This is a library for parsing command-line arguments. It can parse both options and positional arguments. It aims to be easy to use and concise yet powerful and robust.
Overview
Defining options and positional arguments is as simple as:
import com.xenomachina.argparser.ArgParser
class MyArgs(parser: ArgParser) {
val v by parser.flagging("enable verbose mode")
val name by parser.storing("name of the user")
val count by parser.storing("number of the widgets") { toInt() }
val source by parser.positional("source filename")
val destination by parser.positional("destination filename")
}
An instance of MyArgs will represent the set of parsed arguments. Each option and positional argument is declared as a property that delegates through a delegate factory method on an instance of ArgParser.
The name of an option is inferred from the name of the property it is bound to. The options above are named -v, --name and --count, respectively. There are also two positional arguments.
Direct control over an option's name is also possible, and for most types of options it is also possible to have multiple names, like a short and long name:
class MyArgs(parser: ArgParser) {
val verbose by parser.flagging(
"-v", "--verbose",
help = "enable verbose mode")
val name by parser.storing(
"-N", "--name",
help = "name of the user")
val count by parser.storing(
"-c", "--count",
help = "number of widgets") { toInt() }
val source by parser.positional(
"SOURCE",
help = "source filename")
val destination by parser.positional(
"DEST",
help = "destination filename")
}
The unparsed command-line arguments are passed to the ArgParser instance at construction:
fun main(args: Array<String>) = mainBody {
ArgParser(args).parseInto(::MyArgs).run {
println("Hello, ${name}!")
println("I'm going to move ${count} widgets from ${source} to ${destination}.")
// TODO: move widgets
}
}
See kotlin-argparser-example for a complete example project.
Nomenclature
Options, arguments, flags... what's the difference?
An application's main function is passed an array of strings. These are the unparsed command-line arguments, or unparsed arguments for short.
The unparsed arguments can then be parsed into options, which start with a hyphen ("-"), and positional arguments. For example, in the command ls -l /tmp/, the unparsed arguments would be "-l", "/tmp" where -l is an option, while /tmp/ is a positional argument.
Options can also have option arguments. In the command ls --time-style=iso, the option is --time-style and that options argument is iso. Note that in parsing a single unparsed argument can be split into an option and an option argument, or even into multiple options in some cases.
A flag is a boolean option which has no arguments and which is false if not provided, but true if provided. The -l option of ls is a flag.
Option Types
Boolean Flags
Boolean flags are created by asking the parser for a flagging delegate. One or more option names, may be provided:
val verbose by parser.flagging("-v", "--verbose",
help = "enable verbose mode")
Here the presence of either -v or --verbose options in the arguments will cause the Boolean property verbose to be true, otherwise it will be false.
Storing a Single Argument
Single argument options are created by asking the parser for a storing delegate.
val name by parser.storing("-N", "--name",
help = "name of the user")
Here either -N or --name with an argument will cause the name property to have that argument as its value.
A function can also be supplied to transform the argument into the desired type. Here the size property will be an Int rather than a String:
val size by parser.storing("-c", "--count",
help = "number of widgets") { toInt() }
Adding to a Collection
Options that add to a Collection each time they appear in the arguments are created with using the adding delegate. Just like storing delegates, a transform function may optionally be supplied:
val includeDirs by parser.adding(
"-I", help = "directory to search for header files") { File(this) }
Now each time the -I option appears, its transformed argument is appended to includeDirs.
Mapping from an option to a fixed value
For choosing between a fixed set of values (typically, but not necessarily, from an enum), a mapping delegate can be used:
val mode by parser.mapping(
"--fast" to Mode.FAST,
"--small" to Mode.SMALL,
"--quiet" to Mode.QUIET,
help = "mode of operation")
Here the mode property will be set to the corresponding ArgParser.Mode value depending on which of --fast, --small, and --quiet appears (last) in the arguments.
mapping is one of the few cases where it is not possible to infer the option name from the property name.
More advanced options
For all other types of options, the option method should be used. The methods mentioned above are, in fact, convenience methods built on top of the option method.
For example, it is possible to create an option that has multiple arguments:
fun ArgParser.putting(vararg names: String, help: String) =
option<MutableMap<String, String>>(*names,
argNames = listOf("KEY", "VALUE"),
help = help) {
value.orElse { mutableMapOf<String, String>() }.apply {
put(arguments.first(), arguments.last()) }
}
Note that the option method does not have an auto-naming overload. If you need this capability, create a DelegateProvider that creates your Delegate:
fun ArgParser.putting(help: String) =
ArgParser.DelegateProvider { identifier ->
putting(identifierToOptionName(identifier), help = help) }
Positional Arguments
Positional arguments are collected by using the positional and positionalList methods.
