3 # Copyright © 2021 Nick Bowler
5 # Generate definitions helpful when using getopt_long from an options
8 # The options specification file is processed line by line. Any line
9 # beginning with a - character introduces a new option definition. Each
10 # option definition specifies any or all of a short option name, a long
11 # option name, an argument specification, and an action specification.
13 # Only the long option name is mandatory. It is not possible to define
14 # short options without a corresponding long option.
16 # The optional short option name is first, and consists of a hyphen (which
17 # must be the first character on the line) followed by the one character
18 # short option name, followed by a comma.
20 # The long option name is next on the line, which consists of two hyphens
21 # followed by the desired option name. If the short option name was omitted,
22 # then the first hyphen of the long option name must be the first character
25 # The argument specification is next, consisting of an equals sign followed by
26 # the argument name. The argument name can be any sequence of non-whitespace
27 # characters and only relevant for --help text.
29 # If the argument specification is surrounded by square brackets, this
30 # indicates an optional argument. If the argument specification is omitted
31 # completely, this option has no argument. Otherwise, the option has a
34 # Finally, the optional action specification defines how the "flag" and
35 # "val" members are set in the option structure for this option. An action
36 # specification may only be provided for options with no short name.
38 # If the action specification is omitted, then flag will be set to a null
39 # pointer and val is set to the short option character, if any, otherwise the
40 # unique enumeration constant LOPT_xxx for this option (described below).
42 # The action specification can be of the form (val) or (flag, val), where flag
43 # and val are C expressions suitable for use in an initializer for objects
44 # with static storage duration. Neither flag nor val may contain commas or
45 # whitespace. In the first form, the option's flag is set to a null pointer.
47 # Any amount of whitespace may follow the short option name, the argument
48 # specification, the action specification, or the comma within an action
49 # specification. Whitespace is not permitted between a long option name
50 # and a flag specification.
52 # Examples of option specifications:
58 # --parse-only (&parse_only, 1)
60 # Each option is assigned an enumeration constant of the form LOPT_xxx,
61 # where xxx is the long option name with all letters in uppercase and
62 # all non-alphanumeric characters replaced with underscores. The value
63 # of the constants is unspecified, except that they will be unique across
64 # all defined options and distinct from the integer value of any short
67 # The object-like macro SOPT_STRING expands to a string literal suitable
68 # for use as the optstring argument to getopt et al.
70 # The object-like macro LOPTS_INITIALIZER expands to a comma-separated
71 # sequence of struct option initializers, suitable for use in a declaration
72 # of an array of struct option elements with static storage duration. The
73 # all-zero terminating element required by getopt_long must be added by the
76 # static const struct option lopts[] = { LOPTS_INITIALIZER, {0} };
78 # The help text for an individual struct option element may be obtained by
81 # struct lopt_help { const char *desc, *arg; }
82 # *lopt_get_help(const struct option *opt);
84 # The returned desc and arg pointers point to the argument name and help text
85 # for the argument, respectively, as written in the options specification file.
87 # License WTFPL2: Do What The Fuck You Want To Public License, version 2.
88 # This is free software: you are free to do what the fuck you want to.
89 # There is NO WARRANTY, to the extent permitted by law.
94 print " * Automatically generated by gen-options.awk from " FILENAME
96 print " * Automatically generated by gen-options.awk"
98 print " * Do not edit."
109 # Parse option specifier lines
112 arg = lopt = sopt = ""
115 # Extract short option name
116 if (work ~ /^-[^-]/) {
117 sopt = substr(work, 2, 1)
118 sub(/^-.,[ \t]*/, "", work)
121 # Extract long option name
123 if (n = match(work, /[= \t[]/)) {
124 lopt = substr(work, 3, n-3)
125 work = substr(work, n)
127 lopt = substr(work, 3)
132 # Extract argument name
133 if (work ~ /^\[=[^\] \t]+\]/) {
134 if (n = index(work, "]")) {
135 arg = substr(work, 3, n-3)
136 work = substr(work, n+1)
139 } else if (work ~ /^=/) {
140 if (n = match(work, /[ \t]/)) {
141 arg = substr(work, 2, n-2)
