1 /* Permuted index for GNU, with keywords in their context.
2 Copyright (C) 1990, 1991, 1993, 1998-2005 Free Software Foundation, Inc.
3 François Pinard <pinard@iro.umontreal.ca>, 1988.
5 This program is free software; you can redistribute it and/or modify
6 it under the terms of the GNU General Public License as published by
7 the Free Software Foundation; either version 2, or (at your option)
10 This program is distributed in the hope that it will be useful, but
11 WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with this program; if not, write to the Free Software Foundation,
17 Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
19 François Pinard <pinard@iro.umontreal.ca> */
25 #include <sys/types.h>
35 /* The official name of this program (e.g., no `g' prefix). */
36 #define PROGRAM_NAME "ptx"
38 /* Note to translator: Please translate "F. Pinard" to "François
39 Pinard" if "ç" (c-with-cedilla) is available in the
40 translation's character set and encoding. */
41 #define AUTHORS _("F. Pinard")
43 /* Number of possible characters in a byte. */
44 #define CHAR_SET_SIZE 256
46 #define ISODIGIT(C) ((C) >= '0' && (C) <= '7')
47 #define HEXTOBIN(C) ((C) >= 'a' && (C) <= 'f' ? (C)-'a'+10 \
48 : (C) >= 'A' && (C) <= 'F' ? (C)-'A'+10 : (C)-'0')
49 #define OCTTOBIN(C) ((C) - '0')
51 /* Debugging the memory allocator. */
54 # define MALLOC_FUNC_CHECK 1
58 /* Global definitions. */
60 /* FIXME: There are many unchecked integer overflows in this file,
61 that will cause this command to misbehave given large inputs or
62 options. Many of the "int" values below should be "size_t" or
63 something else like that. */
65 /* Reallocation step when swallowing non regular files. The value is not
66 the actual reallocation step, but its base two logarithm. */
67 #define SWALLOW_REALLOC_LOG 12
69 /* Imported from "regex.c". */
72 /* The name this program was run with. */
75 /* Program options. */
79 UNKNOWN_FORMAT, /* output format still unknown */
80 DUMB_FORMAT, /* output for a dumb terminal */
81 ROFF_FORMAT, /* output for `troff' or `nroff' */
82 TEX_FORMAT /* output for `TeX' or `LaTeX' */
85 static bool gnu_extensions = true; /* trigger all GNU extensions */
86 static bool auto_reference = false; /* refs are `file_name:line_number:' */
87 static bool input_reference = false; /* refs at beginning of input lines */
88 static bool right_reference = false; /* output refs after right context */
89 static int line_width = 72; /* output line width in characters */
90 static int gap_size = 3; /* number of spaces between output fields */
91 static const char *truncation_string = "/";
92 /* string used to mark line truncations */
93 static const char *macro_name = "xx"; /* macro name for roff or TeX output */
94 static enum Format output_format = UNKNOWN_FORMAT;
97 static bool ignore_case = false; /* fold lower to upper for sorting */
98 static const char *context_regex_string = NULL;
99 /* raw regex for end of context */
100 static const char *word_regex_string = NULL;
101 /* raw regex for a keyword */
102 static const char *break_file = NULL; /* name of the `Break characters' file */
103 static const char *only_file = NULL; /* name of the `Only words' file */
104 static const char *ignore_file = NULL; /* name of the `Ignore words' file */
106 /* A BLOCK delimit a region in memory of arbitrary size, like the copy of a
107 whole file. A WORD is something smaller, its length should fit in a
108 short integer. A WORD_TABLE may contain several WORDs. */
112 char *start; /* pointer to beginning of region */
113 char *end; /* pointer to end + 1 of region */
119 char *start; /* pointer to beginning of region */
120 short int size; /* length of the region */
126 WORD *start; /* array of WORDs */
127 size_t alloc; /* allocated length */
128 size_t length; /* number of used entries */
132 /* Pattern description tables. */
134 /* For each character, provide its folded equivalent. */
135 static unsigned char folded_chars[CHAR_SET_SIZE];
137 /* Compiled regex for end of context. */
138 static struct re_pattern_buffer *context_regex;
140 /* End of context pattern register indices. */
141 static struct re_registers context_regs;
143 /* Compiled regex for a keyword. */
144 static struct re_pattern_buffer *word_regex;
146 /* Keyword pattern register indices. */
147 static struct re_registers word_regs;
149 /* A word characters fastmap is used only when no word regexp has been
150 provided. A word is then made up of a sequence of one or more characters
151 allowed by the fastmap. Contains !0 if character allowed in word. Not
152 only this is faster in most cases, but it simplifies the implementation
153 of the Break files. */
154 static char word_fastmap[CHAR_SET_SIZE];
156 /* Maximum length of any word read. */
157 static int maximum_word_length;
159 /* Maximum width of any reference used. */
160 static int reference_max_width;
162 /* Ignore and Only word tables. */
164 static WORD_TABLE ignore_table; /* table of words to ignore */
165 static WORD_TABLE only_table; /* table of words to select */
167 /* Source text table, and scanning macros. */
169 static int number_input_files; /* number of text input files */
170 static int total_line_count; /* total number of lines seen so far */
171 static const char **input_file_name; /* array of text input file names */
172 static int *file_line_count; /* array of `total_line_count' values at end */
174 static BLOCK text_buffer; /* file to study */
176 /* SKIP_NON_WHITE used only for getting or skipping the reference. */
178 #define SKIP_NON_WHITE(cursor, limit) \
179 while (cursor < limit && !ISSPACE(*cursor)) \
182 #define SKIP_WHITE(cursor, limit) \
183 while (cursor < limit && ISSPACE(*cursor)) \
186 #define SKIP_WHITE_BACKWARDS(cursor, start) \
187 while (cursor > start && ISSPACE(cursor[-1])) \
190 #define SKIP_SOMETHING(cursor, limit) \
191 if (word_regex_string) \
194 count = re_match (word_regex, cursor, limit - cursor, 0, NULL); \
195 cursor += count <= 0 ? 1 : count; \
197 else if (word_fastmap[to_uchar (*cursor)]) \
198 while (cursor < limit && word_fastmap[to_uchar (*cursor)]) \
203 /* Occurrences table.
205 The `keyword' pointer provides the central word, which is surrounded
206 by a left context and a right context. The `keyword' and `length'
207 field allow full 8-bit characters keys, even including NULs. At other
208 places in this program, the name `keyafter' refers to the keyword
209 followed by its right context.
211 The left context does not extend, towards the beginning of the file,
212 further than a distance given by the `left' value. This value is
213 relative to the keyword beginning, it is usually negative. This
214 insures that, except for white space, we will never have to backward
215 scan the source text, when it is time to generate the final output
218 The right context, indirectly attainable through the keyword end, does
219 not extend, towards the end of the file, further than a distance given
220 by the `right' value. This value is relative to the keyword
221 beginning, it is usually positive.