For a single positional argument:
val destination by parser.positional("destination filename")
An explicit name may also be specified:
val destination by parser.positional("DEST",
help = "destination filename")
The name ("DEST", here) is used in error handling and help text.
For a list of positional arguments:
val sources by parser.positionalList("SOURCE", 1..Int.MAX_VALUE,
help = "source filename")
The range indicates how many arguments should be collected, and defaults to the value shown in this example. As the name suggests, the resulting property will be a List.
Both of these methods accept an optional transform function for converting arguments from String to whatever type is actually desired:
val destination by parser.positional("DEST",
help = "...") { File(this) }
val sources by parser.positionalList("SOURCE", 1..Int.MAX_VALUE,
help = "...") { File(this) }
Modifying Delegates
The delegates returned by any of these methods also have a few methods for setting optional attributes:
Adding a Default Value
Certain types of delegates (notably storing, mapping, and positional) have no default value, and hence will be required options unless a default value is provided. This is done with the default method:
val name by parser.storing("-N", "--name", help = "...").default("John Doe")
Note that it is possible to use null for the default, though this may require specifying the type parameter for default explicitly:
val name by parser.storing("-N", "--name", help = "...").default<String?>(null)
The type of the resulting property be nullable (a String? in this case).
Adding a Validator
Sometimes it's easier to validate an option at the end of parsing, in which case the addValidator method can be used.
val percentages by parser.adding("--percentages", help = "...") { toInt() }
.addValidator {
if (value.sum() != 100)
throw InvalidArgumentException(
"Percentages must add up to 100%")
}
Error Handling
If the parser determines that execution should not continue it will throw a SystemExitException which has a status code appropriate for passing to exitProcess as well as a message for the user.
These exceptions can be caused by user error, or even if the user requests help (eg: via the --help option).
It is recommended that transform functions (given to storing, positionalList, etc.) and post-parsing validation, including that performed via, addValidator also throw a SystemExitException on failure.
As a convenience, these exceptions can be handled by using the mainBody function:
class ParsedArgs(parser: ArgParser) {
val name by positional("The user's name").default("world")
}
fun main(args: Array<String>) = mainBody {
ArgParser(args).parseInto(::ParsedArgs).run {
println("Hello, {name}!")
}
}
Parsing
Parsing of command-line arguments is performed sequentially. So long as option-processing is enabled, each not-yet-processed command-line argument that starts with a hyphen (-) is treated as an option.
Short Options
Short options start with a single hyphen. If the option takes an argument, the argument can either be appended:
# "-o" with argument "ARGUMENT"
my_program -oARGUMENT
or can be the following command-line argument:
# "-o" with argument "ARGUMENT"
my_program -o ARGUMENT
Zero argument short options can also be appended to each other without intermediate hyphens:
# "-x", "-y" and "-z" options
my_program -xyz
An option that accepts arguments is also allowed at the end of such a chain:
# "-x", "-y" and "-z" options, with argument for "-z"
my_program -xyzARGUMENT
Long Options
Long options start with a double hyphen (--). An argument to a long option can either be delimited with an equal sign (=):
# "--foo" with argument "ARGUMENT"
my_program --foo=ARGUMENT
or can be the following command-line argument:
# "--foo" with argument "ARGUMENT"
my_program --foo ARGUMENT
Multi-argument Options
Multi-argument options are supported, though currently not by any of the convenience methods. Option-arguments after the first must be separate command-line arguments, for both an long and short forms of an option.
Positional Arguments
In GNU mode (the default), options can be interspersed with positional arguments, but in POSIX mode the first positional argument that is encountered disables option processing for the remaining arguments. In either mode, if the argument "--" is encountered while option processing is enabled, then option processing is disabled for the rest of the command-line. Once the options and option-arguments have been eliminated, what remains are considered to be positional arguments.
Each positional argument delegate can specify a minimum and maximum number of arguments it is willing to collect.
The positional arguments are distributed to the delegates by allocating each positional delegate at least as many arguments as it requires. If more than the minimum number of positional arguments have been supplied then additional arguments will be allocated to the first delegate up to its maximum, then the second, and so on, until all arguments have been allocated to a delegate.
