1 /* $OpenBSD: fmt.c,v 1.16 2000/06/25 15:35:42 pjanzen Exp $ */ 2 3 /* Sensible version of fmt 4 * 5 * Syntax: fmt [ options ] [ goal [ max ] ] [ filename ... ] 6 * 7 * Since the documentation for the original fmt is so poor, here 8 * is an accurate description of what this one does. It's usually 9 * the same. The *mechanism* used may differ from that suggested 10 * here. Note that we are *not* entirely compatible with fmt, 11 * because fmt gets so many things wrong. 12 * 13 * 1. Tabs are expanded, assuming 8-space tab stops. 14 * If the `-t <n>' option is given, we assume <n>-space 15 * tab stops instead. 16 * Trailing blanks are removed from all lines. 17 * x\b == nothing, for any x other than \b. 18 * Other control characters are simply stripped. This 19 * includes \r. 20 * 2. Each line is split into leading whitespace and 21 * everything else. Maximal consecutive sequences of 22 * lines with the same leading whitespace are considered 23 * to form paragraphs, except that a blank line is always 24 * a paragraph to itself. 25 * If the `-p' option is given then the first line of a 26 * paragraph is permitted to have indentation different 27 * from that of the other lines. 28 * If the `-m' option is given then a line that looks 29 * like a mail message header, if it is not immediately 30 * preceded by a non-blank non-message-header line, is 31 * taken to start a new paragraph, which also contains 32 * any subsequent lines with non-empty leading whitespace. 33 * Unless the `-n' option is given, lines beginning with 34 * a . (dot) are not formatted. 35 * 3. The "everything else" is split into words; a word 36 * includes its trailing whitespace, and a word at the 37 * end of a line is deemed to be followed by a single 38 * space, or two spaces if it ends with a sentence-end 39 * character. (See the `-d' option for how to change that.) 40 * If the `-s' option has been given, then a word's trailing 41 * whitespace is replaced by what it would have had if it 42 * had occurred at end of line. 43 * 4. Each paragraph is sent to standard output as follows. 44 * We output the leading whitespace, and then enough words 45 * to make the line length as near as possible to the goal 46 * without exceeding the maximum. (If a single word would 47 * exceed the maximum, we output that anyway.) Of course 48 * the trailing whitespace of the last word is ignored. 49 * We then emit a newline and start again if there are any 50 * words left. 51 * Note that for a blank line this translates as "We emit 52 * a newline". 53 * If the `-l <n>' option is given, then leading whitespace 54 * is modified slightly: <n> spaces are replaced by a tab. 55 * Indented paragraphs (see above under `-p') make matters 56 * more complicated than this suggests. Actually every paragraph 57 * has two `leading whitespace' values; the value for the first 58 * line, and the value for the most recent line. (While processing 59 * the first line, the two are equal. When `-p' has not been 60 * given, they are always equal.) The leading whitespace 61 * actually output is that of the first line (for the first 62 * line of *output*) or that of the most recent line (for 63 * all other lines of output). 64 * When `-m' has been given, message header paragraphs are 65 * taken as having first-leading-whitespace empty and 66 * subsequent-leading-whitespace two spaces. 67 * 68 * Multiple input files are formatted one at a time, so that a file 69 * never ends in the middle of a line. 70 * 71 * There's an alternative mode of operation, invoked by giving 72 * the `-c' option. In that case we just center every line, 73 * and most of the other options are ignored. This should 74 * really be in a separate program, but we must stay compatible 75 * with old `fmt'. 