build: improve libcrypto library detection
[coreutils.git] / src / fmt.c
blob73ef64ee4022762c764c74b94cb4eb33b8efc066
1 /* GNU fmt -- simple text formatter.
2 Copyright (C) 1994-2024 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 3 of the License, or
7 (at your option) any later version.
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, see <https://www.gnu.org/licenses/>. */
17 /* Written by Ross Paterson <rap@doc.ic.ac.uk>. */
19 #include <config.h>
20 #include <ctype.h>
21 #include <stdio.h>
22 #include <sys/types.h>
23 #include <getopt.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
29 #include "c-ctype.h"
30 #include "system.h"
31 #include "fadvise.h"
32 #include "xdectoint.h"
34 /* The official name of this program (e.g., no 'g' prefix). */
35 #define PROGRAM_NAME "fmt"
37 #define AUTHORS proper_name ("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). */
43 #define WIDTH 75
45 /* Prefer lines to be LEEWAY % shorter than the maximum width, giving
46 room for optimization. */
47 #define LEEWAY 7
49 /* The default secondary indent of tagged paragraph used for unindented
50 one-line paragraphs not preceded by any multi-line paragraphs. */
51 #define DEF_INDENT 3
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. */
59 /* FIXME: "fmt" misbehaves given large inputs or options. One
60 possible workaround for part of the problem is to change COST to be
61 a floating-point type. There are other problems besides COST,
62 though; see MAXWORDS below. */
64 typedef long int COST;
66 #define MAXCOST TYPE_MAXIMUM (COST)
68 #define SQR(n) ((n) * (n))
69 #define EQUIV(n) SQR ((COST) (n))
71 /* Cost of a filled line n chars longer or shorter than goal_width. */
72 #define SHORT_COST(n) EQUIV ((n) * 10)
74 /* Cost of the difference between adjacent filled lines. */
75 #define RAGGED_COST(n) (SHORT_COST (n) / 2)
77 /* Basic cost per line. */
78 #define LINE_COST EQUIV (70)
80 /* Cost of breaking a line after the first word of a sentence, where
81 the length of the word is N. */
82 #define WIDOW_COST(n) (EQUIV (200) / ((n) + 2))
84 /* Cost of breaking a line before the last word of a sentence, where
85 the length of the word is N. */
86 #define ORPHAN_COST(n) (EQUIV (150) / ((n) + 2))
88 /* Bonus for breaking a line at the end of a sentence. */
89 #define SENTENCE_BONUS EQUIV (50)
91 /* Cost of breaking a line after a period not marking end of a sentence.
92 With the definition of sentence we are using (borrowed from emacs, see
93 get_line()) such a break would then look like a sentence break. Hence
94 we assign a very high cost -- it should be avoided unless things are
95 really bad. */
96 #define NOBREAK_COST EQUIV (600)
98 /* Bonus for breaking a line before open parenthesis. */
99 #define PAREN_BONUS EQUIV (40)
101 /* Bonus for breaking a line after other punctuation. */
102 #define PUNCT_BONUS EQUIV(40)
104 /* Credit for breaking a long paragraph one line later. */
105 #define LINE_CREDIT EQUIV(3)
107 /* Size of paragraph buffer, in words and characters. Longer paragraphs
108 are handled neatly (cf. flush_paragraph()), so long as these values
109 are considerably greater than required by the width. These values
110 cannot be extended indefinitely: doing so would run into size limits
111 and/or cause more overflows in cost calculations. FIXME: Remove these
112 arbitrary limits. */
114 #define MAXWORDS 1000
115 #define MAXCHARS 5000
117 /* Extra ctype(3)-style macros. */
119 #define isopen(c) (strchr ("(['`\"", c) != nullptr)
120 #define isclose(c) (strchr (")]'\"", c) != nullptr)
121 #define isperiod(c) (strchr (".?!", c) != nullptr)
123 /* Size of a tab stop, for expansion on input and re-introduction on
124 output. */
125 #define TABWIDTH 8
127 /* Word descriptor structure. */
129 typedef struct Word WORD;
131 struct Word
134 /* Static attributes determined during input. */
136 char const *text; /* the text of the word */
