1 /* GNU fmt -- simple text formatter.
2 Copyright (C) 1994-1999 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
31 #include "long-options.h"
34 /* The official name of this program (e.g., no `g' prefix). */
35 #define PROGRAM_NAME "fmt"
37 /* The following parameters represent the program's idea of what is
38 "best". Adjust to taste, subject to the caveats given. */
40 /* Default longest permitted line length (max_width). */
43 /* Prefer lines to be LEEWAY % shorter than the maximum width, giving
44 room for optimization. */
47 /* The default secondary indent of tagged paragraph used for unindented
48 one-line paragraphs not preceded by any multi-line paragraphs. */
51 /* Costs and bonuses are expressed as the equivalent departure from the
52 optimal line length, multiplied by 10. e.g. assigning something a
53 cost of 50 means that it is as bad as a line 5 characters too short
54 or too long. The definition of SHORT_COST(n) should not be changed.
55 However, EQUIV(n) may need tuning. */
59 #define MAXCOST (~(((unsigned long) 1) << (8 * sizeof (COST) -1)))
61 #define SQR(n) ((n) * (n))
62 #define EQUIV(n) SQR ((COST) (n))
64 /* Cost of a filled line n chars longer or shorter than best_width. */
65 #define SHORT_COST(n) EQUIV ((n) * 10)
67 /* Cost of the difference between adjacent filled lines. */
68 #define RAGGED_COST(n) (SHORT_COST (n) / 2)
70 /* Basic cost per line. */
71 #define LINE_COST EQUIV (70)
73 /* Cost of breaking a line after the first word of a sentence, where
74 the length of the word is N. */
75 #define WIDOW_COST(n) (EQUIV (200) / ((n) + 2))
77 /* Cost of breaking a line before the last word of a sentence, where
78 the length of the word is N. */
79 #define ORPHAN_COST(n) (EQUIV (150) / ((n) + 2))
81 /* Bonus for breaking a line at the end of a sentence. */
82 #define SENTENCE_BONUS EQUIV (50)
84 /* Cost of breaking a line after a period not marking end of a sentence.
85 With the definition of sentence we are using (borrowed from emacs, see
86 get_line()) such a break would then look like a sentence break. Hence
87 we assign a very high cost -- it should be avoided unless things are
89 #define NOBREAK_COST EQUIV (600)
91 /* Bonus for breaking a line before open parenthesis. */
92 #define PAREN_BONUS EQUIV (40)
94 /* Bonus for breaking a line after other punctuation. */
95 #define PUNCT_BONUS EQUIV(40)
97 /* Credit for breaking a long paragraph one line later. */
98 #define LINE_CREDIT EQUIV(3)
100 /* Size of paragraph buffer, in words and characters. Longer paragraphs
101 are handled neatly (cf. flush_paragraph()), so there's little to gain
102 by making these larger. */
103 #define MAXWORDS 1000
104 #define MAXCHARS 5000
106 /* Extra ctype(3)-style macros. */
108 #define isopen(c) (strchr ("([`'\"", c) != NULL)
109 #define isclose(c) (strchr (")]'\"", c) != NULL)
110 #define isperiod(c) (strchr (".?!", c) != NULL)
112 /* Size of a tab stop, for expansion on input and re-introduction on
116 /* Miscellaneous definitions. */
118 typedef unsigned int bool;
124 /* Word descriptor structure. */
126 typedef struct Word WORD;
131 /* Static attributes determined during input. */
133 const char *text; /* the text of the word */
134 short length; /* length of this word */
135 short space; /* the size of the following space */
136 bool paren:1; /* starts with open paren */
137 bool period:1; /* ends in [.?!])* */
138 bool punct:1; /* ends in punctuation */
139 bool final:1; /* end of sentence */
141 /* The remaining fields are computed during the optimization. */
143 short line_length; /* length of the best line starting here */
144 COST best_cost; /* cost of best paragraph starting here */
145 WORD *next_break; /* break which achieves best_cost */
148 /* Forward declarations. */
150 static void set_prefix PARAMS ((char *p));
151 static void fmt PARAMS ((FILE *f));
152 static bool get_paragraph PARAMS ((FILE *f));
153 static int get_line PARAMS ((FILE *f, int c));
154 static int get_prefix PARAMS ((FILE *f));
155 static int get_space PARAMS ((FILE *f, int c));
156 static int copy_rest PARAMS ((FILE *f, int c));
157 static bool same_para PARAMS ((int c));
158 static void flush_paragraph PARAMS ((void));
159 static void fmt_paragraph PARAMS ((void));
160 static void check_punctuation PARAMS ((WORD *w));