223 When automatic references are used, the `reference' value is the
224 overall line number in all input files read so far, in this case, it
225 is of type (int). When input references are used, the `reference'
226 value indicates the distance between the keyword beginning and the
227 start of the reference field, it is of type (DELTA) and usually
230 typedef short int DELTA; /* to hold displacement within one context */
234 WORD key; /* description of the keyword */
235 DELTA left; /* distance to left context start */
236 DELTA right; /* distance to right context end */
237 int reference; /* reference descriptor */
241 /* The various OCCURS tables are indexed by the language. But the time
242 being, there is no such multiple language support. */
244 static OCCURS *occurs_table[1]; /* all words retained from the read text */
245 static size_t occurs_alloc[1]; /* allocated size of occurs_table */
246 static size_t number_of_occurs[1]; /* number of used slots in occurs_table */
249 /* Communication among output routines. */
251 /* Indicate if special output processing is requested for each character. */
252 static char edited_flag[CHAR_SET_SIZE];
254 static int half_line_width; /* half of line width, reference excluded */
255 static int before_max_width; /* maximum width of before field */
256 static int keyafter_max_width; /* maximum width of keyword-and-after field */
257 static int truncation_string_length;/* length of string used to flag truncation */
259 /* When context is limited by lines, wraparound may happen on final output:
260 the `head' pointer gives access to some supplementary left context which
261 will be seen at the end of the output line, the `tail' pointer gives
262 access to some supplementary right context which will be seen at the
263 beginning of the output line. */
265 static BLOCK tail; /* tail field */
266 static int tail_truncation; /* flag truncation after the tail field */
268 static BLOCK before; /* before field */
269 static int before_truncation; /* flag truncation before the before field */
271 static BLOCK keyafter; /* keyword-and-after field */
272 static int keyafter_truncation; /* flag truncation after the keyafter field */
274 static BLOCK head; /* head field */
275 static int head_truncation; /* flag truncation before the head field */
277 static BLOCK reference; /* reference field for input reference mode */
279 /* Miscellaneous routines. */
281 /*------------------------------------------------------.
282 | Duplicate string STRING, while evaluating \-escapes. |
283 `------------------------------------------------------*/
285 /* Loosely adapted from GNU sh-utils printf.c code. */
288 copy_unescaped_string (const char *string)
290 char *result; /* allocated result */
291 char *cursor; /* cursor in result */
292 int value; /* value of \nnn escape */
293 int length; /* length of \nnn escape */
295 result = xmalloc (strlen (string) + 1);
304 case 'x': /* \xhhh escape, 3 chars maximum */
306 for (length = 0, string++;
307 length < 3 && ISXDIGIT (*string);
309 value = value * 16 + HEXTOBIN (*string);
319 case '0': /* \0ooo escape, 3 chars maximum */
321 for (length = 0, string++;
322 length < 3 && ISODIGIT (*string);
324 value = value * 8 + OCTTOBIN (*string);
328 case 'a': /* alert */
337 case 'b': /* backspace */
342 case 'c': /* cancel the rest of the output */
347 case 'f': /* form feed */
352 case 'n': /* new line */
357 case 'r': /* carriage return */
362 case 't': /* horizontal tab */
367 case 'v': /* vertical tab */
378 *cursor++ = *string++;
383 *cursor++ = *string++;
389 /*-------------------------------------------------------------------.
390 | Compile the regex represented by STRING, diagnose and abort if any |
391 | error. Returns the compiled regex structure. |
392 `-------------------------------------------------------------------*/
394 static struct re_pattern_buffer *
395 alloc_and_compile_regex (const char *string)
397 struct re_pattern_buffer *pattern; /* newly allocated structure */
398 const char *message; /* error message returned by regex.c */
400 pattern = xmalloc (sizeof *pattern);
401 memset (pattern, 0, sizeof (struct re_pattern_buffer));
403 pattern->buffer = NULL;
404 pattern->allocated = 0;
405 pattern->translate = ignore_case ? (char *) folded_chars : NULL;
406 pattern->fastmap = xmalloc ((size_t) CHAR_SET_SIZE);
408 message = re_compile_pattern (string, (int) strlen (string), pattern);
410 error (EXIT_FAILURE, 0, _("%s (for regexp `%s')"), message, string);
412 /* The fastmap should be compiled before `re_match'. The following
413 call is not mandatory, because `re_search' is always called sooner,
414 and it compiles the fastmap if this has not been done yet. */
416 re_compile_fastmap (pattern);
418 /* Do not waste extra allocated space. */
420 if (pattern->allocated > pattern->used)
423 = xrealloc (pattern->buffer, (size_t) pattern->used);
424 pattern->allocated = pattern->used;
430 /*------------------------------------------------------------------------.
431 | This will initialize various tables for pattern match and compiles some |
433 `------------------------------------------------------------------------*/
436 initialize_regex (void)
438 int character; /* character value */
440 /* Initialize the case folding table. */
443 for (character = 0; character < CHAR_SET_SIZE; character++)
444 folded_chars[character] = TOUPPER (character);
446 /* Unless the user already provided a description of the end of line or
447 end of sentence sequence, select an end of line sequence to compile.
448 If the user provided an empty definition, thus disabling end of line
449 or sentence feature, make it NULL to speed up tests. If GNU
450 extensions are enabled, use end of sentence like in GNU emacs. If
451 disabled, use end of lines. */
453 if (context_regex_string)
455 if (!*context_regex_string)
456 context_regex_string = NULL;
458 else if (gnu_extensions & !input_reference)
459 context_regex_string = "[.?!][]\"')}]*\\($\\|\t\\| \\)[ \t\n]*";
461 context_regex_string = "\n";
463 if (context_regex_string)
464 context_regex = alloc_and_compile_regex (context_regex_string);
466 /* If the user has already provided a non-empty regexp to describe
467 words, compile it. Else, unless this has already been done through
468 a user provided Break character file, construct a fastmap of
469 characters that may appear in a word. If GNU extensions enabled,
470 include only letters of the underlying character set. If disabled,
471 include almost everything, even punctuations; stop only on white
474 if (word_regex_string && *word_regex_string)
475 word_regex = alloc_and_compile_regex (word_regex_string);
476 else if (!break_file)
483 for (character = 0; character < CHAR_SET_SIZE; character++)
484 word_fastmap[character] = ISALPHA (character) ? 1 : 0;
489 /* Simulate [^ \t\n]+. */
491 memset (word_fastmap, 1, CHAR_SET_SIZE);
492 word_fastmap[' '] = 0;
493 word_fastmap['\t'] = 0;
494 word_fastmap['\n'] = 0;
499 /*------------------------------------------------------------------------.