137 int length; /* length of this word */
138 int space; /* the size of the following space */
139 unsigned int paren:1; /* starts with open paren */
140 unsigned int period:1; /* ends in [.?!])* */
141 unsigned int punct:1; /* ends in punctuation */
142 unsigned int final:1; /* end of sentence */
144 /* The remaining fields are computed during the optimization. */
146 int line_length; /* length of the best line starting here */
147 COST best_cost; /* cost of best paragraph starting here */
148 WORD *next_break; /* break which achieves best_cost */
151 /* Forward declarations. */
153 static void set_prefix (char *p);
154 static bool fmt (FILE *f, char const *);
155 static bool get_paragraph (FILE *f);
156 static int get_line (FILE *f, int c);
157 static int get_prefix (FILE *f);
158 static int get_space (FILE *f, int c);
159 static int copy_rest (FILE *f, int c);
160 static bool same_para (int c);
161 static void flush_paragraph (void);
162 static void fmt_paragraph (void);
163 static void check_punctuation (WORD *w);
164 static COST base_cost (WORD *this);
165 static COST line_cost (WORD *next, int len);
166 static void put_paragraph (WORD *finish);
167 static void put_line (WORD *w, int indent);
168 static void put_word (WORD *w);
169 static void put_space (int space);
171 /* Option values. */
173 /* If true, first 2 lines may have different indent (default false). */
174 static bool crown;
176 /* If true, first 2 lines _must_ have different indent (default false). */
177 static bool tagged;
179 /* If true, each line is a paragraph on its own (default false). */
180 static bool split;
182 /* If true, don't preserve inter-word spacing (default false). */
183 static bool uniform;
185 /* Prefix minus leading and trailing spaces (default ""). */
186 static char const *prefix;
188 /* User-supplied maximum line width (default WIDTH). The only output
189 lines longer than this will each comprise a single word. */
190 static int max_width;
192 /* Values derived from the option values. */
194 /* The length of prefix minus leading space. */
195 static int prefix_full_length;
197 /* The length of the leading space trimmed from the prefix. */
198 static int prefix_lead_space;
200 /* The length of prefix minus leading and trailing space. */
201 static int prefix_length;
203 /* The preferred width of text lines, set to LEEWAY % less than max_width. */
204 static int goal_width;
206 /* Dynamic variables. */
208 /* Start column of the character most recently read from the input file. */
209 static int in_column;
211 /* Start column of the next character to be written to stdout. */
212 static int out_column;
214 /* Space for the paragraph text -- longer paragraphs are handled neatly
215 (cf. flush_paragraph()). */
216 static char parabuf[MAXCHARS];
218 /* A pointer into parabuf, indicating the first unused character position. */
219 static char *wptr;
221 /* The words of a paragraph -- longer paragraphs are handled neatly
222 (cf. flush_paragraph()). */
223 static WORD word[MAXWORDS];
225 /* A pointer into the above word array, indicating the first position
226 after the last complete word. Sometimes it will point at an incomplete
227 word. */
228 static WORD *word_limit;
230 /* If true, current input file contains tab characters, and so tabs can be
231 used for white space on output. */
232 static bool tabs;
234 /* Space before trimmed prefix on each line of the current paragraph. */
235 static int prefix_indent;
237 /* Indentation of the first line of the current paragraph. */
238 static int first_indent;
240 /* Indentation of other lines of the current paragraph */
241 static int other_indent;
243 /* To detect the end of a paragraph, we need to look ahead to the first
244 non-blank character after the prefix on the next line, or the first
245 character on the following line that failed to match the prefix.
