15.9 Command-Line Parsing
|(require racket/cmdline)||package: base|
(command-line optional-name-expr optional-argv-expr flag-clause ... finish-clause)
| #:program name-expr optional-argv-expr =
| #:argv argv-expr flag-clause = #:multi flag-spec ... | #:once-each flag-spec ... | #:once-any flag-spec ... | #:final flag-spec ... | #:usage-help string ... | #:help-labels string ... | #:ps string ... flag-spec = (flags id ... help-spec body ...+) | (flags => handler-expr help-expr) flags = flag-string | (flag-string ...+) help-spec = string | (string-expr ...+) finish-clause =
| #:args arg-formals body ...+ | #:handlers handlers-exprs arg-formals = rest-id | (arg ...) | (arg ...+ . rest-id) arg = id | [id default-expr] handlers-exprs = finish-expr arg-strings-expr | finish-expr arg-strings-expr help-expr |
finish-expr arg-strings-expr help-expr unknown-expr
The name-expr, if provided, should produce a path or string to be used as the program name for reporting errors when the command-line is ill-formed. It defaults to (find-system-path 'run-file). When a path is provided, only the last element of the path is used to report an error.
The argv-expr, if provided, must evaluate to a list or a vector of strings. It defaults to (current-command-line-arguments).
The command-line is disassembled into flags, each possibly with flag-specific arguments, followed by (non-flag) arguments. Command-line strings starting with - or + are parsed as flags, but arguments to flags are never parsed as flags, and integers and decimal numbers that start with - or + are not treated as flags. Non-flag arguments in the command-line must appear after all flags and the flags’ arguments. No command-line string past the first non-flag argument is parsed as a flag. The built-in -- flag signals the end of command-line flags; any command-line string past the -- flag is parsed as a non-flag argument.
A #:multi, #:once-each, #:once-any, or #:final clause introduces a set of command-line flag specifications. The clause tag indicates how many times the flag can appear on the command line:
Each flag specified in the set can be represented any number of times on the command line; i.e., the flags in the set are independent and each flag can be used multiple times.
Each flag specified in the set can be represented once on the command line; i.e., the flags in the set are independent, but each flag should be specified at most once. If a flag specification is represented in the command line more than once, the exn:fail exception is raised.
Only one flag specified in the set can be represented on the command line; i.e., the flags in the set are mutually exclusive. If the set is represented in the command line more than once, the exn:fail exception is raised.
Like #:multi, except that no argument after the flag is treated as a flag. Note that multiple #:final flags can be specified if they have short names; for example, if -a is a #:final flag, then -aa combines two instances of -a in a single command-line argument.
A normal flag specification has four parts:
a flag string, or a set of flag strings. If a set of flags is provided, all of the flags are equivalent. Each flag string must be of the form "-x" or "+x" for some character x, or "–x" or "++x" for some sequence of characters x. An x cannot contain only digits or digits plus a single decimal point, since simple (signed) numbers are not treated as flags. In addition, the flags "--", "-h", and "--help" are predefined and cannot be changed.
identifier that are bound to the flag’s arguments. The number of identifiers determines how many arguments can be provided on the command line with the flag, and the names of these identifiers will appear in the help message describing the flag. The ids are bound to string values in the bodys for handling the flag.
a string or sequence of strings that describes the flag. This string is used in the help message generated by the handler for the built-in -h (or --help) flag. A single literal string can be provided, or any number of expressions that produce strings; in the latter case, strings after the first one are displayed on subsequent lines.
expressions that are evaluated when one of the flags appears on the command line. The flags are parsed left-to-right, and each sequence of bodys is evaluated as the corresponding flag is encountered. When the bodys are evaluated, the preceding ids are bound to the arguments provided for the flag on the command line.
A flag specification using => escapes to a more general method of specifying the handler and help strings. In this case, the handler procedure and help string list returned by handler-expr and help-expr are used as in the table argument of parse-command-line.
A #:usage-help clause inserts text lines immediately after the usage line. Each string in the clause provides a separate line of text.
A #:help-labels clause inserts text lines into the help table of command-line flags. Each string in the clause provides a separate line of text.
A #:ps clause inserts text lines at the end of the help output. Each string in the clause provides a separate line of text.
After the flag clauses, a final clause handles command-line arguments that are not parsed as flags:
Supplying no finish clause is the same as supplying #:args () (void).
For an #:args finish clause, identifiers in arg-formals are bound to the leftover command-line strings in the same way that identifiers are bound for a lambda expression. Thus, specifying a single id (without parentheses) collects all of the leftover arguments into a list. The effective arity of the arg-formals specification determines the number of extra command-line arguments that the user can provide, and the names of the identifiers in arg-formals are used in the help string. When the command-line is parsed, if the number of provided arguments cannot be matched to identifiers in arg-formals, the exn:fail exception is raised. Otherwise, args clause’s bodys are evaluated to handle the leftover arguments, and the result of the last body is the result of the command-line expression.
A #:handlers finish clause escapes to a more general method of handling the leftover arguments. In this case, the values of the expressions are used like the last two to four arguments parse-command-line.
