1 /* GNU fmt -- simple text formatter.
2 Copyright (C) 1994-2001 Free Software Foundation, Inc.
4 This program is free software; you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation; either version 2, or (at your option)
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program; if not, write to the Free Software Foundation,
16 Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */
18 /* Written by Ross Paterson <rap@doc.ic.ac.uk>. */
22 #include <sys/types.h>
25 /* Redefine. Otherwise, systems (Unicos for one) with headers that define
26 it to be a type get syntax errors for the variable declaration below. */
27 #define word unused_word_type
34 /* The official name of this program (e.g., no `g' prefix). */
35 #define PROGRAM_NAME "fmt"
37 #define AUTHORS "Ross Paterson"
39 /* The following parameters represent the program's idea of what is
40 "best". Adjust to taste, subject to the caveats given. */
42 /* Default longest permitted line length (max_width). */
45 /* Prefer lines to be LEEWAY % shorter than the maximum width, giving
46 room for optimization. */
49 /* The default secondary indent of tagged paragraph used for unindented
50 one-line paragraphs not preceded by any multi-line paragraphs. */
53 /* Costs and bonuses are expressed as the equivalent departure from the
54 optimal line length, multiplied by 10. e.g. assigning something a
55 cost of 50 means that it is as bad as a line 5 characters too short
56 or too long. The definition of SHORT_COST(n) should not be changed.
57 However, EQUIV(n) may need tuning. */
61 #define MAXCOST TYPE_MAXIMUM (COST)
63 #define SQR(n) ((n) * (n))
64 #define EQUIV(n) SQR ((COST) (n))
66 /* Cost of a filled line n chars longer or shorter than best_width. */
67 #define SHORT_COST(n) EQUIV ((n) * 10)
69 /* Cost of the difference between adjacent filled lines. */
70 #define RAGGED_COST(n) (SHORT_COST (n) / 2)
72 /* Basic cost per line. */
73 #define LINE_COST EQUIV (70)
75 /* Cost of breaking a line after the first word of a sentence, where
76 the length of the word is N. */
77 #define WIDOW_COST(n) (EQUIV (200) / ((n) + 2))
79 /* Cost of breaking a line before the last word of a sentence, where
80 the length of the word is N. */
81 #define ORPHAN_COST(n) (EQUIV (150) / ((n) + 2))
83 /* Bonus for breaking a line at the end of a sentence. */
84 #define SENTENCE_BONUS EQUIV (50)
86 /* Cost of breaking a line after a period not marking end of a sentence.
87 With the definition of sentence we are using (borrowed from emacs, see
88 get_line()) such a break would then look like a sentence break. Hence
89 we assign a very high cost -- it should be avoided unless things are
91 #define NOBREAK_COST EQUIV (600)
93 /* Bonus for breaking a line before open parenthesis. */
94 #define PAREN_BONUS EQUIV (40)
96 /* Bonus for breaking a line after other punctuation. */
97 #define PUNCT_BONUS EQUIV(40)
99 /* Credit for breaking a long paragraph one line later. */
100 #define LINE_CREDIT EQUIV(3)
102 /* Size of paragraph buffer, in words and characters. Longer paragraphs
103 are handled neatly (cf. flush_paragraph()), so there's little to gain
104 by making these larger. */
105 #define MAXWORDS 1000
106 #define MAXCHARS 5000
108 /* Extra ctype(3)-style macros. */
110 #define isopen(c) (strchr ("([`'\"", c) != NULL)
111 #define isclose(c) (strchr (")]'\"", c) != NULL)
112 #define isperiod(c) (strchr (".?!", c) != NULL)
114 /* Size of a tab stop, for expansion on input and re-introduction on
118 /* Miscellaneous definitions. */
125 /* Word descriptor structure. */
127 typedef struct Word WORD;
132 /* Static attributes determined during input. */
134 const char *text; /* the text of the word */
135 int length; /* length of this word */
136 int space; /* the size of the following space */
137 bool paren:1; /* starts with open paren */
138 bool period:1; /* ends in [.?!])* */
139 bool punct:1; /* ends in punctuation */
140 bool final:1; /* end of sentence */
142 /* The remaining fields are computed during the optimization. */
144 int line_length; /* length of the best line starting here */
