d-cmdline

A tool library similar to commanderjs in Dlang to help parse command line arguments and options.

Manly of the content in this library, including API design and interal implementation, is drawn from learning from commanderjs.

Quick Start

Firstly, before using this library, we should install it using dub.

dub add cmdline

And then we just import cmdline so that we can use all the features that the library exposes to developers.

This library has preset program which is a variable with Command type.

The following is a simple command line application that have some simple string mutations including split and join.

module examples.str_util;

import std.stdio;
import std.string;
import cmdline;

void main(string[] argv) {
    program
        .name("str_util")
        .description("CLI to some string utilities")
        .setVersion("0.0.1");

    program.command("split")
        .description("Split a string into substrings and display as an array.")
        .argument!string("<str>", "string to split")
        .option("-s, --seperator <char>", "separator character", ",")
        .action((opts, _str) {
            string sp = opts("seperator").get!string;
            string str = _str.get!string;
            writeln(split(str, sp));
        });

    program.command("join")
        .description("Join the command-arguments into a single string")
        .argument!string("<strs...>", "one or more string")
        .option("-s, --seperator <char>", "separator character", ",")
        .action((in OptsWrap opts, in ArgWrap _strs) {
            string sp = opts("seperator").get!string;
            auto strs = cast(string[]) _strs;
            writeln(strs.join(sp));
        });

    program.parse(argv);
}

And after compiling this command line application, we can run the following command lines to do some string mutations and display the help or version info.

$ str_util -V # show the version 
$ str_util --version # same as above 
$ str_util -h # display the help info 
$ str_util --help # same as above 
$ str_util help # same as above 
$ str_util help split # show the help info of sub command: split 
$ str_util split "everything is by design" -s " " # split a string by " " 
$ str_util split "maly,jacob" # split a string by "," 
$ str_util join "maly" "jacob" # join a strings with "," 
$ str_util join "maly" "jacob" -s " " # join a strings with " "`

By the way, here is another way to achieve this little command line program below, and the main comamnd name is strutil instead of str_util:

module examples.strutil;

import std.stdio;
import std.string;
import cmdline;

struct StrutilResult {
    mixin DESC!("CLI to some string utilities");
    mixin VERSION!("0.0.1");
    mixin SUB_CMD!(SplitResult, JoinResult);
}

struct SplitResult {
    mixin DESC!("Split a string into substrings and display as an array.");
    mixin DEF_ARG!(
        "str", string,
        Desc_d!"string to split",
    );
    mixin DEF_OPT!(
        "separator", string, "-s <char>",
        Desc_d!"separator character",
        Default_d!","
    );
    void action() {
        writeln(split(str.get, separator.get));
    }
}

struct JoinResult {
    mixin DESC!("Join the command-arguments into a single string.");
    mixin DEF_ARG!(
        "strs", string[],
        Desc_d!"one or more string"
    );
    mixin DEF_OPT!(
        "separator", string, "-s <char>",
        Desc_d!"separator character",
        Default_d!","
    );
    void action() {
        writeln(strs.get.join(separator.get));
    }
}

void main(in string[] argv) {
    argv.run!StrutilResult;
}

Here is the documentations in html, and you can read it by running it in browser.

Options

Option is one of most important parts in command line. Through options, we can pass infomation to the cmd-line application to process it, then make some action and output the result you want.

After parsing, the options' final inner values will be stored in key-value type. we can get the values according to their long flag names without --.

Options have manly type, including bool option, value option, variadic option and negative option.

In d-cmdline, the inner value must be among int, double, bool, string and their array type int[], double[] and string[], not supporting bool[], since bool value are often used as a switch. It is not necessary to set bool[] as one of inner stored value type, which is the same case in Argument .

We can use build-in trait function isOptionValueType to determine whether a type is legal option inner value type and use isBaseOptionValue to determine whether a type is base option inner type, that is int, double, bool and string.

    static assert(isOptionValueType!int);
    static assert(isOptionValueType!string);
    static assert(isOptionValueType!(int[]));
    static assert(isOptionValueType!(string[]));

    static assert(isBaseOptionValueType!int);
    static assert(isBaseOptionValueType!double);
    static assert(isBaseOptionValueType!bool);
    static assert(isBaseOptionValueType!string);

    static assert(!isOptionValueType!(bool[]));
    static assert(!isOptionValueType!float);
    static assert(!isBaseOptionValueType!(int[]));

Necessary to remember that the options's flag pattern must include its long flag, while the short flag is optional.

Here are two main way to create options and apply it to command line application.

1. using member functions like Command.option and Command.requireOption

2. using functions like createOption and createNegateOption and then using member function addOption to apply them on a command variable.

