Jeff Wofford | July 2014 | e: biz@jeffwofford.com | w: http://www.jeffwofford.com

arrrgh is a fast, small, simple, powerful, single-header library for parsing command line arguments in C++ using more-or-less POSIX parsing rules. It is written using modern C++11 paradigms so it's safe, fast, and tends to invite clean, small, to-the-point code.

There are lots of other command line parsers out there. Yet every time I wrote a new command line program, I ended up rolling my own underpowered and error-prone argument parsing. I like TCLAP but it's a full library with makefiles, lots of source and headers, and the rest. Likewise for getopt, boost, and others: the command line parsing code can end up vastly outweighing the program it's meant to serve. Plus, integrating headers and libraries into C/C++ programs is never a lot of fun, especially when porting to different IDEs and compilers. The Lean Mean Option Parser is nice and small but I didn't prefer its style. You can disagree.

I wrote arrrgh to escape the madness. It's a single file. You copy it into your project. You include it. You set up the arguments you want to support. It does the rest.

1. Include the header:

   #include "arrrgh.hpp" 

   There's no .lib or .so or .cpp or anything else to muck around with. It's all here.

2. Create a parser object and give it your program name and description (for the "usage" output):

   arrrgh::parser parser( "<my-prog>", "<description>" ); 

3. Add your arguments, templated on the desired type. Example:

   const auto& myArgument = parser.add< float >( ... ); 

4. In the add() function you'll indicate the long-form ("--example") and short-form ("-e") switches for this argument. You can nullify one or the other if you want. You should specify a description (the second argument). You may also specify whether the argument is optional or required, and what the default value is (in case the user doesn't supply it):

   ... parser.add< float >(	"<long-form>",	// Use "example" if you want the user to say "--example",	// or leave blank for no long-form.	"<description>",	// For "usage" output.	'<short-form>',	// A character. Use '\0' for no short-form.	[arrrgh::Optional|arrrgh::Required], // To indicate whether it's required.	[default-value]	// The default value. Defaults to 0 or its equivalent for this type.	); 

5. Let 'er rip:

   parser.parse( argc, argv ); 

   Use try...catch... if you want to catch problems in a healthy way.

6. Access argument values:

   myArgument.value()	// Returns a float if we templated on <float>. 

   try...catch... helps here too because users may give invalid strings for non-string types.

7. Call parser.show_usage() if you want to print help text.

8. Do other stuff. See "Example usage" below for "unlabeled" arguments, the "--" marker, and such.

#include "arrrgh.hpp" int main( int argc, const char* argv[] ) {	arrrgh::parser parser( "arrrghsample", "Parses a mix of many argument types and combinations." );	const auto& useAscii = parser.add< bool >( "ascii", "Use ASCII instead of that UNICORN thing or whatever it is.", 'a', arrrgh::Optional, true /* defaults to true */ );	const auto& runFast = parser.add< bool >( "fast", "Should this program run fast?", 'f' );	const auto& doCalibrate = parser.add< bool >( "calibrate", "Calibrate sensors." /* no short-form */ );	const auto& kindaOdd = parser.add< bool >( "", "No long-form argument for this one.", 'o' );	const auto& mass = parser.add< float >( "mass", "The mass of the thing.", 'm', arrrgh::Optional, 3.141f );	const auto& mana = parser.add< float >( "mana", "The mana of the thing.", 'M' );	const auto& height = parser.add< int >( "height", "The height of the thing.", 'h', arrrgh::Required );	const auto& name = parser.add< std::string >( "name", "The name of the wind.", 's', arrrgh::Required );	// Unleash the hounds.	//	try	{	// Example command line:	//	arrrghsample --height=16.25 -fo unlabeled --name="Absolom, Absolom" -h=8 -- --weirdly-unlabeled	parser.parse( argc, argv );	}	catch( const std::exception& e )	{	std::cerr << "Error parsing arguments: " << e.what() << std::endl;	parser.show_usage( std::cerr );	exit( 1 );	}	// Get argument values.	//	try	{	std::cout << std::boolalpha;	// So that bool values come through as "true/false" rather than "1/0".	std::cout << "useAscii=" << useAscii.value() << std::endl;	std::cout << "runFast=" << runFast.value() << std::endl;	std::cout << "doCalibrate=" << doCalibrate.value() << std::endl;	std::cout << "kindaOdd=" << kindaOdd.value() << std::endl;	std::cout << "mass=" << mass.value() << std::endl;	std::cout << "mana=" << mana.value() << std::endl;	std::cout << "height=" << height.value() << std::endl;	std::cout << "name=" << name.value() << std::endl;	// What about unlabeled arguments?	//	// Notice that "--weirdly-unlabeled" looks like a switch but comes through nicely as unlabeled.	// That's because the example command line indicated "--" before it, which ends switch parsing	// (everything else becomes unlabeled).	//	std::cout << "Unlabeled:\n";	parser.each_unlabeled_argument( []( const std::string& arg ) { std::cout << "\t" << arg << std::endl; } );	}	catch( const std::exception& e )	{	std::cerr << "Error reading argument values: " << e.what() << std::endl;	}	return 0; } 

Released under the MIT license. See arrrgh.hpp.