76 * 77 * QUERY: Should `-m' also try to do the right thing with quoted text? 78 * QUERY: `-b' to treat backslashed whitespace as old `fmt' does? 79 * QUERY: Option meaning `never join lines'? 80 * QUERY: Option meaning `split in mid-word to avoid overlong lines'? 81 * (Those last two might not be useful, since we have `fold'.) 82 * 83 * Differences from old `fmt': 84 * 85 * - We have many more options. Options that aren't understood 86 * generate a lengthy usage message, rather than being 87 * treated as filenames. 88 * - Even with `-m', our handling of message headers is 89 * significantly different. (And much better.) 90 * - We don't treat `\ ' as non-word-breaking. 91 * - Downward changes of indentation start new paragraphs 92 * for us, as well as upward. (I think old `fmt' behaves 93 * in the way it does in order to allow indented paragraphs, 94 * but this is a broken way of making indented paragraphs 95 * behave right.) 96 * - Given the choice of going over or under |goal_length| 97 * by the same amount, we go over; old `fmt' goes under. 98 * - We treat `?' as ending a sentence, and not `:'. Old `fmt' 99 * does the reverse. 100 * - We return approved return codes. Old `fmt' returns 101 * 1 for some errors, and *the number of unopenable files* 102 * when that was all that went wrong. 103 * - We have fewer crashes and more helpful error messages. 104 * - We don't turn spaces into tabs at starts of lines unless 105 * specifically requested. 106 * - New `fmt' is somewhat smaller and slightly faster than 107 * old `fmt'. 108 * 109 * Bugs: 110 * 111 * None known. There probably are some, though. 112 * 113 * Portability: 114 * 115 * I believe this code to be pretty portable. It does require 116 * that you have `getopt'. If you need to include "getopt.h" 117 * for this (e.g., if your system didn't come with `getopt' 118 * and you installed it yourself) then you should arrange for 119 * NEED_getopt_h to be #defined. 120 * 121 * Everything here should work OK even on nasty 16-bit 122 * machines and nice 64-bit ones. However, it's only really 123 * been tested on my FreeBSD machine. Your mileage may vary. 124 * 125 * $FreeBSD: src/usr.bin/fmt/fmt.c,v 1.12.2.2 2002/01/14 09:48:26 ru Exp $ 126 * $DragonFly: src/usr.bin/fmt/fmt.c,v 1.2 2003/06/17 04:29:26 dillon Exp $ 127 */ 128 129 /* Copyright (c) 1997 Gareth McCaughan. All rights reserved. 130 * 131 * Redistribution and use of this code, in source or binary forms, 132 * with or without modification, are permitted subject to the following 133 * conditions: 134 * 135 * - Redistribution of source code must retain the above copyright 136 * notice, this list of conditions and the following disclaimer. 137 * 138 * - If you distribute modified source code it must also include 139 * a notice saying that it has been modified, and giving a brief 140 * description of what changes have been made. 141 * 142 * Disclaimer: I am not responsible for the results of using this code. 143 * If it formats your hard disc, sends obscene messages to 144 * your boss and kills your children then that's your problem 145 * not mine. I give absolutely no warranty of any sort as to 146 * what the program will do, and absolutely refuse to be held 147 * liable for any consequences of your using it. 148 * Thank you. Have a nice day. 149 */ 150 151 /* RCS change log: 152 * Revision 1.5 1998/03/02 18:02:21 gjm11 153 * Minor changes for portability. 154 * 155 * Revision 1.4 1997/10/01 11:51:28 gjm11 156 * Repair broken indented-paragraph handling. 157 * Add mail message header stuff. 158 * Improve comments and layout. 159 * Make usable with non-BSD systems. 160 * Add revision display to usage message. 161 * 162 * Revision 1.3 1997/09/30 16:24:47 gjm11 163 * Add copyright notice, rcsid string and log message. 164 * 165 * Revision 1.2 1997/09/30 16:13:39 gjm11 166 * Add options: -d <chars>, -l <width>, -p, -s, -t <width>, -h . 167 * Parse options with `getopt'. Clean up code generally. 