145 COST best_cost; /* cost of best paragraph starting here */
146 WORD *next_break; /* break which achieves best_cost */
149 /* Forward declarations. */
151 static void set_prefix PARAMS ((char *p));
152 static void fmt PARAMS ((FILE *f));
153 static bool get_paragraph PARAMS ((FILE *f));
154 static int get_line PARAMS ((FILE *f, int c));
155 static int get_prefix PARAMS ((FILE *f));
156 static int get_space PARAMS ((FILE *f, int c));
157 static int copy_rest PARAMS ((FILE *f, int c));
158 static bool same_para PARAMS ((int c));
159 static void flush_paragraph PARAMS ((void));
160 static void fmt_paragraph PARAMS ((void));
161 static void check_punctuation PARAMS ((WORD *w));
162 static COST base_cost PARAMS ((WORD *this));
163 static COST line_cost PARAMS ((WORD *next, int len));
164 static void put_paragraph PARAMS ((WORD *finish));
165 static void put_line PARAMS ((WORD *w, int indent));
166 static void put_word PARAMS ((WORD *w));
167 static void put_space PARAMS ((int space));
169 /* The name this program was run with. */
170 const char *program_name;
174 /* If TRUE, first 2 lines may have different indent (default FALSE). */
177 /* If TRUE, first 2 lines _must_ have different indent (default FALSE). */
180 /* If TRUE, each line is a paragraph on its own (default FALSE). */
183 /* If TRUE, don't preserve inter-word spacing (default FALSE). */
186 /* Prefix minus leading and trailing spaces (default ""). */
187 static const char *prefix;
189 /* User-supplied maximum line width (default WIDTH). The only output
190 lines longer than this will each comprise a single word. */
191 static int max_width;
193 /* Values derived from the option values. */
195 /* The length of prefix minus leading space. */
196 static int prefix_full_length;
198 /* The length of the leading space trimmed from the prefix. */
199 static int prefix_lead_space;
201 /* The length of prefix minus leading and trailing space. */
202 static int prefix_length;
204 /* The preferred width of text lines, set to LEEWAY % less than max_width. */
205 static int best_width;
207 /* Dynamic variables. */
209 /* Start column of the character most recently read from the input file. */
210 static int in_column;
212 /* Start column of the next character to be written to stdout. */
213 static int out_column;
215 /* Space for the paragraph text -- longer paragraphs are handled neatly
216 (cf. flush_paragraph()). */
217 static char parabuf[MAXCHARS];
219 /* A pointer into parabuf, indicating the first unused character position. */
222 /* The words of a paragraph -- longer paragraphs are handled neatly
223 (cf. flush_paragraph()). */
224 static WORD word[MAXWORDS];
226 /* A pointer into the above word array, indicating the first position
227 after the last complete word. Sometimes it will point at an incomplete
229 static WORD *word_limit;
231 /* If TRUE, current input file contains tab characters, and so tabs can be
232 used for white space on output. */
235 /* Space before trimmed prefix on each line of the current paragraph. */
236 static int prefix_indent;
238 /* Indentation of the first line of the current paragraph. */
239 static int first_indent;
241 /* Indentation of other lines of the current paragraph */
242 static int other_indent;
244 /* To detect the end of a paragraph, we need to look ahead to the first
245 non-blank character after the prefix on the next line, or the first
246 character on the following line that failed to match the prefix.
247 We can reconstruct the lookahead from that character (next_char), its
248 position on the line (in_column) and the amount of space before the
249 prefix (next_prefix_indent). See get_paragraph() and copy_rest(). */
251 /* The last character read from the input file. */
252 static int next_char;
254 /* The space before the trimmed prefix (or part of it) on the next line
255 after the current paragraph. */
256 static int next_prefix_indent;
258 /* If nonzero, the length of the last line output in the current
259 paragraph, used to charge for raggedness at the split point for long
260 paragraphs chosen by fmt_paragraph(). */
