Each arg_xxx struct has it own unique set of constructor functions and while these may differ slightly between arg_xxx structs, they are generally of the form:

struct arg_xxx* arg_xxx0 (const char *shortopts, const char *longopts, const char *datatype, const char *glossary)

struct arg_xxx* arg_xxx1 (const char *shortopts, const char *longopts, const char *datatype, const char *glossary)

struct arg_xxx* arg_xxxn (const char *shortopts, const char *longopts, const char *datatype, int mincount, int maxcount, const char *glossary)

The arg_xxx0() and arg_xxx1() forms are merely abbreviated forms of arg_xxxn() and are provided as a convenience for the most common arrangements of command line options; namely those that have zero-or-one occurrences (mincount=0,maxcount=1) and those that have one exactly one occurrence (mincount=1,maxcount=1) respectively.

The const char* shortopts parameter defines the option's short form tag (eg: -x, -k3, -D"macro"). It can be left as NULL if a short option is not required, otherwise use it to specify the desired short option character in the string (without the leading "-" and without any whitespace). For example, the short option -v is defined simply as "v". In fact, a command line option may have multiple alternate short form tags defined for it by concatenating the desired characters into the shortopts string. For instance "abc" defines an option which will accept any of the three equivalent short forms -a, -b, -c interchangeably.

The const char* longopts parameter is similar to shortopts, except it defines the option's long form tags (eg: --help, --depth=3, --name=myfile.txt). It too can be left as NULL if not required, and it too can have multiple equivalent tags defined but these must be separated by commas. For example, if we wish to define two equivalent long options --quiet and --silent then we would give longopts as "quiet,silent". Remember not to include any whitespace.

If both shortopts and longopts are given as NULL then the resulting option is an untagged argument.

The const char* datatype parameter is a descriptive string you can use to customize the appearance of the argument data type in error messages and so forth. It does not affect the actual data type definition as that is a fixed property of the arg_xxx struct. So for example, defining a datatype of "<bar>" will result in the option being display something like "-x <bar>" or "--foo=<bar>" depending upon your option tags. If given as NULL, the datatype string will revert to the default value for the particular arg_xxx struct. You can effectively disable the default by specifying datatype as an empty string.

The int mincount parameter specifies the minimum number of occurrences that the option must appear on the command line. If the option does not appear at least that many times then the parser reports it as a syntax error. The mincount defaults to 0 for the arg_xxx0() functions and 1 for arg_xxx1() functions.

The int maxcount parameter specifies the maximum number of occurrences that the option may appear on the command line. Any occurrences beyond the maximum are discarded by the parser reported as syntax errors. The maxcount defaults to 1 for both the arg_xxx0() and arg_xxx1() functions.

The const char* glossary parameter is another descriptive string but this one appears in the glossary table summarizing the program's command line options. The glossary table is generated automatically by the arg_print_glossary function (see later). For example, a glossary string of "the foobar factor" would appear in the glossary table along side the option something like:

--foo=<bar> the foobar factor

Specifying a NULL glossary string causes that option to be omitted from the glossary table.

See below for the exact definitions of the individual arg_xxx structs and their constructor functions.

Returns non-zero if the argtable[] array contains any NULL entries up until the terminating arg_end* entry. Returns zero otherwise.

Parse the command line arguments in argv[] using the command line syntax specified in argtable[], returning the number of errors encountered. Error details are recorded in the argument table's arg_end structure from where they can be displayed later with the arg_print_errors function. Upon a successful parse, the arg_xxx structures referenced in argtable[] will contain the argument values extracted from the command line.

This function prints an option's syntax, as in -K|--scalar=<int>, where the short options, long options, and datatype are all given as parameters of this function. It is primarily used within the arg_xxx structures' errorfn functions as a way of displaying an option's syntax inside of error messages. However, it can also be used in user code if desired. The suffix string is provided as a convenience for appending newlines and so forth to the end of the display and can be given as NULL if not required.

Prints the GNU style command line syntax for the given argument table, as in: [-abcv] [--scalar=<n>] [-o myfile] <file> [<file>]

The suffix string is provided as a convenience for appending newlines and so forth to the end of the display and can be given as NULL if not required.

Prints the verbose form of the command line syntax for the given argument table, as in: [-a] [-b] [-c] [--scalar=<n>] [-o myfile] [-v|--verbose] <file> [<file>]

The suffix string is provided as a convenience for appending newlines and so forth to the end of the display and can be given as NULL if not required.

Prints a glossary table describing each option in the given argument table. The format string is passed to printf to control the formatting of each entry in the the glossary. It must have exactly two "%s" format parameters as in "%-25s %s\n", the first is for the option's syntax and the second for its glossary string. If an option's glossary string is NULL then that option in omitted from the glossary display.

Prints the details of all errors stored in the end data structure. The progname string is prepended to each error message.