500 | This routine will attempt to swallow a whole file name FILE_NAME into a |
501 | contiguous region of memory and return a description of it into BLOCK. |
502 | Standard input is assumed whenever FILE_NAME is NULL, empty or "-". |
504 | Previously, in some cases, white space compression was attempted while |
505 | inputting text. This was defeating some regexps like default end of |
506 | sentence, which checks for two consecutive spaces. If white space |
507 | compression is ever reinstated, it should be in output routines. |
508 `------------------------------------------------------------------------*/
511 swallow_file_in_memory (const char *file_name, BLOCK *block)
513 int file_handle; /* file descriptor number */
514 struct stat stat_block; /* stat block for file */
515 size_t allocated_length; /* allocated length of memory buffer */
516 size_t used_length; /* used length in memory buffer */
517 int read_length; /* number of character gotten on last read */
519 /* As special cases, a file name which is NULL or "-" indicates standard
520 input, which is already opened. In all other cases, open the file from
522 bool using_stdin = !file_name || !*file_name || STREQ (file_name, "-");
524 file_handle = STDIN_FILENO;
526 if ((file_handle = open (file_name, O_RDONLY)) < 0)
527 error (EXIT_FAILURE, errno, "%s", file_name);
529 /* If the file is a plain, regular file, allocate the memory buffer all at
530 once and swallow the file in one blow. In other cases, read the file
531 repeatedly in smaller chunks until we have it all, reallocating memory
532 once in a while, as we go. */
534 if (fstat (file_handle, &stat_block) < 0)
535 error (EXIT_FAILURE, errno, "%s", file_name);
537 if (S_ISREG (stat_block.st_mode))
539 size_t in_memory_size;
541 block->start = xmalloc ((size_t) stat_block.st_size);
543 if ((in_memory_size = read (file_handle,
544 block->start, (size_t) stat_block.st_size))
545 != stat_block.st_size)
548 /* On MSDOS, in memory size may be smaller than the file
549 size, because of end of line conversions. But it can
550 never be smaller than half the file size, because the
551 minimum is when all lines are empty and terminated by
553 if (in_memory_size != (size_t)-1
554 && in_memory_size >= stat_block.st_size / 2)
555 block->start = xrealloc (block->start, in_memory_size);
557 #endif /* not MSDOS */
559 error (EXIT_FAILURE, errno, "%s", file_name);
561 block->end = block->start + in_memory_size;
565 block->start = xmalloc ((size_t) 1 << SWALLOW_REALLOC_LOG);
567 allocated_length = (1 << SWALLOW_REALLOC_LOG);
569 while (read_length = read (file_handle,
570 block->start + used_length,
571 allocated_length - used_length),
574 used_length += read_length;
575 if (used_length == allocated_length)
577 allocated_length += (1 << SWALLOW_REALLOC_LOG);
579 = xrealloc (block->start, allocated_length);
584 error (EXIT_FAILURE, errno, "%s", file_name);
586 block->end = block->start + used_length;
589 /* Close the file, but only if it was not the standard input. */
591 if (! using_stdin && close (file_handle) != 0)
592 error (EXIT_FAILURE, errno, "%s", file_name);
595 /* Sort and search routines. */
597 /*--------------------------------------------------------------------------.
598 | Compare two words, FIRST and SECOND, and return 0 if they are identical. |
599 | Return less than 0 if the first word goes before the second; return |
600 | greater than 0 if the first word goes after the second. |
602 | If a word is indeed a prefix of the other, the shorter should go first. |
603 `--------------------------------------------------------------------------*/
606 compare_words (const void *void_first, const void *void_second)
608 #define first ((const WORD *) void_first)
609 #define second ((const WORD *) void_second)
610 int length; /* minimum of two lengths */
611 int counter; /* cursor in words */
612 int value; /* value of comparison */
614 length = first->size < second->size ? first->size : second->size;
618 for (counter = 0; counter < length; counter++)
620 value = (folded_chars [to_uchar (first->start[counter])]
621 - folded_chars [to_uchar (second->start[counter])]);
628 for (counter = 0; counter < length; counter++)
630 value = (to_uchar (first->start[counter])
631 - to_uchar (second->start[counter]));
637 return first->size - second->size;
642 /*-----------------------------------------------------------------------.
643 | Decides which of two OCCURS, FIRST or SECOND, should lexicographically |
644 | go first. In case of a tie, preserve the original order through a |
645 | pointer comparison. |
646 `-----------------------------------------------------------------------*/
649 compare_occurs (const void *void_first, const void *void_second)
651 #define first ((const OCCURS *) void_first)
652 #define second ((const OCCURS *) void_second)
655 value = compare_words (&first->key, &second->key);
656 return value == 0 ? first->key.start - second->key.start : value;
661 /*------------------------------------------------------------.
662 | Return !0 if WORD appears in TABLE. Uses a binary search. |
663 `------------------------------------------------------------*/
666 search_table (WORD *word, WORD_TABLE *table)
668 int lowest; /* current lowest possible index */
669 int highest; /* current highest possible index */
670 int middle; /* current middle index */
671 int value; /* value from last comparison */
674 highest = table->length - 1;
675 while (lowest <= highest)
677 middle = (lowest + highest) / 2;
678 value = compare_words (word, table->start + middle);
680 highest = middle - 1;
689 /*---------------------------------------------------------------------.
690 | Sort the whole occurs table in memory. Presumably, `qsort' does not |
691 | take intermediate copies or table elements, so the sort will be |
692 | stabilized throughout the comparison routine. |
693 `---------------------------------------------------------------------*/
696 sort_found_occurs (void)
699 /* Only one language for the time being. */
701 qsort (occurs_table[0], number_of_occurs[0], sizeof (OCCURS),
705 /* Parameter files reading routines. */
707 /*----------------------------------------------------------------------.
708 | Read a file named FILE_NAME, containing a set of break characters. |
709 | Build a content to the array word_fastmap in which all characters are |
710 | allowed except those found in the file. Characters may be repeated. |
711 `----------------------------------------------------------------------*/
714 digest_break_file (const char *file_name)
716 BLOCK file_contents; /* to receive a copy of the file */
717 char *cursor; /* cursor in file copy */
719 swallow_file_in_memory (file_name, &file_contents);
721 /* Make the fastmap and record the file contents in it. */
723 memset (word_fastmap, 1, CHAR_SET_SIZE);
724 for (cursor = file_contents.start; cursor < file_contents.end; cursor++)
725 word_fastmap[to_uchar (*cursor)] = 0;
730 /* If GNU extensions are enabled, the only way to avoid newline as
731 a break character is to write all the break characters in the
732 file with no newline at all, not even at the end of the file.
733 If disabled, spaces, tabs and newlines are always considered as
734 break characters even if not included in the break file. */
736 word_fastmap[' '] = 0;
737 word_fastmap['\t'] = 0;
738 word_fastmap['\n'] = 0;
741 /* Return the space of the file, which is no more required. */
743 free (file_contents.start);
746 /*-----------------------------------------------------------------------.
747 | Read a file named FILE_NAME, containing one word per line, then |
748 | construct in TABLE a table of WORD descriptors for them. The routine |
749 | swallows the whole file in memory; this is at the expense of space |
750 | needed for newlines, which are useless; however, the reading is fast. |
751 `-----------------------------------------------------------------------*/
754 digest_word_file (const char *file_name, WORD_TABLE *table)
756 BLOCK file_contents; /* to receive a copy of the file */
757 char *cursor; /* cursor in file copy */
758 char *word_start; /* start of the current word */
760 swallow_file_in_memory (file_name, &file_contents);
766 /* Read the whole file. */
768 cursor = file_contents.start;
769 while (cursor < file_contents.end)
772 /* Read one line, and save the word in contains. */
775 while (cursor < file_contents.end && *cursor != '\n')
778 /* Record the word in table if it is not empty. */
780 if (cursor > word_start)
782 if (table->length == table->alloc)
784 if ((SIZE_MAX / sizeof *table->start - 1) / 2 < table->alloc)
786 table->alloc = table->alloc * 2 + 1;
787 table->start = xrealloc (table->start,
788 table->alloc * sizeof *table->start);
791 table->start[table->length].start = word_start;
792 table->start[table->length].size = cursor - word_start;
796 /* This test allows for an incomplete line at end of file. */
798 if (cursor < file_contents.end)
802 /* Finally, sort all the words read. */
804 qsort (table->start, table->length, (size_t) sizeof (WORD), compare_words);
807 /* Keyword recognition and selection. */
809 /*----------------------------------------------------------------------.