This makes it easy to create a program that behaves like grep:
class Args(parser: ArgParser) {
// accept 1 regex followed by n filenames
val regex by parser.positional("REGEX",
help = "regular expression to search for")
val files by parser.positionalList("FILE",
help = "file to search in")
}
And equally easy to create a program that behaves like cp:
class Args(parser: ArgParser) {
// accept n source files followed by 1 destination
val sources by parser.positionalList("SOURCE",
help = "source file")
val destination by parser.positional("DEST",
help = "destination file")
}
Forcing Parsing
Parsing normally does not begin until a delegate's value is accessed. Sometimes this is not desirable, so it is possible to enforce the parsing of arguments into a class of values. This ensures that all arguments that are required are provided, and all arguments provided are consumed.
Forcing can be done in a separate step using the force method:
val parser = ArgParser(args)
val parsedArgs = ParsedArgs(parser)
parser.force()
// now you can use parsedArgs
Alternatively, forcing can be done inline via the parseInto method:
val parsedArgs = ArgParser(args).parseInto(::ParsedArgs)
// now you can use parsedArgs
In both cases exceptions will be thrown where parsing or validation errors are found.
Help Formatting
By default, ArgParser will add a --help option (short name -h) for displaying usage information. If this option is present a ShowHelpException will be thrown. If the default exception handling is being used (see Error Handling) the program will halt and print a help message like the one below, based on the ArgParser configuration:
usage: program_name [-h] [-n] [-I INCLUDE]... -o OUTPUT
[-v]... SOURCE... DEST
This is the prologue. Lorem ipsum dolor sit amet, consectetur
adipiscing elit. Aliquam malesuada maximus eros. Fusce
luctus risus eget quam consectetur, eu auctor est
ullamcorper. Maecenas eget suscipit dui, sed sodales erat.
Phasellus.
required arguments:
-o OUTPUT, directory in which all output should
--output OUTPUT be generated
optional arguments:
-h, --help show this help message and exit
-n, --dry-run don't do anything
-I INCLUDE, search in this directory for header
--include INCLUDE files
-v, --verbose increase verbosity
positional arguments:
SOURCE source file
DEST destination file
This is the epilogue. Lorem ipsum dolor sit amet,
consectetur adipiscing elit. Donec vel tortor nunc. Sed eu
massa sed turpis auctor faucibus. Donec vel pellentesque
tortor. Ut ultrices tempus lectus fermentum vestibulum.
Phasellus.
The creation of the --help option can be disabled by passing null as the helpFormatter when constructing the ArgParser, or configured by manually constructing a HelpFormatter instance. In the above example a DefaultHelpFormatter was created with the prologue and epilogue.
Caveats
-
This library should be considered to be very beta. While there are no plans to make any breaking changes to the API, it's possible that there may be some until it is mature.
-
Upon reading the value any of the delegated properties created by an
ArgParser, the arguments used to construct thatArgParserwill be parsed. This means it's important that you don't attempt to create delegates on anArgParserafter any of its existing delegated properties have been read. Attempting to do so will cause anIllegalStateException. It would be nice if Kotlin had facilities for doing some of the work ofArgParserat compile time rather than run time, but so far the run time errors seem to be reasonably easy to avoid.
Configuring Your Build
Kotlin-argparser binaries are hosted on Maven Central and also Bintray's JCenter.
In Gradle, add something like this in your build.gradle:
// you probably already have this part
buildscript {
repositories {
mavenCentral() // or jcenter()
}
}
dependencies {
compile "com.xenomachina:kotlin-argparser:$kotlin_argparser_version"
}
In Maven add something like this to your pom.xml:
<dependency>
<groupId>com.xenomachina</groupId>
<artifactId>kotlin-argparser</artifactId>
<version>VERSION</version>
</dependency>
Information on setting up other build systems, as well as the current version number, can be found on MVN Repository's page for Kotlin-argparser.
Thanks
Thanks to the creators of Python's argparse module, which provided the initial inspiration for this library.
Thanks also to the team behind Kotlin.
Finally, thanks to all of the people who have contributed code and/or issues.
" (hyphen immediately followed by a number)">
2 Aug 30, 2022
3.8k Jan 3, 2023
367 Nov 22, 2022
447 Dec 26, 2022
1 Jun 28, 2022
25 Nov 16, 2021
7 Jun 8, 2022
439 Dec 29, 2022
509 Dec 18, 2022
4.8k Jan 3, 2023
191 Dec 21, 2022
29 Dec 26, 2022
2 Oct 28, 2021
3 Sep 13, 2022
0 Dec 9, 2021
0 Dec 27, 2021
157 Nov 13, 2022
15 Dec 23, 2022