246 We can reconstruct the lookahead from that character (next_char), its
247 position on the line (in_column) and the amount of space before the
248 prefix (next_prefix_indent). See get_paragraph() and copy_rest(). */
250 /* The last character read from the input file. */
251 static int next_char;
253 /* The space before the trimmed prefix (or part of it) on the next line
254 after the current paragraph. */
255 static int next_prefix_indent;
257 /* If nonzero, the length of the last line output in the current
258 paragraph, used to charge for raggedness at the split point for long
259 paragraphs chosen by fmt_paragraph(). */
260 static int last_line_length;
262 void
263 usage (int status)
265 if (status != EXIT_SUCCESS)
266 emit_try_help ();
267 else
269 printf (_("Usage: %s [-WIDTH] [OPTION]... [FILE]...\n"), program_name);
270 fputs (_("\
271 Reformat each paragraph in the FILE(s), writing to standard output.\n\
272 The option -WIDTH is an abbreviated form of --width=DIGITS.\n\
273 "), stdout);
275 emit_stdin_note ();
276 emit_mandatory_arg_note ();
278 fputs (_("\
279 -c, --crown-margin preserve indentation of first two lines\n\
280 -p, --prefix=STRING reformat only lines beginning with STRING,\n\
281 reattaching the prefix to reformatted lines\n\
282 -s, --split-only split long lines, but do not refill\n\
284 stdout);
285 /* Tell xgettext that the "% o" below is not a printf-style
286 format string: xgettext:no-c-format */
287 fputs (_("\
288 -t, --tagged-paragraph indentation of first line different from second\n\
289 -u, --uniform-spacing one space between words, two after sentences\n\
290 -w, --width=WIDTH maximum line width (default of 75 columns)\n\
291 -g, --goal=WIDTH goal width (default of 93% of width)\n\
292 "), stdout);
293 fputs (HELP_OPTION_DESCRIPTION, stdout);
294 fputs (VERSION_OPTION_DESCRIPTION, stdout);
295 emit_ancillary_info (PROGRAM_NAME);
297 exit (status);
300 /* Decode options and launch execution. */
302 static struct option const long_options[] =
304 {"crown-margin", no_argument, nullptr, 'c'},
305 {"prefix", required_argument, nullptr, 'p'},
306 {"split-only", no_argument, nullptr, 's'},
307 {"tagged-paragraph", no_argument, nullptr, 't'},
308 {"uniform-spacing", no_argument, nullptr, 'u'},
309 {"width", required_argument, nullptr, 'w'},
310 {"goal", required_argument, nullptr, 'g'},
311 {GETOPT_HELP_OPTION_DECL},
312 {GETOPT_VERSION_OPTION_DECL},
313 {nullptr, 0, nullptr, 0},
317 main (int argc, char **argv)
319 int optchar;
320 bool ok = true;
321 char const *max_width_option = nullptr;
322 char const *goal_width_option = nullptr;
324 initialize_main (&argc, &argv);
325 set_program_name (argv[0]);
326 setlocale (LC_ALL, "");
327 bindtextdomain (PACKAGE, LOCALEDIR);
328 textdomain (PACKAGE);
330 atexit (close_stdout);
332 crown = tagged = split = uniform = false;
333 max_width = WIDTH;
334 prefix = "";
335 prefix_length = prefix_lead_space = prefix_full_length = 0;
337 if (argc > 1 && argv[1][0] == '-' && ISDIGIT (argv[1][1]))
339 /* Old option syntax; a dash followed by one or more digits. */
340 max_width_option = argv[1] + 1;
342 /* Make the option we just parsed invisible to getopt. */
343 argv[1] = argv[0];
344 argv++;
345 argc--;
348 while ((optchar = getopt_long (argc, argv, "0123456789cstuw:p:g:",
349 long_options, nullptr))
350 != -1)
351 switch (optchar)
353 default:
354 if (ISDIGIT (optchar))
355 error (0, 0, _("invalid option -- %c; -WIDTH is recognized\
356 only when it is the first\noption; use -w N instead"),
357 optchar);
358 usage (EXIT_FAILURE);
360 case 'c':
361 crown = true;
362 break;
364 case 's':
365 split = true;
366 break;
368 case 't':
369 tagged = true;
370 break;
372 case 'u':
373 uniform = true;
374 break;
376 case 'w':
377 max_width_option = optarg;