810 | For each keyword in the source text, constructs an OCCURS structure. |
811 `----------------------------------------------------------------------*/
814 find_occurs_in_text (void)
816 char *cursor; /* for scanning the source text */
817 char *scan; /* for scanning the source text also */
818 char *line_start; /* start of the current input line */
819 char *line_scan; /* newlines scanned until this point */
820 int reference_length; /* length of reference in input mode */
821 WORD possible_key; /* possible key, to ease searches */
822 OCCURS *occurs_cursor; /* current OCCURS under construction */
824 char *context_start; /* start of left context */
825 char *context_end; /* end of right context */
826 char *word_start; /* start of word */
827 char *word_end; /* end of word */
828 char *next_context_start; /* next start of left context */
830 /* reference_length is always used within `if (input_reference)'.
831 However, GNU C diagnoses that it may be used uninitialized. The
832 following assignment is merely to shut it up. */
834 reference_length = 0;
836 /* Tracking where lines start is helpful for reference processing. In
837 auto reference mode, this allows counting lines. In input reference
838 mode, this permits finding the beginning of the references.
840 The first line begins with the file, skip immediately this very first
841 reference in input reference mode, to help further rejection any word
842 found inside it. Also, unconditionally assigning these variable has
843 the happy effect of shutting up lint. */
845 line_start = text_buffer.start;
846 line_scan = line_start;
849 SKIP_NON_WHITE (line_scan, text_buffer.end);
850 reference_length = line_scan - line_start;
851 SKIP_WHITE (line_scan, text_buffer.end);
854 /* Process the whole buffer, one line or one sentence at a time. */
856 for (cursor = text_buffer.start;
857 cursor < text_buffer.end;
858 cursor = next_context_start)
861 /* `context_start' gets initialized before the processing of each
862 line, or once for the whole buffer if no end of line or sentence
863 sequence separator. */
865 context_start = cursor;
867 /* If a end of line or end of sentence sequence is defined and
868 non-empty, `next_context_start' will be recomputed to be the end of
869 each line or sentence, before each one is processed. If no such
870 sequence, then `next_context_start' is set at the end of the whole
871 buffer, which is then considered to be a single line or sentence.
872 This test also accounts for the case of an incomplete line or
873 sentence at the end of the buffer. */
875 if (context_regex_string
876 && (re_search (context_regex, cursor, text_buffer.end - cursor,
877 0, text_buffer.end - cursor, &context_regs)
879 next_context_start = cursor + context_regs.end[0];
882 next_context_start = text_buffer.end;
884 /* Include the separator into the right context, but not any suffix
885 white space in this separator; this insures it will be seen in
886 output and will not take more space than necessary. */
888 context_end = next_context_start;
889 SKIP_WHITE_BACKWARDS (context_end, context_start);
891 /* Read and process a single input line or sentence, one word at a
898 /* If a word regexp has been compiled, use it to skip at the
899 beginning of the next word. If there is no such word, exit
903 if (re_search (word_regex, cursor, context_end - cursor,
904 0, context_end - cursor, &word_regs)
907 word_start = cursor + word_regs.start[0];
908 word_end = cursor + word_regs.end[0];
912 /* Avoid re_search and use the fastmap to skip to the
913 beginning of the next word. If there is no more word in
914 the buffer, exit the loop. */
918 while (scan < context_end
919 && !word_fastmap[to_uchar (*scan)])
922 if (scan == context_end)
927 while (scan < context_end
928 && word_fastmap[to_uchar (*scan)])
934 /* Skip right to the beginning of the found word. */
938 /* Skip any zero length word. Just advance a single position,
939 then go fetch the next word. */
941 if (word_end == word_start)
947 /* This is a genuine, non empty word, so save it as a possible
948 key. Then skip over it. Also, maintain the maximum length of
949 all words read so far. It is mandatory to take the maximum
950 length of all words in the file, without considering if they
951 are actually kept or rejected, because backward jumps at output
952 generation time may fall in *any* word. */
954 possible_key.start = cursor;
955 possible_key.size = word_end - word_start;
956 cursor += possible_key.size;
958 if (possible_key.size > maximum_word_length)
959 maximum_word_length = possible_key.size;
961 /* In input reference mode, update `line_start' from its previous
962 value. Count the lines just in case auto reference mode is
963 also selected. If it happens that the word just matched is
964 indeed part of a reference; just ignore it. */
968 while (line_scan < possible_key.start)
969 if (*line_scan == '\n')
973 line_start = line_scan;
974 SKIP_NON_WHITE (line_scan, text_buffer.end);
975 reference_length = line_scan - line_start;
979 if (line_scan > possible_key.start)
983 /* Ignore the word if an `Ignore words' table exists and if it is
984 part of it. Also ignore the word if an `Only words' table and
985 if it is *not* part of it.
987 It is allowed that both tables be used at once, even if this
988 may look strange for now. Just ignore a word that would appear
989 in both. If regexps are eventually implemented for these
990 tables, the Ignore table could then reject words that would
991 have been previously accepted by the Only table. */
993 if (ignore_file && search_table (&possible_key, &ignore_table))
995 if (only_file && !search_table (&possible_key, &only_table))
998 /* A non-empty word has been found. First of all, insure
999 proper allocation of the next OCCURS, and make a pointer to
1000 where it will be constructed. */
1002 if (number_of_occurs[0] == occurs_alloc[0])
1004 if ((SIZE_MAX / sizeof *occurs_table[0] - 1) / 2
1007 occurs_alloc[0] = occurs_alloc[0] * 2 + 1;
1008 occurs_table[0] = xrealloc (occurs_table[0],
1009 occurs_alloc[0] * sizeof *occurs_table[0]);
1012 occurs_cursor = occurs_table[0] + number_of_occurs[0];
1014 /* Define the refence field, if any. */
1019 /* While auto referencing, update `line_start' from its
1020 previous value, counting lines as we go. If input
1021 referencing at the same time, `line_start' has been
1022 advanced earlier, and the following loop is never really
1025 while (line_scan < possible_key.start)
1026 if (*line_scan == '\n')
1030 line_start = line_scan;
1031 SKIP_NON_WHITE (line_scan, text_buffer.end);
1036 occurs_cursor->reference = total_line_count;
1038 else if (input_reference)
1041 /* If only input referencing, `line_start' has been computed
1042 earlier to detect the case the word matched would be part
1043 of the reference. The reference position is simply the
1044 value of `line_start'. */
1046 occurs_cursor->reference
1047 = (DELTA) (line_start - possible_key.start);
1048 if (reference_length > reference_max_width)
1049 reference_max_width = reference_length;
1052 /* Exclude the reference from the context in simple cases. */
1054 if (input_reference && line_start == context_start)
1056 SKIP_NON_WHITE (context_start, context_end);
1057 SKIP_WHITE (context_start, context_end);
1060 /* Completes the OCCURS structure. */
1062 occurs_cursor->key = possible_key;
1063 occurs_cursor->left = context_start - possible_key.start;
1064 occurs_cursor->right = context_end - possible_key.start;
1066 number_of_occurs[0]++;
1071 /* Formatting and actual output - service routines. */
1073 /*-----------------------------------------.
1074 | Prints some NUMBER of spaces on stdout. |
1075 `-----------------------------------------*/
1078 print_spaces (int number)
1082 for (counter = number; counter > 0; counter--)
1086 /*-------------------------------------.