(define verbose-mode (make-parameter #f)) (define profiling-on (make-parameter #f)) (define optimize-level (make-parameter 0)) (define link-flags (make-parameter null)) (define file-to-compile (command-line #:program "compiler" #:once-each [("-v" "--verbose") "Compile with verbose messages" (verbose-mode #t)] [("-p" "--profile") "Compile with profiling" (profiling-on #t)] #:once-any [("-o" "--optimize-1") "Compile with optimization level 1" (optimize-level 1)] ["--optimize-2" (; show help on separate lines "Compile with optimization level 2," "which includes all of level 1") (optimize-level 2)] #:multi [("-l" "--link-flags") lf ; flag takes one argument "Add a flag <lf> for the linker" (link-flags (cons lf (link-flags)))] #:args (filename) ; expect one command-line argument: <filename> ; return the argument as a filename to compile filename))
(parse-command-line name argv table finish-proc arg-help-strs [ help-proc unknown-proc]) → any name : (or/c string? path?) argv : (or/c (listof string?) (vectorof string?)) table : (listof (cons/c symbol? list?)) finish-proc : ((list?) () #:rest list? . ->* . any) arg-help-strs : (listof string?) help-proc : (string? . -> . any) = (lambda (str) ....) unknown-proc : (string? . -> . any) = (lambda (str) ...)
The table argument to this procedural form encodes the information in command-line’s clauses, except for the args clause. Instead, arguments are handled by the finish-proc procedure, and help information about non-flag arguments is provided in arg-help-strs. In addition, the finish-proc procedure receives information accumulated while parsing flags. The help-proc and unknown-proc arguments allow customization that is not possible with command-line.
When there are no more flags, finish-proc is called with a list of information accumulated for command-line flags (see below) and the remaining non-flag arguments from the command-line. The arity of finish-proc determines the number of non-flag arguments accepted and required from the command-line. For example, if finish-proc accepts either two or three arguments, then either one or two non-flag arguments must be provided on the command-line. The finish-proc procedure can have any arity (see procedure-arity) except 0 or a list of 0s (i.e., the procedure must at least accept one or more arguments).
The arg-help-strs argument is a list of strings identifying the expected (non-flag) command-line arguments, one for each argument. If an arbitrary number of arguments are allowed, the last string in arg-help-strs represents all of them.
The help-proc procedure is called with a help string if the -h or --help flag is included on the command line. If an unknown flag is encountered, the unknown-proc procedure is called just like a flag-handling procedure (as described below); it must at least accept one argument (the unknown flag), but it may also accept more arguments. The default help-proc displays the string and exits and the default unknown-proc raises the exn:fail exception.
A table is a list of flag specification sets. Each set is represented as a pair of two items: a mode symbol and a list of either help strings or flag specifications. A mode symbol is one of 'once-each, 'once-any, 'multi, 'final, 'help-labels, 'usage-help, or 'ps with the same meanings as the corresponding clause tags in command-line. For the 'help-labels, 'usage-help or 'ps mode, a list of help strings is provided. For the other modes, a list of flag specifications is provided, where each specification maps a number of flags to a single handler procedure. A specification is a list of three items:
A list of strings for the flags defined by the spec. See command-line for information about the format of flag strings.
A procedure to handle the flag and its arguments when one of the flags is found on the command line. The arity of this handler procedure determines the number of arguments consumed by the flag: the handler procedure is called with a flag string plus the next few arguments from the command line to match the arity of the handler procedure. The handler procedure must accept at least one argument to receive the flag. If the handler accepts arbitrarily many arguments, all of the remaining arguments are passed to the handler. A handler procedure’s arity must either be a number or an arity-at-least value.
The return value from the handler is added to a list that is eventually passed to finish-proc. If the handler returns #<void>, no value is added onto this list. For all non-#<void> values returned by handlers, the order of the values in the list is the same as the order of the arguments on the command-line.
A non-empty list for constructing help information for the spec. The first element of the list describes the flag; it can be a string or a non-empty list of strings, and in the latter case, each string is shown on its own line. Additional elements of the main list must be strings to name the expected arguments for the flag. The number of extra help strings provided for a spec must match the number of arguments accepted by the spec’s handler procedure.
The following example is the same as the core example for command-line, translated to the procedural form:
(parse-command-line "compile" (current-command-line-arguments) `((once-each [("-v" "--verbose") ,(lambda (flag) (verbose-mode #t)) ("Compile with verbose messages")] [("-p" "--profile") ,(lambda (flag) (profiling-on #t)) ("Compile with profiling")]) (once-any [("-o" "--optimize-1") ,(lambda (flag) (optimize-level 1)) ("Compile with optimization level 1")] [("--optimize-2") ,(lambda (flag) (optimize-level 2)) (("Compile with optimization level 2," "which implies all optimizations of level 1"))]) (multi [("-l" "--link-flags") ,(lambda (flag lf) (link-flags (cons lf (link-flags)))) ("Add a flag <lf> for the linker" "lf")])) (lambda (flag-accum file) file) '("filename"))