378 break;
380 case 'g':
381 goal_width_option = optarg;
382 break;
384 case 'p':
385 set_prefix (optarg);
386 break;
388 case_GETOPT_HELP_CHAR;
390 case_GETOPT_VERSION_CHAR (PROGRAM_NAME, AUTHORS);
394 if (max_width_option)
396 /* Limit max_width to MAXCHARS / 2; otherwise, the resulting
397 output can be quite ugly. */
398 max_width = xdectoumax (max_width_option, 0, MAXCHARS / 2, "",
399 _("invalid width"), 0);
402 if (goal_width_option)
404 /* Limit goal_width to max_width. */
405 goal_width = xdectoumax (goal_width_option, 0, max_width, "",
406 _("invalid width"), 0);
407 if (max_width_option == nullptr)
408 max_width = goal_width + 10;
410 else
412 goal_width = max_width * (2 * (100 - LEEWAY) + 1) / 200;
415 bool have_read_stdin = false;
417 if (optind == argc)
419 have_read_stdin = true;
420 ok = fmt (stdin, "-");
422 else
424 for (; optind < argc; optind++)
426 char *file = argv[optind];
427 if (STREQ (file, "-"))
429 ok &= fmt (stdin, file);
430 have_read_stdin = true;
432 else
434 FILE *in_stream;
435 in_stream = fopen (file, "r");
436 if (in_stream != nullptr)
437 ok &= fmt (in_stream, file);
438 else
440 error (0, errno, _("cannot open %s for reading"),
441 quoteaf (file));
442 ok = false;
448 if (have_read_stdin && fclose (stdin) != 0)
449 error (EXIT_FAILURE, errno, "%s", _("closing standard input"));
451 return ok ? EXIT_SUCCESS : EXIT_FAILURE;
454 /* Trim space from the front and back of the string P, yielding the prefix,
455 and record the lengths of the prefix and the space trimmed. */
457 static void
458 set_prefix (char *p)
460 char *s;
462 prefix_lead_space = 0;
463 while (*p == ' ')
465 prefix_lead_space++;
466 p++;
468 prefix = p;
469 prefix_full_length = strlen (p);
470 s = p + prefix_full_length;
471 while (s > p && s[-1] == ' ')
472 s--;
473 *s = '\0';
474 prefix_length = s - p;
477 /* Read F and send formatted output to stdout.
478 Close F when done, unless F is stdin. Diagnose input errors, using FILE.
479 If !F, assume F resulted from an fopen failure and diagnose that.
480 Return true if successful. */
482 static bool
483 fmt (FILE *f, char const *file)
485 fadvise (f, FADVISE_SEQUENTIAL);
486 tabs = false;
487 other_indent = 0;
488 next_char = get_prefix (f);
489 while (get_paragraph (f))
491 fmt_paragraph ();
492 put_paragraph (word_limit);
495 int err = ferror (f) ? 0 : -1;
496 if (f == stdin)
497 clearerr (f);
498 else if (fclose (f) != 0 && err < 0)
499 err = errno;
500 if (0 <= err)
501 error (0, err, err ? "%s" : _("read error"), quotef (file));
502 return err < 0;
505 /* Set the global variable 'other_indent' according to SAME_PARAGRAPH
506 and other global variables. */
508 static void
509 set_other_indent (bool same_paragraph)
511 if (split)
512 other_indent = first_indent;
513 else if (crown)
515 other_indent = (same_paragraph ? in_column : first_indent);
517 else if (tagged)
519 if (same_paragraph && in_column != first_indent)
521 other_indent = in_column;
524 /* Only one line: use the secondary indent from last time if it
525 splits, or 0 if there have been no multi-line paragraphs in the
526 input so far. But if these rules make the two indents the same,
527 pick a new secondary indent. */
529 else if (other_indent == first_indent)
530 other_indent = first_indent == 0 ? DEF_INDENT : 0;
532 else
534 other_indent = first_indent;
538 /* Read a paragraph from input file F. A paragraph consists of a
539 maximal number of non-blank (excluding any prefix) lines subject to:
540 * In split mode, a paragraph is a single non-blank line.
541 * In crown mode, the second and subsequent lines must have the
542 same indentation, but possibly different from the indent of the
543 first line.
544 * Tagged mode is similar, but the first and second lines must have
545 different indentations.