1087 | Prints the field provided by FIELD. |
1088 `-------------------------------------*/
1091 print_field (BLOCK field)
1093 char *cursor; /* Cursor in field to print */
1094 int base; /* Base character, without diacritic */
1095 int diacritic; /* Diacritic code for the character */
1097 /* Whitespace is not really compressed. Instead, each white space
1098 character (tab, vt, ht etc.) is printed as one single space. */
1100 for (cursor = field.start; cursor < field.end; cursor++)
1102 unsigned char character = *cursor;
1103 if (edited_flag[character])
1106 /* First check if this is a diacriticized character.
1108 This works only for TeX. I do not know how diacriticized
1109 letters work with `roff'. Please someone explain it to me! */
1111 diacritic = todiac (character);
1112 if (diacritic != 0 && output_format == TEX_FORMAT)
1114 base = tobase (character);
1118 case 1: /* Latin diphthongs */
1122 fputs ("\\oe{}", stdout);
1126 fputs ("\\OE{}", stdout);
1130 fputs ("\\ae{}", stdout);
1134 fputs ("\\AE{}", stdout);
1142 case 2: /* Acute accent */
1143 printf ("\\'%s%c", (base == 'i' ? "\\" : ""), base);
1146 case 3: /* Grave accent */
1147 printf ("\\`%s%c", (base == 'i' ? "\\" : ""), base);
1150 case 4: /* Circumflex accent */
1151 printf ("\\^%s%c", (base == 'i' ? "\\" : ""), base);
1154 case 5: /* Diaeresis */
1155 printf ("\\\"%s%c", (base == 'i' ? "\\" : ""), base);
1158 case 6: /* Tilde accent */
1159 printf ("\\~%s%c", (base == 'i' ? "\\" : ""), base);
1162 case 7: /* Cedilla */
1163 printf ("\\c{%c}", base);
1166 case 8: /* Small circle beneath */
1170 fputs ("\\aa{}", stdout);
1174 fputs ("\\AA{}", stdout);
1182 case 9: /* Strike through */
1186 fputs ("\\o{}", stdout);
1190 fputs ("\\O{}", stdout);
1201 /* This is not a diacritic character, so handle cases which are
1202 really specific to `roff' or TeX. All white space processing
1203 is done as the default case of this switch. */
1208 /* In roff output format, double any quote. */
1218 /* In TeX output format, precede these with a backslash. */
1220 putchar (character);
1225 /* In TeX output format, precede these with a backslash and
1226 force mathematical mode. */
1227 printf ("$\\%c$", character);
1231 /* In TeX output mode, request production of a backslash. */
1232 fputs ("\\backslash{}", stdout);
1236 /* Any other flagged character produces a single space. */
1245 /* Formatting and actual output - planning routines. */
1247 /*--------------------------------------------------------------------.
1248 | From information collected from command line options and input file |
1249 | readings, compute and fix some output parameter values. |
1250 `--------------------------------------------------------------------*/
1253 fix_output_parameters (void)
1255 int file_index; /* index in text input file arrays */
1256 int line_ordinal; /* line ordinal value for reference */
1257 char ordinal_string[12]; /* edited line ordinal for reference */
1258 int reference_width; /* width for the whole reference */
1259 int character; /* character ordinal */
1260 const char *cursor; /* cursor in some constant strings */
1262 /* In auto reference mode, the maximum width of this field is
1263 precomputed and subtracted from the overall line width. Add one for
1264 the column which separate the file name from the line number. */
1268 reference_max_width = 0;
1269 for (file_index = 0; file_index < number_input_files; file_index++)
1271 line_ordinal = file_line_count[file_index] + 1;
1273 line_ordinal -= file_line_count[file_index - 1];
1274 sprintf (ordinal_string, "%d", line_ordinal);
1275 reference_width = strlen (ordinal_string);
1276 if (input_file_name[file_index])
1277 reference_width += strlen (input_file_name[file_index]);
1278 if (reference_width > reference_max_width)
1279 reference_max_width = reference_width;
1281 reference_max_width++;
1282 reference.start = xmalloc ((size_t) reference_max_width + 1);
1285 /* If the reference appears to the left of the output line, reserve some
1286 space for it right away, including one gap size. */
1288 if ((auto_reference | input_reference) & !right_reference)
1289 line_width -= reference_max_width + gap_size;
1291 /* The output lines, minimally, will contain from left to right a left
1292 context, a gap, and a keyword followed by the right context with no
1293 special intervening gap. Half of the line width is dedicated to the
1294 left context and the gap, the other half is dedicated to the keyword
1295 and the right context; these values are computed once and for all here.
1296 There also are tail and head wrap around fields, used when the keyword
1297 is near the beginning or the end of the line, or when some long word
1298 cannot fit in, but leave place from wrapped around shorter words. The
1299 maximum width of these fields are recomputed separately for each line,
1300 on a case by case basis. It is worth noting that it cannot happen that
1301 both the tail and head fields are used at once. */
1303 half_line_width = line_width / 2;
1304 before_max_width = half_line_width - gap_size;
1305 keyafter_max_width = half_line_width;
1307 /* If truncation_string is the empty string, make it NULL to speed up
1308 tests. In this case, truncation_string_length will never get used, so
1309 there is no need to set it. */
1311 if (truncation_string && *truncation_string)
1312 truncation_string_length = strlen (truncation_string);
1314 truncation_string = NULL;
1319 /* When flagging truncation at the left of the keyword, the
1320 truncation mark goes at the beginning of the before field,
1321 unless there is a head field, in which case the mark goes at the
1322 left of the head field. When flagging truncation at the right
1323 of the keyword, the mark goes at the end of the keyafter field,
1324 unless there is a tail field, in which case the mark goes at the
1325 end of the tail field. Only eight combination cases could arise
1326 for truncation marks:
1329 . One beginning the before field.
1330 . One beginning the head field.
1331 . One ending the keyafter field.
1332 . One ending the tail field.
1333 . One beginning the before field, another ending the keyafter field.
1334 . One ending the tail field, another beginning the before field.
1335 . One ending the keyafter field, another beginning the head field.
1337 So, there is at most two truncation marks, which could appear both
1338 on the left side of the center of the output line, both on the
1339 right side, or one on either side. */
1341 before_max_width -= 2 * truncation_string_length;
1342 keyafter_max_width -= 2 * truncation_string_length;
1347 /* I never figured out exactly how UNIX' ptx plans the output width
1348 of its various fields. If GNU extensions are disabled, do not
1349 try computing the field widths correctly; instead, use the
1350 following formula, which does not completely imitate UNIX' ptx,
1353 keyafter_max_width -= 2 * truncation_string_length + 1;
1356 /* Compute which characters need special output processing. Initialize
1357 by flagging any white space character. Some systems do not consider
1358 form feed as a space character, but we do. */
1360 for (character = 0; character < CHAR_SET_SIZE; character++)
1361 edited_flag[character] = ISSPACE (character) != 0;
1362 edited_flag['\f'] = 1;
1364 /* Complete the special character flagging according to selected output
1367 switch (output_format)
1369 case UNKNOWN_FORMAT:
1370 /* Should never happen. */
1377 /* `Quote' characters should be doubled. */
1379 edited_flag['"'] = 1;
1384 /* Various characters need special processing. */
1386 for (cursor = "$%&#_{}\\"; *cursor; cursor++)
1387 edited_flag[to_uchar (*cursor)] = 1;
1389 /* Any character with 8th bit set will print to a single space, unless
1390 it is diacriticized. */
1392 for (character = 0200; character < CHAR_SET_SIZE; character++)
1393 edited_flag[character] = todiac (character) != 0;
1398 /*------------------------------------------------------------------.
