the last shell type

Additionally, when GetLine() is called, it sets Terminaux's reader history to point to the shell's history list. After it's done getting the input, it reverts back to the General history buffer. They are loaded on boot and saved on shutdown or reboot.

You can force a reload on the history by using the loadhistories command across all the shells.

You can manually save the history list for all the shells using the savehistories command.

Base Shell

The BaseShell abstract class, which your shell must override, contains the shell type name (ShellType), the flag to bail from the shell (Bail), and the shell initialization code with the shell arguments (InitializeShell()).

public class YourShell : BaseShell, IShell

The shell initialization code usually waits for the Bail value to become true (the shell requested bailing, usually done by exiting the shell using the exit universal command), as in the below example code.

public override void InitializeShell(params object[] ShellArgs)
{
    while (!Bail)
    {
        // Shell code
    }
}

While it's waiting for this to happen, the shell does what it's programmed to do, but in two conditions:

All shells must call the ShellManager.GetLine() function, which usually is adaptive to your shell type. This is the below example code inside the shell initialization code to illustrate this:

while (!Bail)
{
    ShellManager.GetLine();
}

All shells must also handle both the ThreadInterruptedException, which must set Bail to true, and the general exceptions, which must call continue after dumping the exception to the debugger or to the console. For example, the below example code, inside the InitializeShell() function:

while (!Bail)
{
    try
    {
        ShellManager.GetLine();
    }
    catch (ThreadInterruptedException)
    {
        Flags.CancelRequested = false;
        Bail = true;
    }
    catch (Exception ex)
    {
        DebugWriter.WriteDebugStackTrace(ex);
        TextWriterColor.Write(Translate.DoTranslation("There was an error in the shell.") + CharManager.NewLine + "Error {0}: {1}", true, KernelColorType.Error, ex.GetType().FullName, ex.Message);
        continue;
    }
}

The shell registration is required once you're done implementing the shell and all its required values, which will show you how to implement them in the next three pages. The function responsible for this action is ShellTypeManager.RegisterShell() in the Nitrocid.Shell.ShellBase.Shells namespace.

Be sure to unregister your shell using the UnregisterShell() function, or the shell registry function will not update your BaseShellInfo class in the available shell lists!

Shell Information

Every BaseShell class you create must accompany it with a separate class that implements the BaseShellInfo and IShellInfo classes, as in below:

internal class YourShellInfo : BaseShellInfo, IShellInfo

This is where your commands get together by overriding the Commands variable with the new dictionary containing all your commands, like below (in the UESH shell):

public override List<CommandInfo> Commands => new()
{
    new CommandInfo("adduser", /* Localizable */ "Adds users",
        new[] {
            new CommandArgumentInfo(new[]
            {
                new CommandArgumentPart(true, "username"),
                new CommandArgumentPart(false, "password"),
                new CommandArgumentPart(false, "confirm"),
            }, Array.Empty<SwitchInfo>())
        }, new AddUserCommand(), CommandFlags.Strict),
    (...)
};

In addition, you can override the ShellPresets class with a new dictionary containing all the presets for your shell, like below:

public override Dictionary<string, PromptPresetBase> ShellPresets => new()
{
    { "Default", new DefaultPreset() }
};

ShellBase, however, must be overridden with an instance of your shell in this form:

public override BaseShell ShellBase => new YourShell();

Additionally, CurrentPreset must be overridden with a variable that queries your shell type with the CurrentPresets variable as in below:

public override PromptPresetBase CurrentPreset => PromptPresetManager.CurrentPresets[ShellType];

The ShellType variable found within the BaseShellInfo class is a wrapper for the ShellBase.ShellType variable for easier access. It's not overridable and is defined like this:

public string ShellType => ShellBase.ShellType;

By default, your shells don't accept network connections. To make them accept network connections, you must override the AcceptsNetworkConnection so that it holds the value of true instead of false. This causes the network connection selector, especially OpenConnectionForShell() which can be invoked in your networked shell launch code in your command class, to be able to acknowledge your shell.

public override bool AcceptsNetworkConnection => true;

By default, all the shells provide you a multi-line prompt, but if you want your input to be in one line wrapped mode, you can override the below property:

public override bool OneLineWrap => false;

If your shell meets the following conditions:

You need to handle written text in a way, and
You need to use your commands with a slash character, just like the remote debugger,

Then, you need to override the two properties in order for your special non-slash handler to execute:

public override bool SlashCommand => true;
public override CommandInfo NonSlashCommandInfo => new CommandInfo(...);

If you need to know how to define a command information class, consult the below link:

Command Information

You'll have to adapt your shell to take the first argument, ShellArgs[0], as the network connection instance in your Shell instance. For example, we've done this to the FTP shell and shell info instances:

FTPShell.cs

public override void InitializeShell(params object[] ShellArgs)
{
    // Parse shell arguments
    NetworkConnection ftpConnection = (NetworkConnection)ShellArgs[0];
    FtpClient clientFTP = (FtpClient)ftpConnection.ConnectionInstance;