168 * Make comments more accurate. 169 * 170 * Revision 1.1 1997/09/30 11:29:57 gjm11 171 * Initial revision 172 */ 173 174 #include <ctype.h> 175 #include <err.h> 176 #include <locale.h> 177 #include <stdio.h> 178 #include <stdlib.h> 179 #include <string.h> 180 #include <sysexits.h> 181 #include <unistd.h> 182 183 /* Something that, we hope, will never be a genuine line length, 184 * indentation etc. 185 */ 186 #define SILLY ((size_t)-1) 187 188 /* I used to use |strtoul| for this, but (1) not all systems have it 189 * and (2) it's probably better to use |strtol| to detect negative 190 * numbers better. 191 * If |fussyp==0| then we don't complain about non-numbers 192 * (returning 0 instead), but we do complain about bad numbers. 193 */ 194 static size_t 195 get_positive(const char *s, const char *err_mess, int fussyP) { 196 char *t; 197 long result = strtol(s,&t,0); 198 if (*t) { if (fussyP) goto Lose; else return 0; } 199 if (result<=0) { Lose: errx(EX_USAGE, "%s", err_mess); } 200 return (size_t) result; 201 } 202 203 static size_t 204 get_nonnegative(const char *s, const char *err_mess, int fussyP) { 205 char *t; 206 long result = strtol(s,&t,0); 207 if (*t) { if (fussyP) goto Lose; else return 0; } 208 if (result<0) { Lose: errx(EX_USAGE, "%s", err_mess); } 209 return (size_t) result; 210 } 211 212 /* Global variables */ 213 214 static int centerP=0; /* Try to center lines? */ 215 static size_t goal_length=0; /* Target length for output lines */ 216 static size_t max_length=0; /* Maximum length for output lines */ 217 static int coalesce_spaces_P=0; /* Coalesce multiple whitespace -> ' ' ? */ 218 static int allow_indented_paragraphs=0; /* Can first line have diff. ind.? */ 219 static int tab_width=8; /* Number of spaces per tab stop */ 220 static size_t output_tab_width=8; /* Ditto, when squashing leading spaces */ 221 static const char *sentence_enders=".?!"; /* Double-space after these */ 222 static int grok_mail_headers=0; /* treat embedded mail headers magically? */ 223 static int format_troff=0; /* Format troff? */ 224 225 static int n_errors=0; /* Number of failed files. Return on exit. */ 226 static char *output_buffer=0; /* Output line will be built here */ 227 static size_t x; /* Horizontal position in output line */ 228 static size_t x0; /* Ditto, ignoring leading whitespace */ 229 static size_t pending_spaces; /* Spaces to add before next word */ 230 static int output_in_paragraph=0; /* Any of current para written out yet? */ 231 232 /* Prototypes */ 233 234 static void process_named_file (const char *); 235 static void process_stream (FILE *, const char *); 236 static size_t indent_length (const char *, size_t); 237 static int might_be_header (const unsigned char *); 238 static void new_paragraph (size_t, size_t); 239 static void output_word (size_t, size_t, const char *, size_t, size_t); 240 static void output_indent (size_t); 241 static void center_stream (FILE *, const char *); 242 static char * get_line (FILE *, size_t *); 243 static void * xrealloc (void *, size_t); 244 245 #define XMALLOC(x) xrealloc(0,x) 246 247 /* Here is perhaps the right place to mention that this code is 248 * all in top-down order. Hence, |main| comes first. 249 */ 250 int 251 main(int argc, char *argv[]) { 252 int ch; /* used for |getopt| processing */ 253 254 255 (void) setlocale(LC_CTYPE, ""); 256 257 /* 1. Grok parameters. */ 258 259 while ((ch = getopt(argc, argv, "0123456789cd:hl:mnpst:w:")) != -1) 260 switch(ch) { 261 case 'c': 262 centerP = 1; 263 format_troff = 1; 264 continue; 265 case 'd': 266 sentence_enders = optarg; 267 continue; 268 case 'l': 269 output_tab_width 270 = get_nonnegative(optarg, "output tab width must be non-negative", 1); 271 continue; 272 case 'm': 273 grok_mail_headers = 1; 274 continue; 275 case 'n': 276 format_troff = 1; 277 continue; 278 