Command program = createCommand("program");
program.option("-f, --flag", "description"); // bool option
program.option("-o, --output <dest>", "description", "path-to-dest"); // string option

// variadic option
Option variadic_opt = createOption!int("-a|--array <numbers...>", "desc");
program.addOption(variadic_opt);

// negate option
NegateOption nopt = createNegateOption("-F --no-flag", "disable the `--flag`");
program.addOption(nopt);

bool options

On the command line, bool options are often used to control the behavior of a program.

dotnet tool update dotnet-suggest --verbosity quiet --global

In this example, --global is bool options. Bool arguments that usually default to true if the option is specified on the command line but no value is specified.

When bool flag specified then the value is true, otherwise the value is undefined and cannot be gotten.

Only its related negate option is specified, then the bool option's value is false.

And when related bool option ihas been configured but not specifled, then the bool option's value is true by default.

The following is an example of bool options using d-cmdline

module examples.cheese;
import cmdline;
import std.stdio;

void main(in string[] argv) {
    Command program = createCommand("cheese");
    program.option("--sauce", "Add sauce");
    program.option("--no-sauce", "Remove sauce");

    program.parse(argv);

    OptsWrap opts = program.getOpts();
    bool has_sauce = opts("sauce").get!bool;
    if(has_sauce)
        writeln("Add sauce");
    else
        writeln("Remove sauce");
}

$ cheese # Add sauce
$ cheese --sauce # same as above
$ cheese --no-sauce # Remove sauce

value options

On the command line, the value options are always used to pass value to the program and its value stored in key-value structure, key is its name and value is its inner value.

The flags of value options on command line are like --flag value , -f value, --flag=value, -fvalue.

In d-cmdline value options' inner sotred value type is among int, double and string.

Value options have two type, that is rquired value options and optional required value options. The required value options' flag pattern is like --flag <value-name>, while the optional value options' flag pattern is like --flag [value-name].

For required value options, the inner value of the option must be valid if the flag of the options are found in command line, while the optional value options' ones need not.

And if the optional value option's flag is found while no value be found after it, then the optional value option's inner value would be true, instead of the registered type.

Let's see what the required/optional value options are like:

module examples.greet;

import std.stdio;
import cmdline : program, OptsWrap, ArgWrap;

void main(in string[] argv) {
    program
        .name("greet")
        .description("a simple greeting program")
        .option!string("-p, --person [name]", "the persion you greet to")
        .option!string("-g, --greeting <str>", "the greeting string");

    program.parse(argv);

    OptsWrap opts = program.getOpts();
    ArgWrap raw_person = opts("person");
    bool person_is_bool = raw_person.verifyType!bool;
    string person = raw_person.isValid ?
        person_is_bool ? "guy" : raw_person.get!string : "";
    string greeting = opts("greeting").get!string;
    writefln("Hello %s, %s", person, greeting);
}

$ greet -g "how are you?" # Hello , how are you?
$ greet -g "how are you?" -p # Hello guy, how are you?
$ greet -g "how are you?" -pjack # Hello jack, how are you?

variadic options

On the command line, the variadic options are always used to pass values to the program and its values stored in key-value structure, key is its name and value is stored as an array of the values you pass on command line.

The flags of variadic options on command line are like --flag value1 value2 , -f value1 value2, --flag=value1 --flag=value2, -fvalue1 -fvalue2.

In d-cmdline variadic options' inner sotred value type is among int[], double[] and string[].

Variadic options have two type, that is rquired variadic options and optional required variadic options. The required variadic options' flag pattern is like --flag <value-name...>, while the optional variadic options' flag pattern is like --flag [value-name...].

For required variadic options, the inner value of the option must be valid and its length must not be zero if the flag of the options are found in command line, while the optional value options' ones need not.

And if the optional variadic option's flag is found while no value be found after it, then the optional value option's inner value would be true, instead of the registered type.

Let's see what the required/optional variadic options are like:

module examples.variadic;

import std.stdio;
import std.conv;
import cmdline : program, OptsWrap;


void main(in string[] argv) {
    program
        .name("variadic")
        .description("test the variadic option")
        .option!int("-r, --required <values...>", "")
        .option!(int[])("-o, --optional [values...]", "");

    program.parse(argv);

    OptsWrap opts = program.getOpts();
    auto raw_required = opts("required");
    auto raw_optional = opts("optional");

    string required = raw_required.isValid ?
            raw_required.get!(int[]).to!string : "no required";

    string optional = raw_optional.isValid ?
            raw_optional.verifyType!bool ? true.to!string
            : raw_optional.get!(int[]).to!string : "no optional";

    writefln("required: %s", required);
    writefln("optional: %s", optional);
}

$ variadic -r 12 13 -r14 -r15
# required: [12, 13, 14, 15]
# optional: no optional