546 * Otherwise, all lines of a paragraph must have the same indent.
547 If a prefix is in effect, it must be present at the same indent for
548 each line in the paragraph.
550 Return false if end-of-file was encountered before the start of a
551 paragraph, else true. */
553 static bool
554 get_paragraph (FILE *f)
556 int c;
558 last_line_length = 0;
559 c = next_char;
561 /* Scan (and copy) blank lines, and lines not introduced by the prefix. */
563 while (c == '\n' || c == EOF
564 || next_prefix_indent < prefix_lead_space
565 || in_column < next_prefix_indent + prefix_full_length)
567 c = copy_rest (f, c);
568 if (c == EOF)
570 next_char = EOF;
571 return false;
573 putchar ('\n');
574 c = get_prefix (f);
577 /* Got a suitable first line for a paragraph. */
579 prefix_indent = next_prefix_indent;
580 first_indent = in_column;
581 wptr = parabuf;
582 word_limit = word;
583 c = get_line (f, c);
584 set_other_indent (same_para (c));
586 /* Read rest of paragraph (unless split is specified). */
588 if (split)
590 /* empty */
592 else if (crown)
594 if (same_para (c))
597 { /* for each line till the end of the para */
598 c = get_line (f, c);
600 while (same_para (c) && in_column == other_indent);
603 else if (tagged)
605 if (same_para (c) && in_column != first_indent)
608 { /* for each line till the end of the para */
609 c = get_line (f, c);
611 while (same_para (c) && in_column == other_indent);
614 else
616 while (same_para (c) && in_column == other_indent)
617 c = get_line (f, c);
620 (word_limit - 1)->period = (word_limit - 1)->final = true;
621 next_char = c;
622 return true;
625 /* Copy to the output a line that failed to match the prefix, or that
626 was blank after the prefix. In the former case, C is the character
627 that failed to match the prefix. In the latter, C is \n or EOF.
628 Return the character (\n or EOF) ending the line. */
630 static int
631 copy_rest (FILE *f, int c)
633 char const *s;
635 out_column = 0;
636 if (in_column > next_prefix_indent || (c != '\n' && c != EOF))
638 put_space (next_prefix_indent);
639 for (s = prefix; out_column != in_column && *s; out_column++)
640 putchar (*s++);
641 if (c != EOF && c != '\n')
642 put_space (in_column - out_column);
643 if (c == EOF && in_column >= next_prefix_indent + prefix_length)
644 putchar ('\n');
646 while (c != '\n' && c != EOF)
648 putchar (c);
649 c = getc (f);
651 return c;
654 /* Return true if a line whose first non-blank character after the
655 prefix (if any) is C could belong to the current paragraph,
656 otherwise false. */
658 static bool
659 same_para (int c)
661 return (next_prefix_indent == prefix_indent
662 && in_column >= next_prefix_indent + prefix_full_length
663 && c != '\n' && c != EOF);
666 /* Read a line from input file F, given first non-blank character C
667 after the prefix, and the following indent, and break it into words.
668 A word is a maximal non-empty string of non-white characters. A word
669 ending in [.?!][])"']* and followed by end-of-line or at least two
670 spaces ends a sentence, as in emacs.
672 Return the first non-blank character of the next line. */
674 static int
675 get_line (FILE *f, int c)
677 int start;
678 char *end_of_parabuf;
679 WORD *end_of_word;
681 end_of_parabuf = &parabuf[MAXCHARS];
682 end_of_word = &word[MAXWORDS - 2];
685 { /* for each word in a line */
687 /* Scan word. */
689 word_limit->text = wptr;
692 if (wptr == end_of_parabuf)
694 set_other_indent (true);
695 flush_paragraph ();
697 *wptr++ = c;
698 c = getc (f);
700 while (c != EOF && !c_isspace (c));
701 in_column += word_limit->length = wptr - word_limit->text;
702 check_punctuation (word_limit);
704 /* Scan inter-word space. */
706 start = in_column;
707 c = get_space (f, c);
708 word_limit->space = in_column - start;
709 word_limit->final = (c == EOF
710 || (word_limit->period
711 && (c == '\n' || word_limit->space > 1)));
712 if (c == '\n' || c == EOF || uniform)
713 word_limit->space = word_limit->final ? 2 : 1;
714 if (word_limit == end_of_word)
716 set_other_indent (true);
717 flush_paragraph ();
719 word_limit++;
721 while (c != '\n' && c != EOF);
722 return get_prefix (f);
725 /* Read a prefix from input file F. Return either first non-matching
726 character, or first non-blank character after the prefix. */
728 static int
729 get_prefix (FILE *f)
731 int c;
733 in_column = 0;
734 c = get_space (f, getc (f));
735 if (prefix_length == 0)
736 next_prefix_indent = prefix_lead_space < in_column ?