case 'p': 279 allow_indented_paragraphs = 1; 280 continue; 281 case 's': 282 coalesce_spaces_P = 1; 283 continue; 284 case 't': 285 tab_width = get_positive(optarg, "tab width must be positive", 1); 286 continue; 287 case 'w': 288 goal_length = get_positive(optarg, "width must be positive", 1); 289 max_length = goal_length; 290 continue; 291 case '0': case '1': case '2': case '3': case '4': case '5': 292 case '6': case '7': case '8': case '9': 293 /* XXX this is not a stylistically approved use of getopt() */ 294 if (goal_length==0) { 295 char *p; 296 p = argv[optind - 1]; 297 if (p[0] == '-' && p[1] == ch && !p[2]) 298 goal_length = get_positive(++p, "width must be nonzero", 1); 299 else 300 goal_length = get_positive(argv[optind]+1, 301 "width must be nonzero", 1); 302 max_length = goal_length; 303 } 304 continue; 305 case 'h': default: 306 fprintf(stderr, 307 "Usage: fmt [-cmps] [-d chars] [-l num] [-t num]\n" 308 " [-w width | -width | goal [maximum]] [file ...]\n" 309 "Options: -c center each line instead of formatting\n" 310 " -d <chars> double-space after <chars> at line end\n" 311 " -l <n> turn each <n> spaces at start of line into a tab\n" 312 " -m try to make sure mail header lines stay separate\n" 313 " -n format lines beginning with a dot\n" 314 " -p allow indented paragraphs\n" 315 " -s coalesce whitespace inside lines\n" 316 " -t <n> have tabs every <n> columns\n" 317 " -w <n> set maximum width to <n>\n" 318 " goal set target width to goal\n"); 319 exit(ch=='h' ? 0 : EX_USAGE); 320 } 321 argc -= optind; argv += optind; 322 323 /* [ goal [ maximum ] ] */ 324 325 if (argc>0 && goal_length==0 326 && (goal_length=get_positive(*argv,"goal length must be positive", 0)) 327 != 0) { 328 --argc; ++argv; 329 if (argc>0 330 && (max_length=get_positive(*argv,"max length must be positive", 0)) 331 != 0) { 332 --argc; ++argv; 333 if (max_length<goal_length) 334 errx(EX_USAGE, "max length must be >= goal length"); 335 } 336 } 337 if (goal_length==0) goal_length = 65; 338 if (max_length==0) max_length = goal_length+10; 339 output_buffer = XMALLOC(max_length+1); /* really needn't be longer */ 340 341 /* 2. Process files. */ 342 343 if (argc>0) { 344 while (argc-->0) process_named_file(*argv++); 345 } 346 else { 347 process_stream(stdin, "standard input"); 348 } 349 350 /* We're done. */ 351 352 return n_errors ? EX_NOINPUT : 0; 353 354 } 355 356 /* Process a single file, given its name. 357 */ 358 static void 359 process_named_file(const char *name) { 360 FILE *f=fopen(name, "r"); 361 if (!f) { perror(name); ++n_errors; } 362 else { 363 process_stream(f, name); 364 fclose(f); 365 } 366 } 367 368 /* Types of mail header continuation lines: 369 */ 370 typedef enum { 371 hdr_ParagraphStart = -1, 372 hdr_NonHeader = 0, 373 hdr_Header = 1, 374 hdr_Continuation = 2 375 } HdrType; 376 377 /* Process a stream. This is where the real work happens, 378 * except that centering is handled separately. 379 */ 380 static void 381 process_stream(FILE *stream, const char *name) { 382 size_t last_indent=SILLY; /* how many spaces in last indent? */ 383 size_t para_line_number=0; /* how many lines already read in this para? */ 384 size_t first_indent=SILLY; /* indentation of line 0 of paragraph */ 385 HdrType prev_header_type=hdr_ParagraphStart; 386 /* ^-- header_type of previous line; -1 at para start */ 387 char *line; 388 size_t length; 389 390 if (centerP) { center_stream(stream, name); return; } 391 while ((line=get_line(stream,&length)) != NULL) { 392 size_t np=indent_length(line, length); 393 { HdrType header_type=hdr_NonHeader; 394 if (grok_mail_headers && prev_header_type!