1399 | Compute the position and length of all the output fields, given a |
1400 | pointer to some OCCURS. |
1401 `------------------------------------------------------------------*/
1404 define_all_fields (OCCURS *occurs)
1406 int tail_max_width; /* allowable width of tail field */
1407 int head_max_width; /* allowable width of head field */
1408 char *cursor; /* running cursor in source text */
1409 char *left_context_start; /* start of left context */
1410 char *right_context_end; /* end of right context */
1411 char *left_field_start; /* conservative start for `head'/`before' */
1412 int file_index; /* index in text input file arrays */
1413 const char *file_name; /* file name for reference */
1414 int line_ordinal; /* line ordinal for reference */
1416 /* Define `keyafter', start of left context and end of right context.
1417 `keyafter' starts at the saved position for keyword and extend to the
1418 right from the end of the keyword, eating separators or full words, but
1419 not beyond maximum allowed width for `keyafter' field or limit for the
1420 right context. Suffix spaces will be removed afterwards. */
1422 keyafter.start = occurs->key.start;
1423 keyafter.end = keyafter.start + occurs->key.size;
1424 left_context_start = keyafter.start + occurs->left;
1425 right_context_end = keyafter.start + occurs->right;
1427 cursor = keyafter.end;
1428 while (cursor < right_context_end
1429 && cursor <= keyafter.start + keyafter_max_width)
1431 keyafter.end = cursor;
1432 SKIP_SOMETHING (cursor, right_context_end);
1434 if (cursor <= keyafter.start + keyafter_max_width)
1435 keyafter.end = cursor;
1437 keyafter_truncation = truncation_string && keyafter.end < right_context_end;
1439 SKIP_WHITE_BACKWARDS (keyafter.end, keyafter.start);
1441 /* When the left context is wide, it might take some time to catch up from
1442 the left context boundary to the beginning of the `head' or `before'
1443 fields. So, in this case, to speed the catchup, we jump back from the
1444 keyword, using some secure distance, possibly falling in the middle of
1445 a word. A secure backward jump would be at least half the maximum
1446 width of a line, plus the size of the longest word met in the whole
1447 input. We conclude this backward jump by a skip forward of at least
1448 one word. In this manner, we should not inadvertently accept only part
1449 of a word. From the reached point, when it will be time to fix the
1450 beginning of `head' or `before' fields, we will skip forward words or
1451 delimiters until we get sufficiently near. */
1453 if (-occurs->left > half_line_width + maximum_word_length)
1456 = keyafter.start - (half_line_width + maximum_word_length);
1457 SKIP_SOMETHING (left_field_start, keyafter.start);
1460 left_field_start = keyafter.start + occurs->left;
1462 /* `before' certainly ends at the keyword, but not including separating
1463 spaces. It starts after than the saved value for the left context, by
1464 advancing it until it falls inside the maximum allowed width for the
1465 before field. There will be no prefix spaces either. `before' only
1466 advances by skipping single separators or whole words. */
1468 before.start = left_field_start;
1469 before.end = keyafter.start;
1470 SKIP_WHITE_BACKWARDS (before.end, before.start);
1472 while (before.start + before_max_width < before.end)
1473 SKIP_SOMETHING (before.start, before.end);
1475 if (truncation_string)
1477 cursor = before.start;
1478 SKIP_WHITE_BACKWARDS (cursor, text_buffer.start);
1479 before_truncation = cursor > left_context_start;
1482 before_truncation = 0;
1484 SKIP_WHITE (before.start, text_buffer.end);
1486 /* The tail could not take more columns than what has been left in the
1487 left context field, and a gap is mandatory. It starts after the
1488 right context, and does not contain prefixed spaces. It ends at
1489 the end of line, the end of buffer or when the tail field is full,
1490 whichever comes first. It cannot contain only part of a word, and
1491 has no suffixed spaces. */
1494 = before_max_width - (before.end - before.start) - gap_size;
1496 if (tail_max_width > 0)
1498 tail.start = keyafter.end;
1499 SKIP_WHITE (tail.start, text_buffer.end);
1501 tail.end = tail.start;
1503 while (cursor < right_context_end
1504 && cursor < tail.start + tail_max_width)
1507 SKIP_SOMETHING (cursor, right_context_end);
1510 if (cursor < tail.start + tail_max_width)
1513 if (tail.end > tail.start)
1515 keyafter_truncation = 0;
1516 tail_truncation = truncation_string && tail.end < right_context_end;
1519 tail_truncation = 0;
1521 SKIP_WHITE_BACKWARDS (tail.end, tail.start);
1526 /* No place left for a tail field. */
1530 tail_truncation = 0;
1533 /* `head' could not take more columns than what has been left in the right
1534 context field, and a gap is mandatory. It ends before the left
1535 context, and does not contain suffixed spaces. Its pointer is advanced
1536 until the head field has shrunk to its allowed width. It cannot
1537 contain only part of a word, and has no suffixed spaces. */
1540 = keyafter_max_width - (keyafter.end - keyafter.start) - gap_size;
1542 if (head_max_width > 0)
1544 head.end = before.start;
1545 SKIP_WHITE_BACKWARDS (head.end, text_buffer.start);
1547 head.start = left_field_start;
1548 while (head.start + head_max_width < head.end)
1549 SKIP_SOMETHING (head.start, head.end);
1551 if (head.end > head.start)
1553 before_truncation = 0;
1554 head_truncation = (truncation_string
1555 && head.start > left_context_start);
1558 head_truncation = 0;
1560 SKIP_WHITE (head.start, head.end);
1565 /* No place left for a head field. */
1569 head_truncation = 0;
1575 /* Construct the reference text in preallocated space from the file
1576 name and the line number. Find out in which file the reference
1577 occurred. Standard input yields an empty file name. Insure line
1578 numbers are one based, even if they are computed zero based. */
1581 while (file_line_count[file_index] < occurs->reference)
1584 file_name = input_file_name[file_index];
1588 line_ordinal = occurs->reference + 1;
1590 line_ordinal -= file_line_count[file_index - 1];
1592 sprintf (reference.start, "%s:%d", file_name, line_ordinal);
1593 reference.end = reference.start + strlen (reference.start);
1595 else if (input_reference)
1598 /* Reference starts at saved position for reference and extends right
1599 until some white space is met. */
1601 reference.start = keyafter.start + (DELTA) occurs->reference;
1602 reference.end = reference.start;
1603 SKIP_NON_WHITE (reference.end, right_context_end);
1607 /* Formatting and actual output - control routines. */
1609 /*----------------------------------------------------------------------.
1610 | Output the current output fields as one line for `troff' or `nroff'. |
1611 `----------------------------------------------------------------------*/
1614 output_one_roff_line (void)
1616 /* Output the `tail' field. */
1618 printf (".%s \"", macro_name);
1620 if (tail_truncation)
1621 fputs (truncation_string, stdout);
1624 /* Output the `before' field. */
1626 fputs (" \"", stdout);
1627 if (before_truncation)
1628 fputs (truncation_string, stdout);
1629 print_field (before);
1632 /* Output the `keyafter' field. */
1634 fputs (" \"", stdout);
1635 print_field (keyafter);
1636 if (keyafter_truncation)
1637 fputs (truncation_string, stdout);
1640 /* Output the `head' field. */
1642 fputs (" \"", stdout);
1643 if (head_truncation)
1644 fputs (truncation_string, stdout);
1648 /* Conditionally output the `reference' field. */
1650 if (auto_reference | input_reference)
1652 fputs (" \"", stdout);
1653 print_field (reference);
1660 /*---------------------------------------------------------.