737 prefix_lead_space : in_column;
738 else
740 char const *p;
741 next_prefix_indent = in_column;
742 for (p = prefix; *p != '\0'; p++)
744 unsigned char pc = *p;
745 if (c != pc)
746 return c;
747 in_column++;
748 c = getc (f);
750 c = get_space (f, c);
752 return c;
755 /* Read blank characters from input file F, starting with C, and keeping
756 in_column up-to-date. Return first non-blank character. */
758 static int
759 get_space (FILE *f, int c)
761 while (true)
763 if (c == ' ')
764 in_column++;
765 else if (c == '\t')
767 tabs = true;
768 in_column = (in_column / TABWIDTH + 1) * TABWIDTH;
770 else
771 return c;
772 c = getc (f);
776 /* Set extra fields in word W describing any attached punctuation. */
778 static void
779 check_punctuation (WORD *w)
781 char const *start = w->text;
782 char const *finish = start + (w->length - 1);
783 unsigned char fin = *finish;
785 w->paren = isopen (*start);
786 w->punct = !! ispunct (fin);
787 while (start < finish && isclose (*finish))
788 finish--;
789 w->period = isperiod (*finish);
792 /* Flush part of the paragraph to make room. This function is called on
793 hitting the limit on the number of words or characters. */
795 static void
796 flush_paragraph (void)
798 WORD *split_point;
799 WORD *w;
800 int shift;
801 COST best_break;
803 /* In the special case where it's all one word, just flush it. */
805 if (word_limit == word)
807 fwrite (parabuf, sizeof *parabuf, wptr - parabuf, stdout);
808 wptr = parabuf;
809 return;
812 /* Otherwise:
813 - format what you have so far as a paragraph,
814 - find a low-cost line break near the end,
815 - output to there,
816 - make that the start of the paragraph. */
818 fmt_paragraph ();
820 /* Choose a good split point. */
822 split_point = word_limit;
823 best_break = MAXCOST;
824 for (w = word->next_break; w != word_limit; w = w->next_break)
826 if (w->best_cost - w->next_break->best_cost < best_break)
828 split_point = w;
829 best_break = w->best_cost - w->next_break->best_cost;
831 if (best_break <= MAXCOST - LINE_CREDIT)
832 best_break += LINE_CREDIT;
834 put_paragraph (split_point);
836 /* Copy text of words down to start of parabuf -- we use memmove because
837 the source and target may overlap. */
839 memmove (parabuf, split_point->text, wptr - split_point->text);
840 shift = split_point->text - parabuf;
841 wptr -= shift;
843 /* Adjust text pointers. */
845 for (w = split_point; w <= word_limit; w++)
846 w->text -= shift;
848 /* Copy words from split_point down to word -- we use memmove because
849 the source and target may overlap. */
851 memmove (word, split_point, (word_limit - split_point + 1) * sizeof *word);
852 word_limit -= split_point - word;
855 /* Compute the optimal formatting for the whole paragraph by computing
856 and remembering the optimal formatting for each suffix from the empty
857 one to the whole paragraph. */
859 static void
860 fmt_paragraph (void)
862 WORD *start, *w;
863 int len;
864 COST wcost, best;
865 int saved_length;
867 word_limit->best_cost = 0;
868 saved_length = word_limit->length;
869 word_limit->length = max_width; /* sentinel */
871 for (start = word_limit - 1; start >= word; start--)
873 best = MAXCOST;
874 len = start == word ? first_indent : other_indent;
876 /* At least one word, however long, in the line. */
878 w = start;
879 len += w->length;
882 w++;
884 /* Consider breaking before w. */
886 wcost = line_cost (w, len) + w->best_cost;
887 if (start == word && last_line_length > 0)
888 wcost += RAGGED_COST (len - last_line_length);
889 if (wcost < best)
891 best = wcost;
892 start->next_break = w;
893 start->line_length = len;
896 /* This is a kludge to keep us from computing 'len' as the
897 sum of the sentinel length and some non-zero number.