=hdr_NonHeader) { 395 if (np==0 && might_be_header(line)) 396 header_type = hdr_Header; 397 else if (np>0 && prev_header_type>hdr_NonHeader) 398 header_type = hdr_Continuation; 399 } 400 /* We need a new paragraph if and only if: 401 * this line is blank, 402 * OR it's a troff request (and we don't format troff), 403 * OR it's a mail header, 404 * OR it's not a mail header AND the last line was one, 405 * OR the indentation has changed 406 * AND the line isn't a mail header continuation line 407 * AND this isn't the second line of an indented paragraph. 408 */ 409 if ( length==0 410 || (line[0]=='.' && !format_troff) 411 || header_type==hdr_Header 412 || (header_type==hdr_NonHeader && prev_header_type>hdr_NonHeader) 413 || (np!=last_indent 414 && header_type != hdr_Continuation 415 && (!allow_indented_paragraphs || para_line_number != 1)) ) { 416 new_paragraph(output_in_paragraph ? last_indent : first_indent, np); 417 para_line_number = 0; 418 first_indent = np; 419 last_indent = np; 420 if (header_type==hdr_Header) last_indent=2; /* for cont. lines */ 421 if (length==0 || (line[0]=='.' && !format_troff)) { 422 if (length==0) 423 putchar('\n'); 424 else 425 printf("%.*s\n", (int)length, line); 426 prev_header_type=hdr_ParagraphStart; 427 continue; 428 } 429 } 430 else { 431 /* If this is an indented paragraph other than a mail header 432 * continuation, set |last_indent|. 433 */ 434 if (np != last_indent && header_type != hdr_Continuation) 435 last_indent=np; 436 } 437 prev_header_type = header_type; 438 } 439 440 { size_t n=np; 441 while (n<length) { 442 /* Find word end and count spaces after it */ 443 size_t word_length=0, space_length=0; 444 while (n+word_length < length && line[n+word_length] != ' ') 445 ++word_length; 446 space_length = word_length; 447 while (n+space_length < length && line[n+space_length] == ' ') 448 ++space_length; 449 /* Send the word to the output machinery. */ 450 output_word(first_indent, last_indent, 451 line+n, word_length, space_length-word_length); 452 n += space_length; 453 } 454 } 455 ++para_line_number; 456 } 457 new_paragraph(output_in_paragraph ? last_indent : first_indent, 0); 458 if (ferror(stream)) { perror(name); ++n_errors; } 459 } 460 461 /* How long is the indent on this line? 462 */ 463 static size_t 464 indent_length(const char *line, size_t length) { 465 size_t n=0; 466 while (n<length && *line++ == ' ') ++n; 467 return n; 468 } 469 470 /* Might this line be a mail header? 471 * We deem a line to be a possible header if it matches the 472 * Perl regexp /^[A-Z][-A-Za-z0-9]*:\s/. This is *not* the same 473 * as in RFC whatever-number-it-is; we want to be gratuitously 474 * conservative to avoid mangling ordinary civilised text. 475 */ 476 static int 477 might_be_header(const unsigned char *line) { 478 if (!isupper(*line++)) return 0; 479 while (*line && (isalnum(*line) || *line=='-')) ++line; 480 return (*line==':' && isspace(line[1])); 481 } 482 483 /* Begin a new paragraph with an indent of |indent| spaces. 484 */ 485 static void 486 new_paragraph(size_t old_indent, size_t indent) { 487 if (x0) { 488 if (old_indent>0) output_indent(old_indent); 489 fwrite(output_buffer, 1, x0, stdout); 490 putchar('\n'); 491 } 492 x=indent; x0=0; pending_spaces=0; 493 output_in_paragraph = 0; 494 } 495 496 /* Output spaces or tabs for leading indentation. 497 */ 498 static void 499 output_indent(size_t n_spaces) { 500 if (output_tab_width) { 501 while (n_spaces >= output_tab_width) { 502 putchar('\t'); 503 n_spaces -= output_tab_width; 504 } 505 } 506 while (n_spaces-- > 0) putchar(' '); 507 } 508 509 /* Output a single word, or add it to the buffer. 510 * indent0 and indent1 are the indents to use on the first and subsequent 511 * lines of a paragraph. They'll often be the same, of course. 512 */ 513 static void 514 output_word(size_t indent0, size_t indent1, const char *word, size_t length, size_t spaces) { 515 size_t new_x = x+pending_spaces+length; 516 size_t indent = output_in_paragraph ? indent1 : indent0; 517 518 /* If either |spaces==0| (at end of line) or |coalesce_spaces_P| 519 * (squashing internal whitespace), then add just one space; 520 * except that if the last character was a sentence-ender we 521 * actually add two spaces. 