1661 | Output the current output fields as one line for `TeX'. |
1662 `---------------------------------------------------------*/
1665 output_one_tex_line (void)
1667 BLOCK key; /* key field, isolated */
1668 BLOCK after; /* after field, isolated */
1669 char *cursor; /* running cursor in source text */
1671 printf ("\\%s ", macro_name);
1674 fputs ("}{", stdout);
1675 print_field (before);
1676 fputs ("}{", stdout);
1677 key.start = keyafter.start;
1678 after.end = keyafter.end;
1679 cursor = keyafter.start;
1680 SKIP_SOMETHING (cursor, keyafter.end);
1682 after.start = cursor;
1684 fputs ("}{", stdout);
1685 print_field (after);
1686 fputs ("}{", stdout);
1689 if (auto_reference | input_reference)
1692 print_field (reference);
1698 /*-------------------------------------------------------------------.
1699 | Output the current output fields as one line for a dumb terminal. |
1700 `-------------------------------------------------------------------*/
1703 output_one_dumb_line (void)
1705 if (!right_reference)
1710 /* Output the `reference' field, in such a way that GNU emacs
1711 next-error will handle it. The ending colon is taken from the
1712 gap which follows. */
1714 print_field (reference);
1716 print_spaces (reference_max_width
1718 - (reference.end - reference.start)
1724 /* Output the `reference' field and its following gap. */
1726 print_field (reference);
1727 print_spaces (reference_max_width
1729 - (reference.end - reference.start));
1733 if (tail.start < tail.end)
1735 /* Output the `tail' field. */
1738 if (tail_truncation)
1739 fputs (truncation_string, stdout);
1741 print_spaces (half_line_width - gap_size
1742 - (before.end - before.start)
1743 - (before_truncation ? truncation_string_length : 0)
1744 - (tail.end - tail.start)
1745 - (tail_truncation ? truncation_string_length : 0));
1748 print_spaces (half_line_width - gap_size
1749 - (before.end - before.start)
1750 - (before_truncation ? truncation_string_length : 0));
1752 /* Output the `before' field. */
1754 if (before_truncation)
1755 fputs (truncation_string, stdout);
1756 print_field (before);
1758 print_spaces (gap_size);
1760 /* Output the `keyafter' field. */
1762 print_field (keyafter);
1763 if (keyafter_truncation)
1764 fputs (truncation_string, stdout);
1766 if (head.start < head.end)
1768 /* Output the `head' field. */
1770 print_spaces (half_line_width
1771 - (keyafter.end - keyafter.start)
1772 - (keyafter_truncation ? truncation_string_length : 0)
1773 - (head.end - head.start)
1774 - (head_truncation ? truncation_string_length : 0));
1775 if (head_truncation)
1776 fputs (truncation_string, stdout);
1781 if ((auto_reference | input_reference) & right_reference)
1782 print_spaces (half_line_width
1783 - (keyafter.end - keyafter.start)
1784 - (keyafter_truncation ? truncation_string_length : 0));
1786 if ((auto_reference | input_reference) & right_reference)
1788 /* Output the `reference' field. */
1790 print_spaces (gap_size);
1791 print_field (reference);
1797 /*------------------------------------------------------------------------.
1798 | Scan the whole occurs table and, for each entry, output one line in the |
1799 | appropriate format. |
1800 `------------------------------------------------------------------------*/
1803 generate_all_output (void)
1805 size_t occurs_index; /* index of keyword entry being processed */
1806 OCCURS *occurs_cursor; /* current keyword entry being processed */
1808 /* The following assignments are useful to provide default values in case
1809 line contexts or references are not used, in which case these variables
1810 would never be computed. */
1814 tail_truncation = 0;
1818 head_truncation = 0;
1820 /* Loop over all keyword occurrences. */
1822 occurs_cursor = occurs_table[0];
1824 for (occurs_index = 0; occurs_index < number_of_occurs[0]; occurs_index++)
1826 /* Compute the exact size of every field and whenever truncation flags
1827 are present or not. */
1829 define_all_fields (occurs_cursor);
1831 /* Produce one output line according to selected format. */
1833 switch (output_format)
1835 case UNKNOWN_FORMAT:
1836 /* Should never happen. */
1839 output_one_dumb_line ();
1843 output_one_roff_line ();
1847 output_one_tex_line ();
1851 /* Advance the cursor into the occurs table. */
1857 /* Option decoding and main program. */
1859 /*------------------------------------------------------.
1860 | Print program identification and options, then exit. |
1861 `------------------------------------------------------*/
1866 if (status != EXIT_SUCCESS)
1867 fprintf (stderr, _("Try `%s --help' for more information.\n"),
1872 Usage: %s [OPTION]... [INPUT]... (without -G)\n\
1873 or: %s -G [OPTION]... [INPUT [OUTPUT]]\n"),
1874 program_name, program_name);
1876 Output a permuted index, including context, of the words in the input files.\n\
1880 Mandatory arguments to long options are mandatory for short options too.\n\
1883 -A, --auto-reference output automatically generated references\n\
1884 -C, --copyright display Copyright and copying conditions\n\
1885 -G, --traditional behave more like System V `ptx'\n\
1886 -F, --flag-truncation=STRING use STRING for flagging line truncations\n\
1889 -M, --macro-name=STRING macro name to use instead of `xx'\n\
1890 -O, --format=roff generate output as roff directives\n\
1891 -R, --right-side-refs put references at right, not counted in -w\n\
1892 -S, --sentence-regexp=REGEXP for end of lines or end of sentences\n\
1893 -T, --format=tex generate output as TeX directives\n\
1896 -W, --word-regexp=REGEXP use REGEXP to match each keyword\n\
1897 -b, --break-file=FILE word break characters in this FILE\n\
1898 -f, --ignore-case fold lower case to upper case for sorting\n\
1899 -g, --gap-size=NUMBER gap size in columns between output fields\n\
1900 -i, --ignore-file=FILE read ignore word list from FILE\n\
1901 -o, --only-file=FILE read only word list from this FILE\n\
1904 -r, --references first field of each line is a reference\n\
1905 -t, --typeset-mode - not implemented -\n\
1906 -w, --width=NUMBER output width in columns, reference excluded\n\
1908 fputs (HELP_OPTION_DESCRIPTION, stdout);
1909 fputs (VERSION_OPTION_DESCRIPTION, stdout);
1912 With no FILE or if FILE is -, read Standard Input. `-F /' by default.\n\
1914 printf (_("\nReport bugs to <%s>.\n"), PACKAGE_BUGREPORT);
1919 /*----------------------------------------------------------------------.