Deallocates the memory used by each arg_xxx struct referenced by argtable[]. It does this by calling free for each of the n pointers in the argtable array and then nulling them for safety.

-x, -y, -z, --help, --verbose

struct arg_lit
   {
   struct arg_hdr hdr;
   int count;
   };

struct arg_lit* arg_lit0 (const char *shortopts, const char *longopts, const char *glossary) struct arg_lit* arg_lit1 (const char *shortopts, const char *longopts, const char *glossary) struct arg_lit* arg_litn (const char *shortopts, const char *longopts, int mincount, int maxcount, const char *glossary)

Literal options take no argument values so all that is to be seen in the arg_lit struct is the count of the number of times the option was present on the command line. Upon a successful parse, count is guaranteed to be within the mincount and maxcount limits set for the option at construction.

-x2, -y 7, -z-3, --size=734, --count 124

struct arg_int
   {
   struct arg_hdr hdr;
   int count;
   int *ival;
   };

struct arg_int* arg_int0 (const char *shortopts, const char *longopts, const char *datatype, const char *glossary) struct arg_int* arg_int1 (const char *shortopts, const char *longopts, const char *datatype, const char *glossary) struct arg_int* arg_intn (const char *shortopts, const char *longopts, const char *datatype, int mincount, int maxcount, const char *glossary

The arg_int struct contains the count of the number of times the option was present on the command line and a pointer (ival) to an array containing the integer values given with those particular options. The array is fixed at construction time to hold maxcount integers at most.

Upon a successful parse, count is guaranteed to be within the mincount and maxcount limits set for the option at construction with the appropriate values store in the ival array. The parser will not accept any values beyond that limit.

It is quite acceptable to set default values in the ival array prior to calling arg_parse if desired as the parser does alter ival entries for which no command line argument is received.

-x2.234, -y 7e-03, -z-3.3E+6, --pi=3.1415, --tolerance 1.0E-6

struct arg_dbl
   {
   struct arg_hdr hdr;
   int count;
   double *dval;
   };

struct arg_dbl* arg_dbl0 (const char *shortopts, const char *longopts, const char *datatype, const char *glossary) struct arg_dbl* arg_dbl1 (const char *shortopts, const char *longopts, const char *datatype, const char *glossary) struct arg_dbl* arg_dbln (const char *shortopts, const char *longopts, const char *datatype, int mincount, int maxcount, const char *glossary

Like arg_int but the arguments values are stored as doubles in dval.

-Dmacro, -t mytitle, -m "my message string", --title="hello world"

struct arg_str
   {
   struct arg_hdr hdr;
   int count;
   const char **sval;
   };

struct arg_str* arg_str0 (const char *shortopts, const char *longopts, const char *datatype, const char *glossary) struct arg_str* arg_str1 (const char *shortopts, const char *longopts, const char *datatype, const char *glossary) struct arg_str* arg_strn (const char *shortopts, const char *longopts, const char *datatype, int mincount, int maxcount, const char *glossary)

The arg_str struct contains the count of the number of times the option was present on the command line and a pointer (sval) to an array containing pointers to the parsed string values. The array is fixed at construction time to hold maxcount string pointers at most. These pointers in this array reference the actual command line string buffers stored in argv[], so the string contents should not be should not be altered. Although it is quite acceptable to set default string pointers in the sval array prior to calling arg_parse as the parser does alter them if no matching command line argument is received.

"hello world", -t mytitle, -m "my message string", --title="hello world"

struct arg_rex
   {
   struct arg_hdr hdr;
   int count;
   const char **sval;
   };

struct arg_rex* arg_rex0 (const char *shortopts, const char *longopts, const char *pattern, const char *datatype, int flags, const char *glossary) struct arg_rex* arg_rex1 (const char *shortopts, const char *longopts, const char *pattern, const char *datatype, int flags, const char *glossary) struct arg_rex* arg_rexn (const char *shortopts, const char *longopts, const char *pattern, const char *datatype, int mincount, int maxcount, int flags, const char *glossary)

Like arg_str but but the string argument values are only accepted if they match a predefined regular expression. The regular expression is defined by the pattern parameter passed to the arg_rex constructor. The regular expression parsing is done using regex, and its behaviour can be controlled via standard regex bit flags which are passed to argtable via the flags parameter in the arg_rex conbstructors. However the only two regex flags that are relevant to argtable are REG_EXTENDED (use extended regular expressions rather than basic ones) and REG_ICASE (ignore case). These may be logically ORed if desired. This argument type is useful for matching command line keywords, particularly if case insensitive strings or pattern matching is required. See regex(3) for more details of regular expression matching.

Argtable does not support arg_date functionality under Microsoft Windows platforms as the Microsoft compilers do include the necessary regex support as standard.