$ variadic -r 12 13 -r14 -r15 -o
# required: [12, 13, 14, 15]
# optional: true

$ variadic -r 12 13 -r14 -r15 -o 34
# required: [12, 13, 14, 15]
# optional: [34]

$ variadic -r 12 13 -r14 -r15 -o34
# same as above

$ variadic -o 12 13 -o14 -o15
# required: no required
# optional: [12, 13, 14, 15]

negate options

Negate options are a kind of special option type that don't store any inner value. A negate option is just a switch related to the option that has same name.

Negate options flag is similar to bool options, but their flags start with --no-.

When we add a negate option, a related bool option would be added and be set default value true if this related bool option is not added before, or a related bool option would be set default value true if this related bool option is added before.

If a related value or variadic option is added before, this value or variadic option would be disable, that is being removed from the options list of command, when we add a negate option. Therefore, it is better not to add a option after its related negate option is specified on command line.

Here is an example of the usage of negate options:

module examples.cheese;

import std.stdio;
import cmdline;

void main(in string[] argv) {
    program.name("cheese");
    program.sortOptions();

    program.option("--no-sauce", "Remove sauce");
    program.option("--cheese <flavor>", "cheese flavor", "mozzarella");
    program.option("--no-cheese", "plain with no cheese");

    program.action((opts) {
        string sause_str = opts("sauce") ? "sauce" : "no sauce";
        string cheese_str = opts("cheese").isValid ?
            opts("cheese").get!string ~ " cheese" : "no cheese";
        writefln("You ordered apizza with %s and %s", sause_str, cheese_str);
    });

    program.addHelpText(AddHelpPos.Before, `
Try the following:
    $ cheese
    $ cheese --sauce
    $ cheese --cheese=blue
    $ cheese --no-sauce --no-cheese
    `);

    program.parse(argv);
}

$ cheese
# You ordered apizza with sauce and mozzarella cheese
$ cheese --sauce
# You ordered apizza with sauce and mozzarella cheese
$ cheese --cheese=blue
# You ordered apizza with sauce and blue cheese
$ cheese --no-sauce --no-cheese
# You ordered apizza with no sauce and no cheese

the ways to assign value to option

There are manly other ways to assign value to option except for command line.

The option will decide the final inner value based on the priorities of these way after being initialized. And you can use Command.getOptionValSource to get a enum Source to see the final inner value is obtained from which source/way.

The definition of enum Source is below:

/// the source of the final option value gotten
enum Source {
    /// default value
    None,
    /// from client terminal
    Cli,
    /// from env
    Env,
    /// from impled value by other options
    Imply,
    /// from config file
    Config,
    /// from the value that is set by user using `defaultVal` 
    Default,
    /// from the value that is set by user using `preset`
    Preset
}

I will introduce these ways to you.

Option.defaultVal

We can use Option.defaultVal to set a default value to a option (not negate option).

if the flag of an option does not exist on command line, no amy other way to assign value to it, and its default value is set, then the final inner value would be the default value and its source is Source.Default.

Since setting default value to a option is often to be used, we can use Command.option or Command.requiredOption to add n option and set its default value at the same time.

For value options and variadic options, we cannot set their default value bool type.

Here is an example about the usage of setting default value of options:

module examples.defaultval;

import std.stdio;
import cmdline;


void main(in string[] argv) {
    program
        .name("defaultval")
        .option(
            "-c, --cheese <type>",
            "Add the specified type of cheese",
            "blue" 
            // default value, autodetect the option value type is `string`
    );

    Option sauce_opt = createOption!string("-s, --sauce <intig>");
    sauce_opt.description("Add the specified type of sauce");
    sauce_opt.defaultVal("pepper");

    program.addOption(sauce_opt);

    program.parse(argv);

    auto opts = program.getOpts();
    string cheese = opts("cheese").get!string;
    string sauce = opts("sauce").get!string;
    writefln("You ordered apizza with %s and %s cheese", sauce, cheese);
}

$ defaultval
# You ordered apizza with pepper and blue cheese
$ defaultval -ckabpa
# You ordered apizza with pepper and kabpa cheese
$ default -ckabpa -smint
# You ordered apizza with mint and kabpa cheese 

Option.preset

we can use Option.preset to assign value to an optional option. And if the flag of an optional option exists on command line but not its value next to, the value set by Option.preset would be the final value of this option.

In fact, when we defined or create an optional option, we implicitly call Option.preset(true) to it.