    // Finalize current connection
    FTPShellCommon.clientConnection = ftpConnection;

FTPShellInfo.cs

internal class FTPShellInfo : BaseShellInfo, IShellInfo
{
    (...)
    public override bool AcceptsNetworkConnection => true;
}

Base Command

The base command is required to be implemented, since it contains overridable command execution code. Your command must implement the command base class below:

class YourCommand : BaseCommand, ICommand

The only function that you need to override is Execute(), which you can override like below:

public override int Execute(CommandParameters parameters, ref string variableValue)

To support dumb consoles that don't support positioning or complex console functions, you can override ExecuteDumb():

public override int ExecuteDumb(CommandParameters parameters, ref string variableValue)

Additionally, you can override the extra help function, HelpHelper(), like this:

public override void HelpHelper()

If you want to support redirection or wrapping, you must either take dumb console support into account on the Execute() function by not calling any of the below console wrappers, or you must override the ExecuteDumb() function shown above to be compatible with the dumb consoles.

The following wrappers should not be called (explicitly and implicitly) on that function:

CursorLeft (set)
CursorTop (set)
ForegroundColor
BackgroundColor
CursorVisible
OutputEncoding
InputEncoding
KeyAvailable
Beep()
Clear()
OpenStandardError()
OpenStandardInput()
OpenStandardOutput()
ReadKey()
ResetColor()
SetCursorPosition()
SetOut()

Registering your command

In order for your command to be usable, your mods are now required to register the commands manually using a function that helps doing this. That function is defined in the CommandManager class.

CommandManager.cs

public static void RegisterCustomCommand(ShellType ShellType, CommandInfo commandBase)
public static void RegisterCustomCommand(string ShellType, CommandInfo commandBase)
public static void RegisterCustomCommands(ShellType ShellType, CommandInfo[] commandBases)
public static void RegisterCustomCommands(string ShellType, CommandInfo[] commandBases)

Similarly, if your mod is going to stop, you must unregister all your mod commands, including those that it created in the middle of the kernel uptime. You can use the following functions:

CommandManager.cs

public static void UnregisterCustomCommand(ShellType ShellType, string commandName)
public static void UnregisterCustomCommand(string ShellType, string commandName)
public static void UnregisterCustomCommands(ShellType ShellType, string[] commandNames)
public static void UnregisterCustomCommands(string ShellType, string[] commandNames)

If you've registered your commands correctly, the help command list should list your mod command that you've registered using one of the RegisterCustomCommand functions.

If you're migrating your mods, your mod code can still contain the old Commands list, but you must use the following properties:

Commands.Keys (command names for unregistration)
Commands.Values (command info instances for registration)

Since Nitrocid no longer uses this list, we recommend that you use it as a mutable list of commands just for your mods, since you could be generating commands and registering them.

Setting command values

There is a special switch called set that allows your command to set the final variable value to any value. For example, if you run calc with the -set switch to a variable called result, that variable will be set to an output value (in this case an arithmetic result) using the variableValue argument.

To take advantage of the feature, just write the following code at the end of Execute():

variableValue = myValue;

...where myValue is a string representation of the resulting value that the command produces. A real-world example of this is provided (from the echo command code):

string result = PlaceParse.ProbePlaces(StringArgs);
TextWriterColor.Write(result);
variableValue = result;

Return codes

Your commands all feature return codes. The return code is zero by default, which means that the command has executed successfully. In case of a failure, some commands may return numbers other than zero, which indicate that there is something wrong when executing a command, possibly due to either a failed operation, a general error, or some other error.

The kernel exceptions can also be used as return codes, though you'll have to reference to a class, called KernelExceptionTools, to be able to get an error code by exception type using the GetErrorCode() function. In addition to that, you can also use it with an instance of a kernel exception instance, in case you've wrapped the code with the try...catch clause and that you're catching all the KernelException errors, which almost all APIs throw on either a failure condition or an invalid operation.

All kernel exception codes consist of 10000 added with the enumeration number of the kernel exception type.

Here's a minimal example of what exit does when you're trying to exit the mother shell:

public override int Execute(CommandParameters parameters, ref string variableValue)
{
    if (ShellManager.IsOnMotherShell())
    {
        TextWriters.Write(Translate.DoTranslation("You can't exit the mother shell. Did you mean to log out of your account, shut the kernel down, or reboot it?"), KernelColorType.Error);
        return KernelExceptionTools.GetErrorCode(KernelExceptionType.ShellOperation);
    }
    ShellManager.KillShell();
    return 0;
}

Command flags

Finally, the command flags (CommandFlags) can be defined. One or more of the command flags can be defined using the OR (|) operator when defining the command flags. These flags are available:

Strict: The command is strict, meaning that it's only available for administrators.
- The flag value is 1
NoMaintenance: This command can't run in maintenance mode.
- The flag value is 2
Obsolete: The command is obsolete.
- The flag value is 4
RedirectionSupported: Redirection is supported, meaning that all the output to the commands can be redirected to a file.
- The flag value is 8
Wrappable: This command is wrappable to pages.
- The flag value is 16