other language: 中文简体
Another Comptime-argparse for Zig! Let's start to build your command line!
const std = @import("std");
const zargs = @import("zargs");
const Command = zargs.Command;
const Arg = zargs.Arg;
const Ranges = zargs.Ranges;
const ztype = @import("ztype");
const String = ztype.String;
pub fn main() !void {
// Like Py3 argparse, https://docs.python.org/3.13/library/argparse.html
const remove = Command.new("remove")
.about("Remove something")
.alias("rm").alias("uninstall").alias("del")
.opt("verbose", u32, .{ .short = 'v' })
.optArg("count", u32, .{ .short = 'c', .argName = "CNT", .default = 9 })
.posArg("name", String, .{});
// Like Rust clap, https://docs.rs/clap/latest/clap/
const cmd = Command.new("demo").requireSub("action")
.about("This is a demo intended to be showcased in the README.")
.author("KiozWang")
.homepage("https://github.com/kioz-wang/zargs")
.arg(Arg.opt("verbose", u32).short('v').help("help of verbose"))
.arg(Arg.optArg("logfile", ?ztype.OpenLazy(.fileCreate, .{ .read = true })).long("log").help("Store log into a file"))
.sub(Command.new("install")
.about("Install something")
.arg(Arg.optArg("count", u32).default(10)
.short('c').short('n').short('t')
.long("count").long("cnt")
.ranges(Ranges(u32).new().u(5, 7).u(13, null)).choices(&.{ 10, 11 }))
.arg(Arg.posArg("name", String).rawChoices(&.{ "gcc", "clang" }))
.arg(Arg.optArg("output", String).short('o').long("out"))
.arg(Arg.optArg("vector", ?@Vector(3, i32)).long("vec")))
.sub(remove);
var gpa: std.heap.GeneralPurposeAllocator(.{}) = .init;
const allocator = gpa.allocator();
var args = cmd.config(.{ .style = .classic }).parse(allocator) catch |e|
zargs.exitf(e, 1, "\n{s}\n", .{cmd.usageString()});
defer cmd.destroy(&args, allocator);
if (args.logfile) |logfile| std.debug.print("Store log into {}\n", .{logfile});
switch (args.action) {
.install => |a| {
std.debug.print("Installing {s}\n", .{a.name});
},
.remove => |a| {
std.debug.print("Removing {s}\n", .{a.name});
std.debug.print("{any}\n", .{a});
},
}
std.debug.print("Success to do {s}\n", .{@tagName(args.action)});
}As a system level programming language, there should be an elegant solution for parsing command line arguments.
zargs draws inspiration from the API styles of Py3 argparse and Rust clap. It provides all parameter information during editing, reflects the parameter structure and parser at compile time, along with everything else needed, and supports dynamic memory allocation for parameters at runtime.
Get the latest version:
zig fetch --save git+https://github.com/kioz-wang/zargsTo fetch a specific version (e.g., v0.14.3):
zig fetch --save https://github.com/kioz-wang/zargs/archive/refs/tags/v0.14.3.tar.gzThe version number follows the format vx.y.z[-alpha.n]:
- x: Currently fixed at 0. It will increment to 1 when the project stabilizes. Afterward, it will increment by 1 for any breaking changes.
- y: Represents the supported Zig version. For example,
vx.14.zsupports Zig 0.14.0. - z: Iteration version, indicating releases with new features or significant changes (see milestones).
- n: Minor version, indicating releases with fixes or minor updates.
In your build.zig, use addImport (for example):
const exe = b.addExecutable(.{
.name = "your_app",
.root_source_file = b.path("src/main.zig"),
.target = b.standardTargetOptions(.{}),
.optimize = b.standardOptimizeOption(.{}),
});
exe.root_module.addImport("zargs", b.dependency("zargs", .{}).module("zargs"));
b.installArtifact(exe);
const run_cmd = b.addRunArtifact(exe);
run_cmd.step.dependOn(b.getInstallStep());
if (b.args) |args| {
run_cmd.addArgs(args);
}
const run_step = b.step("run", "Run the app");
run_step.dependOn(&run_cmd.step);After importing in your source code, you will gain access to the following features:
- Command and argument builders:
Command,Arg - Versatile iterator support:
TokenIter - Convenient exit functions:
exit,exitf
See the documentation for details.
const zargs = @import("zargs");In addition to the core module zargs, I also exported the fmt and par modules.
any, which provides a more flexible and powerful formatting scheme.
stringify, if a class contains a method such as fname(self, writer), then you can obtain a compile-time string like this:
pub fn getString(self: Self) *const [stringify(self, “fname”).count():0]u8 {
return stringify(self, “fname”).literal();
}comptimeUpperString converts a compile-time string to uppercase.
any, parses the string into any type instance you want.
For struct, you need to implement pub fn parse(s: String, a_maybe: ?Allocator) ?Self. For enum, the default parser is std.meta.stringToEnum, but if parse is implemented, it will be used instead.
destroy, releases the parsed type instance.
Safe release: for instances where no memory allocation occurred during parsing, no actual release action is performed. For struct and enum, actual release actions are performed only when pub fn destroy(self: Self, a: Allocator) void is implemented.
Provides String, LiteralString and checker.
Provides the wrappers of some struct in std:
Open/OpenLazy(...):std.fs.File/Dir- See ex-06 for usage
...