161 static COST base_cost PARAMS ((WORD *this));
162 static COST line_cost PARAMS ((WORD *next, int len));
163 static void put_paragraph PARAMS ((WORD *finish));
164 static void put_line PARAMS ((WORD *w, int indent));
165 static void put_word PARAMS ((WORD *w));
166 static void put_space PARAMS ((int space));
168 /* The name this program was run with. */
169 const char *program_name;
173 /* If TRUE, first 2 lines may have different indent (default FALSE). */
176 /* If TRUE, first 2 lines _must_ have different indent (default FALSE). */
179 /* If TRUE, each line is a paragraph on its own (default FALSE). */
182 /* If TRUE, don't preserve inter-word spacing (default FALSE). */
185 /* Prefix minus leading and trailing spaces (default ""). */
186 static const char *prefix;
188 /* User-supplied maximum line width (default WIDTH). The only output
190 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\
280 -t, --tagged-paragraph indentation of first line different from second\n\
281 -u, --uniform-spacing one space between words, two after sentences\n\
282 -w, --width=NUMBER maximum line width (default of 75 columns)\n\
283 --help display this help and exit\n\
284 --version output version information and exit\n\
286 In -wNUMBER, the letter `w' may be omitted.\n"),
288 puts (_("\nReport bugs to <bug-textutils@gnu.org>."));
290 exit (status == 0 ? EXIT_SUCCESS : EXIT_FAILURE);
293 /* Decode options and launch execution. */
295 static const struct option long_options[] =
297 {"crown-margin", no_argument, NULL, 'c'},
298 {"prefix", required_argument, NULL, 'p'},
299 {"split-only", no_argument, NULL, 's'},
300 {"tagged-paragraph", no_argument, NULL, 't'},
301 {"uniform-spacing", no_argument, NULL, 'u'},
302 {"width", required_argument, NULL, 'w'},
307 main (register int argc, register char **argv)
311 program_name = argv[0];
312 setlocale (LC_ALL, "");
313 bindtextdomain (PACKAGE, LOCALEDIR);
314 textdomain (PACKAGE);
316 parse_long_options (argc, argv, "fmt", GNU_PACKAGE, VERSION,
317 "Ross Paterson", usage);
319 crown = tagged = split = uniform = FALSE;
322 prefix_length = prefix_lead_space = prefix_full_length = 0;
324 if (argc > 1 && argv[1][0] == '-' && ISDIGIT (argv[1][1]))
327 /* Old option syntax; a dash followed by one or more digits.
328 Move past the number. */
329 for (++argv[1]; ISDIGIT (*argv[1]); ++argv[1])
331 /* FIXME: use strtol to detect overflow. */
332 max_width = max_width * 10 + *argv[1] - '0';
334 /* Make the options we just parsed invisible to getopt. */
340 while ((optchar = getopt_long (argc, argv, "0123456789cstuw:p:",
370 if (xstrtol (optarg, NULL, 10, &tmp_long, "") != LONGINT_OK
371 || tmp_long <= 0 || tmp_long > INT_MAX)
372 error (EXIT_FAILURE, 0, _("invalid line number increment: `%s'"),
374 max_width = (int) tmp_long;
384 best_width = max_width * (2 * (100 - LEEWAY) + 1) / 200;
390 for (; optind < argc; optind++)
392 char *file = argv[optind];
393 if (STREQ (file, "-"))
398 in_stream = fopen (file, "r");
399 if (in_stream != NULL)
402 if (fclose (in_stream) == EOF)
403 error (EXIT_FAILURE, errno, "%s", file);
406 error (0, errno, "%s", file);
411 if (ferror (stdout) || fclose (stdout) == EOF)
412 error (EXIT_FAILURE, errno, _("write error"));
417 /* Trim space from the front and back of the string P, yielding the prefix,
418 and record the lengths of the prefix and the space trimmed. */
421 set_prefix (register char *p)
425 prefix_lead_space = 0;
432 prefix_full_length = strlen (p);
433 s = p + prefix_full_length;
434 while (s > p && s[-1] == ' ')
437 prefix_length = s - p;
440 /* read file F and send formatted output to stdout. */
447 next_char = get_prefix (f);
448 while (get_paragraph (f))
451 put_paragraph (word_limit);
455 /* Read a paragraph from input file F. A paragraph consists of a
456 maximal number of non-blank (excluding any prefix) lines subject to:
457 * In split mode, a paragraph is a single non-blank line.
458 * In crown mode, the second and subsequent lines must have the
459 same indentation, but possibly different from the indent of the
461 * Tagged mode is similar, but the first and second lines must have
462 different indentations.
463 * Otherwise, all lines of a paragraph must have the same indent.
464 If a prefix is in effect, it must be present at the same indent for
465 each line in the paragraph.