1920 | Main program. Decode ARGC arguments passed through the ARGV array of |
1921 | strings, then launch execution. |
1922 `----------------------------------------------------------------------*/
1924 /* Long options equivalences. */
1925 static const struct option long_options[] =
1927 {"auto-reference", no_argument, NULL, 'A'},
1928 {"break-file", required_argument, NULL, 'b'},
1929 {"copyright", no_argument, NULL, 'C'},
1930 {"flag-truncation", required_argument, NULL, 'F'},
1931 {"ignore-case", no_argument, NULL, 'f'},
1932 {"gap-size", required_argument, NULL, 'g'},
1933 {"ignore-file", required_argument, NULL, 'i'},
1934 {"macro-name", required_argument, NULL, 'M'},
1935 {"only-file", required_argument, NULL, 'o'},
1936 {"references", no_argument, NULL, 'r'},
1937 {"right-side-refs", no_argument, NULL, 'R'},
1938 {"format", required_argument, NULL, 10},
1939 {"sentence-regexp", required_argument, NULL, 'S'},
1940 {"traditional", no_argument, NULL, 'G'},
1941 {"typeset-mode", no_argument, NULL, 't'},
1942 {"width", required_argument, NULL, 'w'},
1943 {"word-regexp", required_argument, NULL, 'W'},
1944 {GETOPT_HELP_OPTION_DECL},
1945 {GETOPT_VERSION_OPTION_DECL},
1949 static char const* const format_args[] =
1954 static enum Format const format_vals[] =
1956 ROFF_FORMAT, TEX_FORMAT
1960 main (int argc, char **argv)
1962 int optchar; /* argument character */
1963 int file_index; /* index in text input file arrays */
1965 /* Decode program options. */
1967 initialize_main (&argc, &argv);
1968 program_name = argv[0];
1969 setlocale (LC_ALL, "");
1970 bindtextdomain (PACKAGE, LOCALEDIR);
1971 textdomain (PACKAGE);
1973 atexit (close_stdout);
1975 #if HAVE_SETCHRCLASS
1979 while (optchar = getopt_long (argc, argv, "ACF:GM:ORS:TW:b:i:fg:o:trw:",
1980 long_options, NULL),
1986 usage (EXIT_FAILURE);
1990 This program is free software; you can redistribute it and/or modify\n\
1991 it under the terms of the GNU General Public License as published by\n\
1992 the Free Software Foundation; either version 2, or (at your option)\n\
1993 any later version.\n\
1997 This program is distributed in the hope that it will be useful,\n\
1998 but WITHOUT ANY WARRANTY; without even the implied warranty of\n\
1999 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the\n\
2000 GNU General Public License for more details.\n\
2004 You should have received a copy of the GNU General Public License\n\
2005 along with this program; if not, write to the Free Software Foundation,\n\
2006 Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.\n"),
2009 exit (EXIT_SUCCESS);
2012 gnu_extensions = false;
2016 break_file = optarg;
2025 unsigned long int tmp_ulong;
2026 if (xstrtoul (optarg, NULL, 0, &tmp_ulong, NULL) != LONGINT_OK
2027 || ! (0 < tmp_ulong && tmp_ulong <= INT_MAX))
2028 error (EXIT_FAILURE, 0, _("invalid gap width: %s"),
2030 gap_size = tmp_ulong;
2035 ignore_file = optarg;
2043 input_reference = true;
2047 /* Yet to understand... */
2052 unsigned long int tmp_ulong;
2053 if (xstrtoul (optarg, NULL, 0, &tmp_ulong, NULL) != LONGINT_OK
2054 || ! (0 < tmp_ulong && tmp_ulong <= INT_MAX))
2055 error (EXIT_FAILURE, 0, _("invalid line width: %s"),
2057 line_width = tmp_ulong;
2062 auto_reference = true;
2066 truncation_string = copy_unescaped_string (optarg);
2070 macro_name = optarg;
2074 output_format = ROFF_FORMAT;
2078 right_reference = true;
2082 context_regex_string = copy_unescaped_string (optarg);
2086 output_format = TEX_FORMAT;
2090 word_regex_string = copy_unescaped_string (optarg);
2094 output_format = XARGMATCH ("--format", optarg,
2095 format_args, format_vals);
2096 case_GETOPT_HELP_CHAR;
2098 case_GETOPT_VERSION_CHAR (PROGRAM_NAME, AUTHORS);
2102 /* Change the default Ignore file if one is defined. */
2104 #ifdef DEFAULT_IGNORE_FILE
2106 ignore_file = DEFAULT_IGNORE_FILE;
2109 /* Process remaining arguments. If GNU extensions are enabled, process
2110 all arguments as input parameters. If disabled, accept at most two
2111 arguments, the second of which is an output parameter. */
2116 /* No more argument simply means: read standard input. */
2118 input_file_name = xmalloc (sizeof *input_file_name);
2119 file_line_count = xmalloc (sizeof *file_line_count);
2120 number_input_files = 1;
2121 input_file_name[0] = NULL;
2123 else if (gnu_extensions)
2125 number_input_files = argc - optind;
2126 input_file_name = xmalloc (number_input_files * sizeof *input_file_name);
2127 file_line_count = xmalloc (number_input_files * sizeof *file_line_count);
2129 for (file_index = 0; file_index < number_input_files; file_index++)
2131 input_file_name[file_index] = argv[optind];
2132 if (!*argv[optind] || STREQ (argv[optind], "-"))
2133 input_file_name[0] = NULL;
2135 input_file_name[0] = argv[optind];
2142 /* There is one necessary input file. */
2144 number_input_files = 1;
2145 input_file_name = xmalloc (sizeof *input_file_name);
2146 file_line_count = xmalloc (sizeof *file_line_count);
2147 if (!*argv[optind] || STREQ (argv[optind], "-"))
2148 input_file_name[0] = NULL;
2150 input_file_name[0] = argv[optind];
2153 /* Redirect standard output, only if requested. */
2157 /* FIXME: don't fclose here? */
2159 if (fopen (argv[optind], "w") == NULL)
2160 error (EXIT_FAILURE, errno, "%s", argv[optind]);
2164 /* Diagnose any other argument as an error. */
2168 error (0, 0, _("extra operand %s"), quote (argv[optind]));
2169 usage (EXIT_FAILURE);
2173 /* If the output format has not been explicitly selected, choose dumb
2174 terminal format if GNU extensions are enabled, else `roff' format. */
2176 if (output_format == UNKNOWN_FORMAT)
2177 output_format = gnu_extensions ? DUMB_FORMAT : ROFF_FORMAT;
2179 /* Initialize the main tables. */
2181 initialize_regex ();
2183 /* Read `Break character' file, if any. */
2186 digest_break_file (break_file);
2188 /* Read `Ignore words' file and `Only words' files, if any. If any of
2189 these files is empty, reset the name of the file to NULL, to avoid
2190 unnecessary calls to search_table. */
2194 digest_word_file (ignore_file, &ignore_table);
2195 if (ignore_table.length == 0)
2201 digest_word_file (only_file, &only_table);
2202 if (only_table.length == 0)
2206 /* Prepare to study all the input files. */
2208 number_of_occurs[0] = 0;
2209 total_line_count = 0;
2210 maximum_word_length = 0;
2211 reference_max_width = 0;
2213 for (file_index = 0; file_index < number_input_files; file_index++)
2216 /* Read the file in core, than study it. */
2218 swallow_file_in_memory (input_file_name[file_index], &text_buffer);
2219 find_occurs_in_text ();
2221 /* Maintain for each file how many lines has been read so far when its
2222 end is reached. Incrementing the count first is a simple kludge to
2223 handle a possible incomplete line at end of file. */
2226 file_line_count[file_index] = total_line_count;
2229 /* Do the output process phase. */
2231 sort_found_occurs ();
2232 fix_output_parameters ();
2233 generate_all_output ();
2237 exit (EXIT_SUCCESS);