- Option (
opt)- Single Option (
singleOpt)- Boolean Option (
boolOpt),T == bool - Accumulative Option (
repeatOpt),@typeInfo(T) == .int
- Boolean Option (
- Option with Argument (
argOpt)- Option with Single Argument (
singleArgOpt), T,?T - Option with Fixed Number of Arguments (
arrayArgOpt),[n]T - Option with Variable Number of Arguments (
multiArgOpt),[]T
- Option with Single Argument (
- Single Option (
- Argument (
arg)- Option Argument (
optArg) (equivalent to Option with Argument) - Positional Argument (
posArg)- Single Positional Argument (
singlePosArg), T,?T - Fixed Number of Positional Arguments (
arrayPosArg),[n]T
- Single Positional Argument (
- Option Argument (
- Subcommand (
subCmd)
Matching and parsing are driven by an iterator. For options, the option is always matched first, and if it takes an argument, the argument is then parsed. For positional arguments, parsing is attempted directly.
For arguments, T must be the smallest parsable unit: []const u8 -> T
.int.float.booltrue: 'y', 't', "yes", "true" (case insensitive)false: 'n', 'f', "no", "false" (case insensitive)
.enum: Usesstd.meta.stringToEnumby default, butparsemethod takes priority.struct: Struct withparsemethod.vector- Only supports base types of
.int,.float, and.bool @Vector{1,1}:[\(\[\{][ ]*1[ ]*[;:,][ ]*1[ ]*[\)\]\}]@Vector{true,false}:[\(\[\{][ ]*y[ ]*[;:,][ ]*no[ ]*[\)\]\}]
- Only supports base types of
If type T has no associated default parser or parse method, you can specify a custom parser (.parseFn) for the parameter. Obviously, single-option parameters cannot have parsers as it would be meaningless.
Options and arguments can be configured with default values (.default). Once configured, the option or argument becomes optional.
- Even if not explicitly configured, single options always have default values: boolean options default to
false, and accumulative options default to0. - Options or arguments with an optional type
?Tcannot be explicitly configured: they are forced to default tonull.
Single options, options with a single argument of optional type, or single positional arguments of optional type are always optional.
Default values must be determined at comptime. For argOpt and posArg, if the value cannot be determined at comptime (e.g., std.fs.cwd() at Windows), you can configure the default input (.rawDefault), which will determine the default value in the perser.
Value ranges (.ranges, .choices) can be configured for arguments, which are validated after parsing.
Default values are not validated (intentional feature? 😄)
If constructing value ranges is cumbersome, .rawChoices can be used to filter values before parsing.
When T implements compare, value .ranges can be configured for the argument.
Choices
See helper.Compare.compare
When T implements equal, value .choices can be configured for the argument.
See helper.Compare.equal
A callback (.callbackFn) can be configured, which will be executed after matching and parsing.
A command cannot have both positional arguments and subcommands simultaneously.
For the parser, except for accumulative options and options with a variable number of arguments, no option can appear more than once.
Options are further divided into short options and long options:
- Short Option:
-v - Long Option:
--verbose
Options with a single argument can use a connector to link the option and the argument:
- Short Option:
-o=hello,-o hello - Long Option:
--output=hello,--output hello
Dropping the connector or whitespace for short options is not allowed, as it results in poor readability!
For options with a fixed number of arguments, connectors cannot be used, and all arguments must be provided at once. For example, with a long option:
--files f0 f1 f2 # [3]const TOptions with a variable number of arguments are similar to options with a single argument but can appear multiple times, e.g.:
--file f0 -v --file=f1 --greet # []const TMultiple short options can share a prefix, but if an option takes an argument, it must be placed last, e.g.:
-Rns
-Rnso hello
-Rnso=helloOnce a positional argument appears, the parser informs the iterator to only return positional arguments, even if the arguments might have an option prefix, e.g.:
-o hello a b -v # -o is an option with a single argument, so a, b, -v are all positional argumentsAn option terminator can be used to inform the iterator to only return positional arguments, e.g.:
--output hello -- a b -vDouble quotes can be used to avoid iterator ambiguity, e.g., to pass a negative number -1, double quotes must be used:
--num \"-1\"Since the shell removes double quotes, escape characters are also required! If a connector is used, escaping is unnecessary:
--num="-1".
As shown in the example at the beginning of the article, command construction can be completed in a single line of code through chaining.
Of course, if needed, you can also build it step by step. Simply declare it as comptime var cmd = Command.new(...).
const install = Command.new("install");
const _demo = Command.new("demo").requireSub("action")
.sub(install.callBack(struct {
fn f(_: *install.Result()) void {
std.debug.print("CallBack of {s}\n", .{install.name});
}
}.f));
const demo = _demo.callBack(struct {
fn f(_: *_demo.Result()) void {
std.debug.print("CallBack of {s}\n", .{_demo.name});
}
}.f);const args = try cmd.parse(allocator);
defer cmd.destroy(&args, allocator);Simply call parse to generate the parser and argument structure. This method internally creates a system iterator, which is destroyed after use.
Additionally, parseFrom supports passing a custom iterator and optionally avoids using a memory allocator. If no allocator is used, there is no need to defer destroy.
When the parser has completed its task, if you still need to handle the remaining arguments manually, you can call the iterator's nextAllBase method.
If further parsing of the arguments is required, you can use the parseAny function.
Flexible for real and test scenarios
- System iterator (
init): get real command line arguments. - General iterator (
initGeneral): splits command line arguments from a one-line string. - Line iterator (
initLine): same as regular iterator, but you can specify delimiters. - List iterator (
initList): iterates over a list of strings.
Short option prefixes (-), long option prefixes (--), connectors (=), option terminators (--) can be customized for iterators (see ex-05).
_ = cmd.usageString();
_ = cmd.helpString();See https://kioz-wang.github.io/zargs/#doc
Look at here
To build all examples:
zig build examplesTo list all examples (all step prefixed ex- are examples):
zig build -lTo execute an example:
zig build ex-01.add -- -hWelcome to submit PRs to link your project that use
zargs!
More real-world examples are coming!
MIT © Kioz Wang