467 Return FALSE if end-of-file was encountered before the start of a
468 paragraph, else TRUE. */
471 get_paragraph (FILE *f)
475 last_line_length = 0;
478 /* Scan (and copy) blank lines, and lines not introduced by the prefix. */
480 while (c == '\n' || c == EOF
481 || next_prefix_indent < prefix_lead_space
482 || in_column < next_prefix_indent + prefix_full_length)
484 c = copy_rest (f, c);
494 /* Got a suitable first line for a paragraph. */
496 prefix_indent = next_prefix_indent;
497 first_indent = in_column;
502 /* Read rest of paragraph (unless split is specified). */
505 other_indent = first_indent;
510 other_indent = in_column;
512 { /* for each line till the end of the para */
515 while (same_para (c) && in_column == other_indent);
518 other_indent = first_indent;
522 if (same_para (c) && in_column != first_indent)
524 other_indent = in_column;
526 { /* for each line till the end of the para */
529 while (same_para (c) && in_column == other_indent);
532 /* Only one line: use the secondary indent from last time if it
533 splits, or 0 if there have been no multi-line paragraphs in the
534 input so far. But if these rules make the two indents the same,
535 pick a new secondary indent. */
537 else if (other_indent == first_indent)
538 other_indent = first_indent == 0 ? DEF_INDENT : 0;
542 other_indent = first_indent;
543 while (same_para (c) && in_column == other_indent)
546 (word_limit - 1)->period = (word_limit - 1)->final = TRUE;
551 /* Copy to the output a line that failed to match the prefix, or that
552 was blank after the prefix. In the former case, C is the character
553 that failed to match the prefix. In the latter, C is \n or EOF.
554 Return the character (\n or EOF) ending the line. */
557 copy_rest (FILE *f, register int c)
559 register const char *s;
562 if (in_column > next_prefix_indent && c != '\n' && c != EOF)
564 put_space (next_prefix_indent);
565 for (s = prefix; out_column != in_column && *s; out_column++)
567 put_space (in_column - out_column);
569 while (c != '\n' && c != EOF)
577 /* Return TRUE if a line whose first non-blank character after the
578 prefix (if any) is C could belong to the current paragraph,
582 same_para (register int c)
584 return (next_prefix_indent == prefix_indent
585 && in_column >= next_prefix_indent + prefix_full_length
586 && c != '\n' && c != EOF);
589 /* Read a line from input file F, given first non-blank character C
590 after the prefix, and the following indent, and break it into words.
591 A word is a maximal non-empty string of non-white characters. A word
592 ending in [.?!]["')\]]* and followed by end-of-line or at least two
593 spaces ends a sentence, as in emacs.
595 Return the first non-blank character of the next line. */
598 get_line (FILE *f, register int c)
601 register char *end_of_parabuf;
602 register WORD *end_of_word;
604 end_of_parabuf = ¶buf[MAXCHARS];
605 end_of_word = &word[MAXWORDS - 2];
608 { /* for each word in a line */
612 word_limit->text = wptr;
615 if (wptr == end_of_parabuf)
620 while (c != EOF && !ISSPACE (c));
621 in_column += word_limit->length = wptr - word_limit->text;
622 check_punctuation (word_limit);
624 /* Scan inter-word space. */
627 c = get_space (f, c);
628 word_limit->space = in_column - start;
629 word_limit->final = (c == EOF
630 || (word_limit->period
631 && (c == '\n' || word_limit->space > 1)));
632 if (c == '\n' || c == EOF || uniform)
633 word_limit->space = word_limit->final ? 2 : 1;
634 if (word_limit == end_of_word)
641 return get_prefix (f);
644 /* Read a prefix from input file F. Return either first non-matching
645 character, or first non-blank character after the prefix. */
651 register const char *p;
654 c = get_space (f, getc (f));
655 if (prefix_length == 0)
656 next_prefix_indent = prefix_lead_space < in_column ?