898 Since the sentinel w->length may be INT_MAX, adding
899 to that would give a negative result. */
900 if (w == word_limit)
901 break;
903 len += (w - 1)->space + w->length; /* w > start >= word */
905 while (len < max_width);
906 start->best_cost = best + base_cost (start);
909 word_limit->length = saved_length;
912 /* Work around <https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109628>. */
913 #if __GNUC__ == 13
914 # pragma GCC diagnostic ignored "-Wanalyzer-use-of-uninitialized-value"
915 #endif
917 /* Return the constant component of the cost of breaking before the
918 word THIS. */
920 static COST
921 base_cost (WORD *this)
923 COST cost;
925 cost = LINE_COST;
927 if (this > word)
929 if ((this - 1)->period)
931 if ((this - 1)->final)
932 cost -= SENTENCE_BONUS;
933 else
934 cost += NOBREAK_COST;
936 else if ((this - 1)->punct)
937 cost -= PUNCT_BONUS;
938 else if (this > word + 1 && (this - 2)->final)
939 cost += WIDOW_COST ((this - 1)->length);
942 if (this->paren)
943 cost -= PAREN_BONUS;
944 else if (this->final)
945 cost += ORPHAN_COST (this->length);
947 return cost;
950 /* Return the component of the cost of breaking before word NEXT that
951 depends on LEN, the length of the line beginning there. */
953 static COST
954 line_cost (WORD *next, int len)
956 int n;
957 COST cost;
959 if (next == word_limit)
960 return 0;
961 n = goal_width - len;
962 cost = SHORT_COST (n);
963 if (next->next_break != word_limit)
965 n = len - next->line_length;
966 cost += RAGGED_COST (n);
968 return cost;
971 /* Output to stdout a paragraph from word up to (but not including)
972 FINISH, which must be in the next_break chain from word. */
974 static void
975 put_paragraph (WORD *finish)
977 WORD *w;
979 put_line (word, first_indent);
980 for (w = word->next_break; w != finish; w = w->next_break)
981 put_line (w, other_indent);
984 /* Output to stdout the line beginning with word W, beginning in column
985 INDENT, including the prefix (if any). */
987 static void
988 put_line (WORD *w, int indent)
990 WORD *endline;
992 out_column = 0;
993 put_space (prefix_indent);
994 fputs (prefix, stdout);
995 out_column += prefix_length;
996 put_space (indent - out_column);
998 endline = w->next_break - 1;
999 for (; w != endline; w++)
1001 put_word (w);
1002 put_space (w->space);
1004 put_word (w);
1005 last_line_length = out_column;
1006 putchar ('\n');
1009 /* Output to stdout the word W. */
1011 static void
1012 put_word (WORD *w)
1014 char const *s;
1015 int n;
1017 s = w->text;
1018 for (n = w->length; n != 0; n--)
1019 putchar (*s++);
1020 out_column += w->length;
1023 /* Output to stdout SPACE spaces, or equivalent tabs. */
1025 static void
1026 put_space (int space)
1028 int space_target, tab_target;
1030 space_target = out_column + space;
1031 if (tabs)
1033 tab_target = space_target / TABWIDTH * TABWIDTH;
1034 if (out_column + 1 < tab_target)
1035 while (out_column < tab_target)
1037 putchar ('\t');
1038 out_column = (out_column / TABWIDTH + 1) * TABWIDTH;
1041 while (out_column < space_target)
1043 putchar (' ');
1044 out_column++;