522 */ 523 if (coalesce_spaces_P || spaces==0) 524 spaces = strchr(sentence_enders, word[length-1]) ? 2 : 1; 525 526 if (new_x<=goal_length) { 527 /* After adding the word we still aren't at the goal length, 528 * so clearly we add it to the buffer rather than outputing it. 529 */ 530 memset(output_buffer+x0, ' ', pending_spaces); 531 x0 += pending_spaces; x += pending_spaces; 532 memcpy(output_buffer+x0, word, length); 533 x0 += length; x += length; 534 pending_spaces = spaces; 535 } 536 else { 537 /* Adding the word takes us past the goal. Print the line-so-far, 538 * and the word too iff either (1) the lsf is empty or (2) that 539 * makes us nearer the goal but doesn't take us over the limit, 540 * or (3) the word on its own takes us over the limit. 541 * In case (3) we put a newline in between. 542 */ 543 if (indent>0) output_indent(indent); 544 fwrite(output_buffer, 1, x0, stdout); 545 if (x0==0 || (new_x <= max_length && new_x-goal_length <= goal_length-x)) { 546 printf("%*s", (int)pending_spaces, ""); 547 goto write_out_word; 548 } 549 else { 550 /* If the word takes us over the limit on its own, just 551 * spit it out and don't bother buffering it. 552 */ 553 if (indent+length > max_length) { 554 putchar('\n'); 555 if (indent>0) output_indent(indent); 556 write_out_word: 557 fwrite(word, 1, length, stdout); 558 x0 = 0; x = indent1; pending_spaces = 0; 559 } 560 else { 561 memcpy(output_buffer, word, length); 562 x0 = length; x = length+indent1; pending_spaces = spaces; 563 } 564 } 565 putchar('\n'); 566 output_in_paragraph = 1; 567 } 568 } 569 570 /* Process a stream, but just center its lines rather than trying to 571 * format them neatly. 572 */ 573 static void 574 center_stream(FILE *stream, const char *name) { 575 char *line; 576 size_t length; 577 while ((line=get_line(stream, &length)) != 0) { 578 size_t l=length; 579 while (l>0 && isspace(*line)) { ++line; --l; } 580 length=l; 581 while (l<goal_length) { putchar(' '); l+=2; } 582 fwrite(line, 1, length, stdout); 583 putchar('\n'); 584 } 585 if (ferror(stream)) { perror(name); ++n_errors; } 586 } 587 588 /* Get a single line from a stream. Expand tabs, strip control 589 * characters and trailing whitespace, and handle backspaces. 590 * Return the address of the buffer containing the line, and 591 * put the length of the line in |lengthp|. 592 * This can cope with arbitrarily long lines, and with lines 593 * without terminating \n. 594 * If there are no characters left or an error happens, we 595 * return 0. 596 * Don't confuse |spaces_pending| here with the global 597 * |pending_spaces|. 598 */ 599 static char * 600 get_line(FILE *stream, size_t *lengthp) { 601 static char *buf=NULL; 602 static size_t length=0; 603 size_t len=0; 604 int ch; 605 size_t spaces_pending=0; 606 int troff=0; 607 608 if (buf==NULL) { length=100; buf=XMALLOC(length); } 609 while ((ch=getc(stream)) != '\n' && ch != EOF) { 610 if (len+spaces_pending==0 && ch=='.' && !format_troff) troff=1; 611 if (ch==' ') ++spaces_pending; 612 else if (troff || isprint(ch)) { 613 while (len+spaces_pending >= length) { 614 length*=2; buf=xrealloc(buf, length); 615 } 616 while (spaces_pending > 0) { --spaces_pending; buf[len++]=' '; } 617 buf[len++] = ch; 618 } 619 else if (ch=='\t') 620 spaces_pending += tab_width - (len+spaces_pending)%tab_width; 621 else if (ch=='\b') { if (len) --len; } 622 } 623 *lengthp=len; 624 return (len>0 || ch!=EOF) ? buf : 0; 625 } 626 627 /* (Re)allocate some memory, exiting with an error if we can't. 628 */ 629 static void * 630 xrealloc(void *ptr, size_t nbytes) { 631 void *p = realloc(ptr, nbytes); 632 if (p == NULL) errx(EX_OSERR, "out of memory"); 633 return p; 634 } 635