Here is an example:

module examples.preset;

import std.stdio;

import cmdline;

void main(in string[] argv) {
    program.name("preset");

    Option pre_opt = createOption!string("-p,--pre [name]");
    pre_opt.defaultVal("dmd");
    pre_opt.preset("dub");
    program.addOption(pre_opt);

    program.addHelpText(AddHelpPos.Before, `
Try the following:
    $ preset
    $ preset --pre
    $ preset -prdmd
    `);

    program.parse(argv);

    OptsWrap opts = program.getOpts;
    string pre_info = opts("pre").get!string;
    writefln("--pre [%s]", pre_info);
}

$ preset
# --pre [dmd]
$ preset --pre
# --pre [dub]
$ preset -prdmd
# --pre [rdmd]

Option.env

Option.env can assign value to option via environment variable. when we pass a key of an environment variable, this option will automatically parse the environment variable in string type to its inner value type and assign this parsed value.

Here is an example about Option.env:

module examples.env;

import std.stdio;
import std.conv;
import std.process;

import cmdline;

void main(in string[] argv) {
    environment["BAR"] = "env";

    program.name("env");

    Option env_opt = createOption!string("-f, --foo <required-arg>");
    env_opt.env("BAR");

    program.addOption(env_opt);
    program.addHelpText(AddHelpPos.Before, `
Try the following:
    $ env
    $ env -fxx
    `);

    program.parse(argv);

    OptsWrap opts = program.getOpts;
    string foo = opts("foo").get!string;
    auto src = program.getOptionValSource("foo");
    writefln("-f, --foo <%s>, from %s", foo, src.to!string);
}

$ env
# -f, --foo <env>, from Env
$ env -fxx
# -f, --foo <xx>, from Cli

Option.implies

Option.implies can assign value to an option whose name is the same as this member function's first parameter if its function interface is Self implies(T)(string key, T value) or the elements of this member function's first parameter if its function interface is Self implies(string[] names).

we can implicitly decided an option's final value using this function.

Here is an example:

module examples.implies;

import std.stdio;
import std.format;
import std.conv;
import cmdline;

void main(in string[] argv) {
    program.name("implies");
    program.option("--foo", "foo bool to be implied");
    program.option("--bar <int-num>", "bar for int implied", 0);

    auto im_opt = createOption!int("--imply <int-num>", "implier with int num");
    im_opt.choices(0, 1);

    im_opt.implies(["foo", "fox"]);
    im_opt.implies("bar", 13);
    im_opt.implies("baz", false);

    program.addOption(im_opt);

    program.parse(argv);

    OptsWrap opts = program.getOpts;
    auto im_raw = opts("imply");
    auto foo_raw = opts("foo");
    auto fox_raw = opts("fox");
    auto baz_raw = opts("baz");

    string bar_info = format!"set bar <%d>"(opts("bar").get!int);
    string im_info = im_raw.isValid ?
        format!"set imply <%d>"(im_raw.get!int) : ("unset imply");
    string foo_info = foo_raw.isValid ?
        format!"set foo `%s`"(foo_raw.get!bool
                .to!string) : ("unset foo");
    string fox_info = fox_raw.isValid ?
        format!"set fox `%s`"(fox_raw.get!bool
                .to!string) : ("unset fox");
    string baz_info = baz_raw.isValid ?
        format!"set baz %s"(baz_raw.get!bool
                .to!string) : ("unset baz");

    writefln("%s, %s, %s, %s, %s.", bar_info, im_info, foo_info, fox_info, baz_info);
}

$ implies
# set bar <0>, unset imply, unset foo, unset fox, unset baz.
$ implies --imply=1
# set bar <13>, set imply <1>, set foo `true`, set fox `true`, set baz false.

Command.setConfigOption

In fact, we can also obtain values from the configuration file to options.

To use this feature, we must call Command.setConfigOption to your target command.

By default, it will auto-detect the config file on the directory where the command line program binary is and the default config file's name is ${YOUR_PROGRAM_NAME}.config.json.

the configuration file is json file, you can see a example of config file for details.

