Repository

Command-line Arguments

Parses command-line arguments to statically typed options or a string map with the help of usage description.

  • Legible configuration inferred from the usage instructions text.
  • Full control over the usage text.
  • Checks for unknown and required options, invalid value types and arithmetic overflow.
  • Compatibility with the getopt and getopt_long standards.

Synopsis

Specify usage description and version of the command-line tool. Declare a structure with all command-line options. Import the command-line parser and parse the options and arguments:

import prantlf.cargs { parse, Input }

// Describe usage of the command-line tool.
usage := 'Converts YAML input to JSON output.

Usage: yaml2json [options] [<yaml-file>]

  <yaml-file>         read the YAML input from a file

Options:
  -o|--output <file>   write the JSON output to a file
  -i|--indent <count>  write the JSON output to a file
  -p|--pretty          print the JSON output with line breaks and indented
  -V|--version         print the version of the executable and exit
  -h|--help            print the usage information and exit

If no input file is specified, it will be read from standard input.

Examples:
  $ yaml2json config.yaml -o config.json -p
  $ cat config.yaml | yaml2json > config.json'

// Declare a structure with all command-line options.
struct Opts {
  output string
  indent int
  pretty bool
}

// Parse command-line options and arguments.
opts, args := parse[Opts](usage, Input{ version: '0.0.1' })!
if args.len > 0 {
  // Process file names from the args array.
} else {
  // Read from the standard input.
}

Installation

You can install this package either from VPM or from GitHub:

v install prantlf.cargs
v install --git https://github.com/prantlf/v-cargs

API

The following functions and types are exported:

parse[T](usage string, input Input) !(T, []string)

Parses the command line, separating options and other arguments. Options will be set to the statically-typed structure and other arguments returned as an array of strings. The list of options will be inferred from the usage description.

import prantlf.cargs { parse, Input }

usage := '...

Options:
  -o|--output <file>  write the JSON output to a file
  -p|--pretty         print the JSON output with line breaks and indented

...'

struct Opts {
  output string
  pretty bool
}

opts, args := parse[Opts](usage, Input{ version: '0.0.1' })!

parse_to[T](usage string, input Input, mut opts T) ![]string

Parses the command line, separating options and other arguments, while setting the field values in an already created object. It can be used to override options initially read from configuration file from the command-line or defaults, for example.

import prantlf.cargs { parse_to, Input }

usage := '...

Options:
  -o|--output <file>  write the JSON output to a file
  -p|--pretty         print the JSON output with line breaks and indented

...'

struct Opts {
  output string
  pretty bool
}

mut opts := Opts{ output: 'out.json' }
args := parse_to[Opts](usage, Input{ version: '0.0.1' }, mut opts)!

See prantlf.config for more information.

scan(usage string, input Input) !Scanned

Parses the command line, only analysing the usage description and splitting command-line argument groups. It can be used for splitting the two phases - analysis and command-line argument processing. The argument processing can be finished by parse_scanned or parse_scanned_to . Before that, a single option can be obtained by get_val , for example.

import prantlf.cargs { scan, parse_scanned_to, Input }

usage := '...

Options:
  -o|--output <file>  write the JSON output to a file
  -p|--pretty         print the JSON output with line breaks and indented

...'

struct Opts {
  output string
  pretty bool
}

input := Input{ version: '0.0.1' }
scanned := scan(usage, input)!
...
mut opts := Opts{ output: 'out.json' }
args := parse_scanned_to[Opts](scanned, input, mut opts)!

get_val(scanned Scanned, input Input, arg_name string, def_val string) !string

Partially parses the command line to get only the value of one argument. This argument usually decides about default values or other processing before the other command-line arguments will be parsed.

import prantlf.cargs { scan, get_val, parse_scanned_to, Input }
import prantlf.config { read_config_to }

usage := '...

Options:
  -c|--config <name>  name or path to the configuration file
  -p|--pretty         print the JSON output with line breaks and indented

...'

struct Opts {
  pretty bool
}

input := Input{ version: '0.0.1' }
scanned := scan(usage, input)!
config := get_val(scanned, input, 'config', '')!
mut opts := Opts{}
if config.len > 0 {
  read_config_to(config_name, mut opts)!
}
args := parse_scanned_to(scanned, input, mut opts)!

parse_scanned[T](scanned Scanned, input Input) !(T, []string)

Finishes parsing the command line using previously analysed usage instructions. Calling the combination of scan and parse_scanned is the same as calling parse .

import prantlf.cargs { scan, parse_scanned, Input }

usage := '...

Options:
  -o|--output <file>  write the JSON output to a file
  -p|--pretty         print the JSON output with line breaks and indented

...'

struct Opts {
  output string
  pretty bool
}

input := Input{ version: '0.0.1' }
scanned := scan(usage, input)!
...
opts, args := parse_scanned[Opts](scanned, input)!

parse_scanned_to[T](scanned Scanned, input Input, mut opts T) ![]string

Finishes parsing the command line using previously analysed usage instructions. Calling the combination of scan and parse_scanned_to is the same as calling parse_to .

import prantlf.cargs { scan, parse_scanned_to, Input }

usage := '...