657 prefix_lead_space : in_column;
660 next_prefix_indent = in_column;
661 for (p = prefix; *p != '\0'; p++)
668 c = get_space (f, c);
673 /* Read blank characters from input file F, starting with C, and keeping
674 in_column up-to-date. Return first non-blank character. */
677 get_space (FILE *f, register int c)
686 in_column = (in_column / TABWIDTH + 1) * TABWIDTH;
694 /* Set extra fields in word W describing any attached punctuation. */
697 check_punctuation (register WORD *w)
699 const unsigned char *start, *finish;
701 start = (unsigned char *) w->text;
702 finish = start + (w->length - 1);
703 w->paren = isopen (*start);
704 w->punct = ISPUNCT (*finish);
705 while (isclose (*finish) && finish > start)
707 w->period = isperiod (*finish);
710 /* Flush part of the paragraph to make room. This function is called on
711 hitting the limit on the number of words or characters. */
714 flush_paragraph (void)
721 /* In the special case where it's all one word, just flush it. */
723 if (word_limit == word)
725 printf ("%*s", wptr - parabuf, parabuf);
731 - format what you have so far as a paragraph,
732 - find a low-cost line break near the end,
734 - make that the start of the paragraph. */
738 /* Choose a good split point. */
740 split_point = word_limit;
741 best_break = MAXCOST;
742 for (w = word->next_break; w != word_limit; w = w->next_break)
744 if (w->best_cost - w->next_break->best_cost < best_break)
747 best_break = w->best_cost - w->next_break->best_cost;
749 if (best_break <= MAXCOST - LINE_CREDIT)
750 best_break += LINE_CREDIT;
752 put_paragraph (split_point);
754 /* Copy text of words down to start of parabuf -- we use memmove because
755 the source and target may overlap. */
757 memmove (parabuf, split_point->text, (size_t) (wptr - split_point->text));
758 shift = split_point->text - parabuf;
761 /* Adjust text pointers. */
763 for (w = split_point; w <= word_limit; w++)
766 /* Copy words from split_point down to word -- we use memmove because
767 the source and target may overlap. */
769 memmove ((char *) word, (char *) split_point,
770 (word_limit - split_point + 1) * sizeof (WORD));
771 word_limit -= split_point - word;
774 /* Compute the optimal formatting for the whole paragraph by computing
775 and remembering the optimal formatting for each suffix from the empty
776 one to the whole paragraph. */
781 register WORD *start, *w;
783 register COST wcost, best;
786 word_limit->best_cost = 0;
787 saved_length = word_limit->length;
788 word_limit->length = max_width; /* sentinel */
790 for (start = word_limit - 1; start >= word; start--)
793 len = start == word ? first_indent : other_indent;
795 /* At least one word, however long, in the line. */
803 /* Consider breaking before w. */
805 wcost = line_cost (w, len) + w->best_cost;
806 if (start == word && last_line_length > 0)
807 wcost += RAGGED_COST (len - last_line_length);
811 start->next_break = w;
812 start->line_length = len;
814 len += (w - 1)->space + w->length; /* w > start >= word */
816 while (len < max_width);
817 start->best_cost = best + base_cost (start);
820 word_limit->length = saved_length;
823 /* Return the constant component of the cost of breaking before the
827 base_cost (register WORD *this)
835 if ((this - 1)->period)
837 if ((this - 1)->final)
838 cost -= SENTENCE_BONUS;
840 cost += NOBREAK_COST;
842 else if ((this - 1)->punct)
844 else if (this > word + 1 && (this - 2)->final)
845 cost += WIDOW_COST ((this - 1)->length);
850 else if (this->final)
851 cost += ORPHAN_COST (this->length);
856 /* Return the component of the cost of breaking before word NEXT that
857 depends on LEN, the length of the line beginning there. */
860 line_cost (register WORD *next, register int len)
865 if (next == word_limit)
867 n = best_width - len;
868 cost = SHORT_COST (n);
869 if (next->next_break != word_limit)
871 n = len - next->line_length;
872 cost += RAGGED_COST (n);
877 /* Output to stdout a paragraph from word up to (but not including)
878 FINISH, which must be in the next_break chain from word. */
881 put_paragraph (register WORD *finish)
885 put_line (word, first_indent);
886 for (w = word->next_break; w != finish; w = w->next_break)
887 put_line (w, other_indent);
890 /* Output to stdout the line beginning with word W, beginning in column
891 INDENT, including the prefix (if any). */
894 put_line (register WORD *w, int indent)
896 register WORD *endline;
899 put_space (prefix_indent);
900 fputs (prefix, stdout);
901 out_column += prefix_length;
902 put_space (indent - out_column);
904 endline = w->next_break - 1;
905 for (; w != endline; w++)
908 put_space (w->space);
911 last_line_length = out_column;
915 /* Output to stdout the word W. */
918 put_word (register WORD *w)
920 register const char *s;
924 for (n = w->length; n != 0; n--)
926 out_column += w->length;
929 /* Output to stdout SPACE spaces, or equivalent tabs. */
932 put_space (int space)
934 register int space_target, tab_target;
936 space_target = out_column + space;
939 tab_target = space_target / TABWIDTH * TABWIDTH;
940 if (out_column + 1 < tab_target)
941 while (out_column < tab_target)
944 out_column = (out_column / TABWIDTH + 1) * TABWIDTH;
947 while (out_column < space_target)