142 work = substr(work, n)
144 arg = substr(work, 2)
151 sub(/^[ \t]*/, "", work)
152 if (!sopt && work ~ /^\([^, \t]+(,[ \t]*[^, \t]+)?\)/) {
153 n = split(work, a, /,[ \t]*/)
155 flag = substr(a[1], 2) ", " substr(a[2], 1, length(a[2])-1)
157 flag = "NULL, " substr(a[1], 2, length(a[1])-2)
159 sub(/^\([^, \t]+(,[ \t]*[^, \t]+)?/, "", work)
161 flag = "NULL, '" sopt "'"
163 flag = "NULL, " to_enum(lopt)
167 print "invalid option specification:", $0 > "/dev/stderr"
173 sopt_string = sopt_string sopt substr("::", 1, has_arg)
175 options[num_options++] = lopt
176 optionspec[lopt] = has_arg ", " flag
178 optionarg[lopt] = arg
184 # Ignore any line beginning with a #
191 optionhelp[lopt] = (lopt in optionhelp ? optionhelp[lopt] "\n" : "") $0
194 # Exit immediately on error
195 END { if (err) { exit err } }
198 print "#include <stddef.h>"
199 print "#include <limits.h>\n"
200 print "#define SOPT_STRING \"" sopt_string "\"\n"
203 # Generate the main options tables
207 count = bucketsort(sorted_options, options)
208 for (i = 0; i < count; i++) {
209 lopt_strings = add_to_strtab(lopt_strings, sorted_options[i], offsets)
211 gsub(/[^ ]+/, "\"&", lopt_strings)
212 gsub(/ /, "\\0\"\n\t", lopt_strings)
214 print "static const char lopt_strings[] ="
215 print "\t" lopt_strings "\";\n"
217 for (i = 0; i < count; i++) {
219 sep = (i+1 == count ? "" : ",")
221 print "\t" to_enum(opt), "= UCHAR_MAX+1 +", offsets[opt] sep
224 print "#define lopt_str(x) (lopt_strings + (LOPT_ ## x - UCHAR_MAX - 1))\n"
226 print "#define LOPTS_INITIALIZER \\"
227 for (i = 0; i < count; i++) {
229 sep = (i+1 == count ? "" : ", \\")
231 print "\t/* --" opt, "*/ \\"
232 print "\t{ lopt_strings+" offsets[opt] ",", optionspec[opt] " }" sep
236 # Generate the help strings
238 # First, sort out the argument names
241 count = bucketsort(sorted_args, optionarg)
242 for (i = 0; i < count; i++) {
243 arg_strings = add_to_strtab(arg_strings, sorted_args[i], arg_offsets)
246 n = split(arg_strings, arg_split)
248 for (i = 1; i <= n; i++) {
249 for (opt in optionarg) {
250 if (optionarg[opt] == arg_split[i]) {
256 sep = (i < n ? "\"\\0\"" : "")
257 arg_strings = arg_strings "\n\tPN_(\"" opt "\", \"" arg_split[i] "\")" sep
260 print "\n#define ARG_L10N_(x)"
262 print "# define PN_(c, x) x"
265 print "static const char arg_strings[] = " arg_strings "\"\";"
266 for (opt in optionarg) {
267 if (opt in l10narg) {
270 print "\tARG_L10N_(PN_(\"" opt "\", \"" optionarg[opt] "\"))"
273 # Then add in the actual descriptions
274 print "\nstatic const char help_strings[] ="
277 for (opt in options) {
279 if (opt in optionhelp) {
284 help = optionhelp[opt]
285 help_offsets[opt] = help_pos
286 help_pos += length(help) + 1
288 gsub(/"/, "\\\"", help)
289 gsub(/\n/, "\\n\"\n\t \"", help)
290 help = "\tPN_(\"" opt "\",\n\t \"" help "\")"
294 for (opt in options) {
296 if (!(opt in optionhelp)) {
297 print "\tARG_L10N_(PN_(\"" opt "\", \"\"))"
298 help_offsets[opt] = help_pos - 1
302 print "\nstatic struct lopt_help { const char *desc, *arg; }"
303 print "*lopt_get_help(const struct option *opt, struct lopt_help *out)\n{"
304 print "\tswitch ((opt->name - lopt_strings) + UCHAR_MAX + 1) {"
305 for (opt in options) {
307 print "\tcase", to_enum(opt) ":"
308 print "\t\tout->desc = help_strings +", help_offsets[opt] ";"
309 if (opt in optionarg) {
310 print "\t\tout->arg = arg_strings +", arg_offsets[optionarg[opt]] ";"
312 print "\t\treturn out;"
314 print "\t}\n\n\treturn NULL;"
318 # bucketsort(dst, src)
320 # Sort the elements of src by descending string length,
321 # placing them into dst[0] ... dst[n].
323 # Returns the number of elements.
324 function bucketsort(dst, src, buckets, max, count, i, t)
328 if (i > max) { max = i }
332 for (i = max; i > 0; i--) {
341 i = length(t = src[t])
342 dst[buckets[i]++] = t
350 # Return the string LOPT_xxx, where xxx is the argument with all lowercase
351 # letters converted to uppercase, and all non-alphanumeric characters replaced
353 function to_enum(lopt)
356 gsub(/[^ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789]/, "_", lopt)
360 # add_to_strtab(strtab, str, offsets)
362 # Append string to strtab if there is not already a matching string present
363 # in the table. Newly-added strings are separated by spaces, which must be
364 # translated into null bytes afterwards. The updated strtab is returned, and
365 # the offsets[str] array member is updated with the position (counting from 0)
366 # of str in the strtab.
368 # For optimal results, strings should be added in descending length order.
369 function add_to_strtab(strtab, str, offsets, pos)
371 if ( (pos = index(strtab, str " ") - 1) < 0) {
374 strtab = strtab " " str