261 static int last_line_length;
267 fprintf (stderr, _("Try `%s --help' for more information.\n"),
271 printf (_("Usage: %s [-DIGITS] [OPTION]... [FILE]...\n"), program_name);
273 Reformat each paragraph in the FILE(s), writing to standard output.\n\
274 If no FILE or if FILE is `-', read standard input.\n\
276 Mandatory arguments to long options are mandatory for short options too.\n\
277 -c, --crown-margin preserve indentation of first two lines\n\
278 -p, --prefix=STRING combine only lines having STRING as prefix\n\
279 -s, --split-only split long lines, but do not refill\n\
283 -t, --tagged-paragraph indentation of first line different from second\n\
284 -u, --uniform-spacing one space between words, two after sentences\n\
285 -w, --width=NUMBER maximum line width (default of 75 columns)\n\
286 --help display this help and exit\n\
287 --version output version information and exit\n\
289 In -wNUMBER, the letter `w' may be omitted.\n"),
291 puts (_("\nReport bugs to <bug-textutils@gnu.org>."));
293 exit (status == 0 ? EXIT_SUCCESS : EXIT_FAILURE);
296 /* Decode options and launch execution. */
298 static const struct option long_options[] =
300 {"crown-margin", no_argument, NULL, 'c'},
301 {"prefix", required_argument, NULL, 'p'},
302 {"split-only", no_argument, NULL, 's'},
303 {"tagged-paragraph", no_argument, NULL, 't'},
304 {"uniform-spacing", no_argument, NULL, 'u'},
305 {"width", required_argument, NULL, 'w'},
306 {GETOPT_HELP_OPTION_DECL},
307 {GETOPT_VERSION_OPTION_DECL},
312 main (register int argc, register char **argv)
316 program_name = argv[0];
317 setlocale (LC_ALL, "");
318 bindtextdomain (PACKAGE, LOCALEDIR);
319 textdomain (PACKAGE);
321 atexit (close_stdout);
323 crown = tagged = split = uniform = FALSE;
326 prefix_length = prefix_lead_space = prefix_full_length = 0;
328 if (argc > 1 && argv[1][0] == '-' && ISDIGIT (argv[1][1]))
330 const char *s = argv[1] + 1;
332 /* Old option syntax; a dash followed by one or more digits.
333 Move past the number. */
334 for (; ISDIGIT (*s); ++s)
336 int old_max = max_width;
337 max_width = max_width * 10 + *s - '0';
338 if (INT_MAX / 10 < old_max || max_width < old_max)
339 error (EXIT_FAILURE, 0, _("invalid width option: `%s'"), argv[1]);
341 /* Make the options we just parsed invisible to getopt. */
347 while ((optchar = getopt_long (argc, argv, "0123456789cstuw:p:",
377 if (xstrtol (optarg, NULL, 10, &tmp_long, "") != LONGINT_OK
378 || tmp_long <= 0 || tmp_long > INT_MAX)
379 error (EXIT_FAILURE, 0, _("invalid width: `%s'"),
381 max_width = (int) tmp_long;
389 case_GETOPT_HELP_CHAR;
391 case_GETOPT_VERSION_CHAR (PROGRAM_NAME, AUTHORS);
395 best_width = max_width * (2 * (100 - LEEWAY) + 1) / 200;
401 for (; optind < argc; optind++)
403 char *file = argv[optind];
404 if (STREQ (file, "-"))
409 in_stream = fopen (file, "r");
410 if (in_stream != NULL)
413 if (fclose (in_stream) == EOF)
414 error (EXIT_FAILURE, errno, "%s", file);
417 error (0, errno, "%s", file);
425 /* Trim space from the front and back of the string P, yielding the prefix,
426 and record the lengths of the prefix and the space trimmed. */
429 set_prefix (register char *p)
433 prefix_lead_space = 0;
440 prefix_full_length = strlen (p);
441 s = p + prefix_full_length;
442 while (s > p && s[-1] == ' ')
445 prefix_length = s - p;
448 /* read file F and send formatted output to stdout. */
455 next_char = get_prefix (f);
456 while (get_paragraph (f))
459 put_paragraph (word_limit);
463 /* Read a paragraph from input file F. A paragraph consists of a
464 maximal number of non-blank (excluding any prefix) lines subject to:
465 * In split mode, a paragraph is a single non-blank line.
466 * In crown mode, the second and subsequent lines must have the
467 same indentation, but possibly different from the indent of the
469 * Tagged mode is similar, but the first and second lines must have
470 different indentations.
471 * Otherwise, all lines of a paragraph must have the same indent.
472 If a prefix is in effect, it must be present at the same indent for
473 each line in the paragraph.