-o myfile, -Ihome/foo/bar, --input=~/doc/letter.txt, --name a.out

struct arg_file
   {
   struct arg_hdr hdr;
   int count;
   const char **filename;
   const char **basename;
   const char **extension;
   };

struct arg_file* arg_file0 (const char *shortopts, const char *longopts, const char *datatype, const char *glossary) struct arg_file* arg_file1 (const char *shortopts, const char *longopts, const char *datatype, const char *glossary) struct arg_file* arg_filen (const char *shortopts, const char *longopts, const char *datatype, int mincount, int maxcount, const char *glossary)

Like arg_str but the argument strings are presumed to have filename qualities so some additional pasring is done to separate out the filename's basename and extension (if they exist). The three arrays filename[], basename[], extension[] each store up to maxcount entries, and the i'th entry of each of these arrays refer to different components of the same string buffer.

For instance, -o /home/heitmann/mydir/foo.txt would be parsed as:

filename[i]  = "/home/heitmann/mydir/foo.txt"
basename[i]  =                      "foo.txt"
extension[i] =                          "txt"

If the filename has no leading path then the basename is the same as the filename, and if no extension could be identified then it is given as NULL. Note that filename extensions are defined as all text following the last "." in the filename. Thus -o foo would be parsed as:

filename[i]  = "foo"
basename[i]  = "foo"
extension[i] = NULL

As with arg_str, the string pointers in filename[], basename[], and extension[] actually refer to the original argv[] command line string buffers so you should not attempt to alter them.

Note also that the parser only ever treats the filenames as strings and never attempts to open them as files or perform any directory lookups on them.

12/31/04, -d 1982-11-28, --time 23:59

struct arg_date
   {
   struct arg_hdr hdr;
   const char *format; 
   int count;
   struct tm *tmval;
   };

struct arg_date* arg_date0 (const char *shortopts, const char *longopts, const char *format, const char *datatype, const char *glossary) struct arg_date* arg_date1 (const char *shortopts, const char *longopts, const char *format, const char *datatype, const char *glossary) struct arg_date* arg_daten (const char *shortopts, const char *longopts, const char *format, const char *datatype, int mincount, int maxcount, const char *glossary)

Accepts a timestamp string from the command line and converts it to struct tm format using the system strptime function. The time format is defined by the format string passed to the arg_date constructor, and is passed directly to strptime. See strptime(3) for more details on the format string.

Argtable does not support arg_date functionality under Microsoft Windows as the Microsoft compilers do include the necessary strptime support as standard.

struct arg_rem
   {
   struct arg_hdr hdr;
   };

struct arg_rem* arg_rem (const char* datatype, const char* glossary)

The arg_rem struct is a dummy struct in the sense it does not represent a command line option to be parsed. Instead it provides a means to include additional datatype and glossary strings in the output of the arg_print_syntax, arg_print_syntaxv, and arg_print_glossary functions. As such, arg_rem structs may be used in the argument table to insert additional lines of text into the glossary descriptions or to insert additional text fields into the syntax description. It has no data members apart from the mandatory arg_hdr struct.

struct arg_end
   {
   struct arg_hdr hdr;
   int count;
   int *error;
   void **parent;
   const char **argval;
   };

struct arg_end* arg_end (int maxerrors)

The arg_end struct is primarily used to mark the end of an argument table and doesn't represent any command line option. Every argument table must have an arg_end structure as its last entry.

Apart from terminating the argument table, the arg_end structure also stores the error codes generated by the arg_parse function as it attempts to parse the command line with the given argument table. The maxerrors parameter passed to the arg_end constructor specifies the maximum number of errors that the structure can store. Any further errors are discarded and replaced with the single error code ARG_ELIMIT which is later reported to the user by the message "too many errors". A maxerrors limit of 20 is quite reasonable.

The arg_print_errors function will print the errors stored in the arg_end struct in the same order as they occurred, so there is no need to understand the internals of the arg_end struct.

For those that are curious, the three arrays error[], parent[], and argval[] are each allocated maxerrors entries at construction. As usual, the count variable gives the number of entries actually stored in these arrays. The same value applies to all three arrays as the i'th entry of each all refer to different aspects of the same error condition.

The error[i] entry holds the error code returned by the hdr.scanfn function of the particular arg_xxx that is reporting the error. The meaning if the code is usually known only to the issuing arg_xxx struct. The predefined error codes that arg_end handles from the parser itself are the exceptions.

The parent[i] entry points to the parent arg_xxx structure that reported the error. That same arg_xxx structure is also responsible for displaying a pertinent error message when called on to do so by the arg_print_errors function. It calls the hdr.errorfn function of each parent arg_xxx struct listed in the arg_end structure.

Lastly, the argval[i] entry points to the command line argument at which the error occurred, although this may be NULL when there is no relevant command line value. For instance, if an error reports a missing option then there will be no matching command line argument value.