Options:
  -o|--output <file>  write the JSON output to a file
  -p|--pretty         print the JSON output with line breaks and indented

...'

struct Opts {
  output string
  pretty bool
}

input := Input{ version: '0.0.1' }
scanned := scan(usage, input)!
...
mut opts := Opts{ output: 'out.json' }
args := parse_scanned_to[Opts](scanned, input, mut opts)!

Usage Instructions

The usage parameter is the formatted text to be presented as usage instructions. It's supposed to contain a line Starting with Options: , which is followed by links listing the options:

Options:
  -o|--output <file>  write the JSON output to a file
  -p|--pretty         print the JSON output with line breaks and indented

An option-line can contain a short (single-letter) option, a long option or both. The option can be either a boolean flag or an variable with a value.

-p             a boolean flag, short variant only
--line-breaks  a boolean flag, long variant only
-v|--verbose   a boolean flag, both short and long variants
-o <file>      a variable with a value

Short and long option variants can be delimited either by | or by , , which can be followed by a space. A value of a variable can be enclose either in < and > , or in [ and ] .

If a negative option is entered, the field in the options structure has to be still positive - without the no_ prefix. This is usually used to declare flags, which are enabled by default and can be disabled by the negative option:

--no-line-breaks  a boolean flag, long variant only

Options

Two command-line options will be recognised and processed by the parse function itself:

  • -V|--version - print the version of the executable and exit
  • -h|--help - print the usage information and exit

Short (single-letter) options can be condensed together. For example, instead of -l -p , you can write -lp on the command line.

Names of fields in the options structure are inferred from the command-line option names with several changes to ensure valid V syntax:

  • The long variant of an option will be mapped to its field name. The short variant will be used only if the long variant is missing.
  • Upper-case letters will converted to lower-case.
  • Dashes ( - ) in an option name will be converted to underscores ( _ ) in its field name.

If you write a short (single-letter) option for a boolean flag in upper-case, it will set the value false to the boolean field instead of true . If you write a long option for a boolean flag, you can negate its value by prefixing the option with no- :

-P --no-line-breaks

Enum field types can be filled either by an integer or by the (string) name of the enum value.

Assigning boolean flags or variable values to option fields may fail. For example:

  • If there's no field with the long name of the option.
  • If the field type is boolean but the option isn't a boolean flag or vice versa.
  • If the field type isn't a string and the field value cannot be converted from the string value.
  • If the numeric field type is too small to accommodate the number converted from the string value.

An option with a value can be entered multiple times. All values can be stored in an array, for example:

usage := '...

Options:
  -n, --numbers <number>  a list of numbers to use

...'

struct Opts {
  numbers []int
}

Other Arguments

An option starts with - or -- and has to consist of at least one more letter. A single dash ( - ) isn't an option, but another argument. An argument not starting with a dash ( - ) is a plain argument and not an option.

If you want to handle some argument as other arguments and not as options, put two dashes ( -- ) on the command line and appends such arguments behind it. The two dashes ( -- ) will be ignored. If you need the two dashes ( -- ) as another argument, append them once more after the first ones to the command line.

Input Fields

The following input fields are available:

Field Type Default Description
version string 'unknown' version of the tool to print if -V|--version is requested
args ?[]string none raw command-line arguments, defaults to os.args[1..]
disable_short_negative bool false disables handling uppercase letters as negated options
ignore_number_overflow bool false ignores an overflow when converting numbers to option fields
options_anywhere bool false do not look for options only after the line with Options:

Advanced

If the transformation of options name to field name described above is not enough, the argument name can be assigned to a specific field by the attribute arg . For example, set the command-line argument type to a field typ :

usage := '...

Options:
  -t|--type <type>  file type (text or binary)

...'

struct Opts {
  typ string [arg: @type]
}

If you don't want to disable the checks for arithmetic overflow globally, but only for one field, it's possible by the attribute nooverflow . For example, set the command-line argument type to a field typ :

usage := '...

Options:
  -r|--random <number>  the value to initialize random number generator with

...'

struct Opts {
  random i16 [nooverflow]
}

If you require an option to be always entered, it's possible by the attribute required . For example:

usage := '...

Options:
  -f|--file <name>  the name of the output file

...'

struct Opts {
  file string [required]
}

If you need to supply multiple values for an option and you want to use more condensed syntax then repeating the option on the command line, you can supply all values only once, if there's a separator, which otherwise cannot be present within a value. For example, you can supply two comma-delimited integers as -n 1,2 by the attribute split :

usage := '...

Options:
  -n, --numbers <number>  a list of numbers to use

...'

struct Opts {
  numbers []int [split]
}

The default separator is , (comma). If you need a different one, you can choose the separator by the same attribute. For example, you can supply two semicolon-delimited characters as -c a;b by the attribute split :

usage := '...

Options:
  -c, --chars <char>  allowed characters

...'

struct Opts {
  numbers []rune [split: ';']
}

TODO

This is a work in progress.

  • Add a string map as the parsing target.

About

Parses command-line arguments to statically typed options or a string map with the help of usage description.

0
354
last Jun 11

Author

prantlf