475 Return FALSE if end-of-file was encountered before the start of a
476 paragraph, else TRUE. */
479 get_paragraph (FILE *f)
483 last_line_length = 0;
486 /* Scan (and copy) blank lines, and lines not introduced by the prefix. */
488 while (c == '\n' || c == EOF
489 || next_prefix_indent < prefix_lead_space
490 || in_column < next_prefix_indent + prefix_full_length)
492 c = copy_rest (f, c);
502 /* Got a suitable first line for a paragraph. */
504 prefix_indent = next_prefix_indent;
505 first_indent = in_column;
510 /* Read rest of paragraph (unless split is specified). */
513 other_indent = first_indent;
518 other_indent = in_column;
520 { /* for each line till the end of the para */
523 while (same_para (c) && in_column == other_indent);
526 other_indent = first_indent;
530 if (same_para (c) && in_column != first_indent)
532 other_indent = in_column;
534 { /* for each line till the end of the para */
537 while (same_para (c) && in_column == other_indent);
540 /* Only one line: use the secondary indent from last time if it
541 splits, or 0 if there have been no multi-line paragraphs in the
542 input so far. But if these rules make the two indents the same,
543 pick a new secondary indent. */
545 else if (other_indent == first_indent)
546 other_indent = first_indent == 0 ? DEF_INDENT : 0;
550 other_indent = first_indent;
551 while (same_para (c) && in_column == other_indent)
554 (word_limit - 1)->period = (word_limit - 1)->final = TRUE;
559 /* Copy to the output a line that failed to match the prefix, or that
560 was blank after the prefix. In the former case, C is the character
561 that failed to match the prefix. In the latter, C is \n or EOF.
562 Return the character (\n or EOF) ending the line. */
565 copy_rest (FILE *f, register int c)
567 register const char *s;
570 if (in_column > next_prefix_indent && c != '\n' && c != EOF)
572 put_space (next_prefix_indent);
573 for (s = prefix; out_column != in_column && *s; out_column++)
574 putchar (*(unsigned char *)s++);
575 put_space (in_column - out_column);
577 while (c != '\n' && c != EOF)
585 /* Return TRUE if a line whose first non-blank character after the
586 prefix (if any) is C could belong to the current paragraph,
590 same_para (register int c)
592 return (next_prefix_indent == prefix_indent
593 && in_column >= next_prefix_indent + prefix_full_length
594 && c != '\n' && c != EOF);
597 /* Read a line from input file F, given first non-blank character C
598 after the prefix, and the following indent, and break it into words.
599 A word is a maximal non-empty string of non-white characters. A word
600 ending in [.?!]["')\]]* and followed by end-of-line or at least two
601 spaces ends a sentence, as in emacs.
603 Return the first non-blank character of the next line. */
606 get_line (FILE *f, register int c)
609 register char *end_of_parabuf;
610 register WORD *end_of_word;
612 end_of_parabuf = ¶buf[MAXCHARS];
613 end_of_word = &word[MAXWORDS - 2];
616 { /* for each word in a line */
620 word_limit->text = wptr;
623 if (wptr == end_of_parabuf)
628 while (c != EOF && !ISSPACE (c));
629 in_column += word_limit->length = wptr - word_limit->text;
630 check_punctuation (word_limit);
632 /* Scan inter-word space. */
635 c = get_space (f, c);
636 word_limit->space = in_column - start;
637 word_limit->final = (c == EOF
638 || (word_limit->period
639 && (c == '\n' || word_limit->space > 1)));
640 if (c == '\n' || c == EOF || uniform)
641 word_limit->space = word_limit->final ? 2 : 1;
642 if (word_limit == end_of_word)
649 return get_prefix (f);
652 /* Read a prefix from input file F. Return either first non-matching
653 character, or first non-blank character after the prefix. */
661 c = get_space (f, getc (f));
662 if (prefix_length == 0)
663 next_prefix_indent = prefix_lead_space < in_column ?