Here is an example:

module examples.config;

import std.stdio;

import cmdline;

void main(in string[] argv) {

    program
        .name("config")
        .description("test the feature  of config option")
        .setConfigOption
        .arguments!(int, int)("<first> <second>")
        .argumentDesc("first", "the  first num")
        .argumentDesc("second", "the  second num")
        .option("-m|--multi <num>", "the multi num", 12)
        .option("-N, --negate", "decide to negate", false)
        .action((opts, fst, snd) {
            int first = fst.get!int;
            int second = snd.get!int;
            int multi = opts("multi").get!int;
            bool negate = opts("negate").get!bool;
            multi = negate ? -multi : multi;
            writefln("%d * (%d + %d) = %d", multi, first, second,
                multi * (first + second));
        });

    auto farg = program.findArgument("first");
    auto sarg = program.findArgument("second");
    farg.defaultVal(65);
    sarg.defaultVal(35);

    program
        .command("sub")
        .description("sub the two numbers")
        .option("-f, --first <int>", "", 65)
        .option("-s, --second <int>", "", 35)
        .action((opts) {
            int first = opts("first").get!int;
            int second = opts("second").get!int;
            writefln("sub:\t%d - %d = %d", first, second, first - second);
        });

    program.parse(argv);
}

the config file config.config.json is:

{
    "sub": {
        "options": {
            "first": 12,
            "second": 13
        }
    },
    "arguments": [12, 13],
    "options": {
        "multi": 4
    }
}

You can run this exmple program.

$ config
# 4 * (12 + 13) = 100
$ config -N
# -4 * (12 + 13) = -100
$ config -Nm3
# -3 * (12 + 13) = -75
$ config sub
# sub:    12 - 13 = -1 
$ config sub -f34 -s43
# sub:    34 - 43 = -9
$ config sub -f34
# sub:    34 - 13 = 21

The priority of option value source

According to above, we can assign value to an option by manly way from different source. However, what if we assign value to an option by different ways at the same time? Then, what the final value it is?

To cope with this problem, d-cmdline give different priorities to different source.

Roughly, the priority rank is below:

(option flag is found only ? Preset : Cli)
> Env > Config > Imply > Default

And if an option have implied other options' value while its final value's Source 's priority is Imply or lower, the final value of these other options would not take the value that be implied.

Option value filter and parsing behavior

d-cmdline also give some input value checker, if the input value of command line is not satisfied with its checker, then an error will be invoked.

And you can also customize the options' parsing behavior, which will change what the final value is.

Option.choices

we can use it to make sure that the option value is one of what Option.choices specifies is. If the input value is not among what Option.choices, then an error will be invoked. This member function should not be used on bool options.

Option.rangeOf

And if the option's inner valut type is int,double, int[] and double[], we can use Option.rangeOf to make sure the option value is in the range between [min, max], min and max are the two parameters of Option.rangeOf.

If you want to use Option.choices and Option.rangeOf on the same option, you'd be better to use Option.rangeOf before Option.choices, so that what the Option.choices specifies mustt also be in the range between [min, max].

Option.parser, Option.processor, Option.processReducer

These three member functions will change the behavior the parsing of option value.

Option.parser is used to set how to transform the string input value from command line to its inner value type.

The default parsing function are std.conv.to!(string, InnerOptionValueType).

Option.processor is used to set how to process the value parsed by function that is specified by Option.parser.

the default processor function are (T v) => v.

Option.processReducer is used only for variadic option so that we can get its value in its Element Type which is processed by the function that Option.processReducer speicifies.

the default process-reducer function pointer is null. So, if you don't set it, you can get the value of array type.

Here is an example:

module examples.optx;

import std.stdio;
import std.conv;
import cmdline;

void main(in string[] argv) {
    program
        .name("optx")
        .description("test the value trait of options");

    auto range_opt = createOption!int("-r, --rgnum <num>");
    range_opt.rangeOf(0, 9);
    range_opt.defaultVal(0);
    range_opt.parser!((string s) => s.to!int);
    range_opt.processor!((int v) => cast(int) v + 1);

    auto choice_opt = createOption!string("-c, --chstr <str>");
    choice_opt.choices("a", "b", "c", "d", "e", "f", "g", "h");
    choice_opt.defaultVal("a");
    choice_opt.processor!((string v) => v ~ "+");

    auto reduce_opt = createOption!int("-n, --nums <numbers...>");
    reduce_opt.defaultVal(12, 13, 14);
    reduce_opt.processReducer!((int a, int b) => a + b);

    program
        .addOption(range_opt)
        .addOption(choice_opt)
        .addOption(reduce_opt);

    program.parse(argv);

    int range_int = program.findOption("rgnum").get!int;
    string choice_str = program.findOption("chstr").get!string;
    int reduce_int = program.findOption("nums").get!int;

    writefln("%d, %s, %d", range_int, choice_str, reduce_int);
}

$ optx
# 1, a+, 39
$ optx -r9
# 10, a+, 39
$ optx -r9 -ch
# 10, h+, 39
$ optx -r9 -ch -n1
# 10, h+, 1
$ optx -r9 -ch -n1 -n 12 12
# 10, h+, 25

Arguments

....... To Be Continued.