664 prefix_lead_space : in_column;
668 next_prefix_indent = in_column;
669 for (p = prefix; *p != '\0'; p++)
671 if (c != *(unsigned char *)p)
676 c = get_space (f, c);
681 /* Read blank characters from input file F, starting with C, and keeping
682 in_column up-to-date. Return first non-blank character. */
685 get_space (FILE *f, register int c)
694 in_column = (in_column / TABWIDTH + 1) * TABWIDTH;
702 /* Set extra fields in word W describing any attached punctuation. */
705 check_punctuation (register WORD *w)
707 const unsigned char *start, *finish;
709 start = (unsigned char *) w->text;
710 finish = start + (w->length - 1);
711 w->paren = isopen (*start);
712 w->punct = ISPUNCT (*finish);
713 while (isclose (*finish) && finish > start)
715 w->period = isperiod (*finish);
718 /* Flush part of the paragraph to make room. This function is called on
719 hitting the limit on the number of words or characters. */
722 flush_paragraph (void)
729 /* In the special case where it's all one word, just flush it. */
731 if (word_limit == word)
733 printf ("%*s", wptr - parabuf, parabuf);
739 - format what you have so far as a paragraph,
740 - find a low-cost line break near the end,
742 - make that the start of the paragraph. */
746 /* Choose a good split point. */
748 split_point = word_limit;
749 best_break = MAXCOST;
750 for (w = word->next_break; w != word_limit; w = w->next_break)
752 if (w->best_cost - w->next_break->best_cost < best_break)
755 best_break = w->best_cost - w->next_break->best_cost;
757 if (best_break <= MAXCOST - LINE_CREDIT)
758 best_break += LINE_CREDIT;
760 put_paragraph (split_point);
762 /* Copy text of words down to start of parabuf -- we use memmove because
763 the source and target may overlap. */
765 memmove (parabuf, split_point->text, (size_t) (wptr - split_point->text));
766 shift = split_point->text - parabuf;
769 /* Adjust text pointers. */
771 for (w = split_point; w <= word_limit; w++)
774 /* Copy words from split_point down to word -- we use memmove because
775 the source and target may overlap. */
777 memmove ((char *) word, (char *) split_point,
778 (word_limit - split_point + 1) * sizeof (WORD));
779 word_limit -= split_point - word;
782 /* Compute the optimal formatting for the whole paragraph by computing
783 and remembering the optimal formatting for each suffix from the empty
784 one to the whole paragraph. */
789 register WORD *start, *w;
791 register COST wcost, best;
794 word_limit->best_cost = 0;
795 saved_length = word_limit->length;
796 word_limit->length = max_width; /* sentinel */
798 for (start = word_limit - 1; start >= word; start--)
801 len = start == word ? first_indent : other_indent;
803 /* At least one word, however long, in the line. */
811 /* Consider breaking before w. */
813 wcost = line_cost (w, len) + w->best_cost;
814 if (start == word && last_line_length > 0)
815 wcost += RAGGED_COST (len - last_line_length);
819 start->next_break = w;
820 start->line_length = len;
823 /* This is a kludge to keep us from computing `len' as the
824 sum of the sentinel length and some non-zero number.
825 Since the sentinel w->length may be INT_MAX, adding
826 to that would give a negative result. */
830 len += (w - 1)->space + w->length; /* w > start >= word */
832 while (len < max_width);
833 start->best_cost = best + base_cost (start);
836 word_limit->length = saved_length;
839 /* Return the constant component of the cost of breaking before the
843 base_cost (register WORD *this)
851 if ((this - 1)->period)
853 if ((this - 1)->final)
854 cost -= SENTENCE_BONUS;
856 cost += NOBREAK_COST;
858 else if ((this - 1)->punct)
860 else if (this > word + 1 && (this - 2)->final)
861 cost += WIDOW_COST ((this - 1)->length);
866 else if (this->final)
867 cost += ORPHAN_COST (this->length);
872 /* Return the component of the cost of breaking before word NEXT that
873 depends on LEN, the length of the line beginning there. */
876 line_cost (register WORD *next, register int len)
881 if (next == word_limit)
883 n = best_width - len;
884 cost = SHORT_COST (n);
885 if (next->next_break != word_limit)
887 n = len - next->line_length;
888 cost += RAGGED_COST (n);
893 /* Output to stdout a paragraph from word up to (but not including)
894 FINISH, which must be in the next_break chain from word. */
897 put_paragraph (register WORD *finish)
901 put_line (word, first_indent);
902 for (w = word->next_break; w != finish; w = w->next_break)
903 put_line (w, other_indent);
906 /* Output to stdout the line beginning with word W, beginning in column
907 INDENT, including the prefix (if any). */
910 put_line (register WORD *w, int indent)
912 register WORD *endline;
915 put_space (prefix_indent);
916 fputs (prefix, stdout);
917 out_column += prefix_length;
918 put_space (indent - out_column);
920 endline = w->next_break - 1;
921 for (; w != endline; w++)
924 put_space (w->space);
927 last_line_length = out_column;
931 /* Output to stdout the word W. */
934 put_word (register WORD *w)
936 register const char *s;
940 for (n = w->length; n != 0; n--)
942 out_column += w->length;
945 /* Output to stdout SPACE spaces, or equivalent tabs. */
948 put_space (int space)
950 register int space_target, tab_target;
952 space_target = out_column + space;
955 tab_target = space_target / TABWIDTH * TABWIDTH;
956 if (out_column + 1 < tab_target)
957 while (out_column < tab_target)
960 out_column = (out_column / TABWIDTH + 1) * TABWIDTH;
963 while (out_column < space_target)