[BACK]Return to fmt.c CVS log [TXT][DIR] Up to [local] / src / usr.bin / fmt

Annotation of src/usr.bin/fmt/fmt.c, Revision 1.16

1.16    ! pjanzen     1: /*     $OpenBSD: fmt.c,v 1.15 1999/12/30 18:00:42 provos Exp $ */
1.1       deraadt     2:
1.10      millert     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:  * 3. The "everything else" is split into words; a word
                     34:  *    includes its trailing whitespace, and a word at the
                     35:  *    end of a line is deemed to be followed by a single
                     36:  *    space, or two spaces if it ends with a sentence-end
                     37:  *    character. (See the `-d' option for how to change that.)
                     38:  *    If the `-s' option has been given, then a word's trailing
                     39:  *    whitespace is replaced by what it would have had if it
                     40:  *    had occurred at end of line.
                     41:  * 4. Each paragraph is sent to standard output as follows.
                     42:  *    We output the leading whitespace, and then enough words
                     43:  *    to make the line length as near as possible to the goal
                     44:  *    without exceeding the maximum. (If a single word would
                     45:  *    exceed the maximum, we output that anyway.) Of course
                     46:  *    the trailing whitespace of the last word is ignored.
                     47:  *    We then emit a newline and start again if there are any
                     48:  *    words left.
                     49:  *    Note that for a blank line this translates as "We emit
                     50:  *    a newline".
                     51:  *    If the `-l <n>' option is given, then leading whitespace
                     52:  *    is modified slightly: <n> spaces are replaced by a tab.
                     53:  *    Indented paragraphs (see above under `-p') make matters
                     54:  *    more complicated than this suggests. Actually every paragraph
                     55:  *    has two `leading whitespace' values; the value for the first
                     56:  *    line, and the value for the most recent line. (While processing
                     57:  *    the first line, the two are equal. When `-p' has not been
                     58:  *    given, they are always equal.) The leading whitespace
                     59:  *    actually output is that of the first line (for the first
                     60:  *    line of *output*) or that of the most recent line (for
                     61:  *    all other lines of output).
                     62:  *    When `-m' has been given, message header paragraphs are
                     63:  *    taken as having first-leading-whitespace empty and
                     64:  *    subsequent-leading-whitespace two spaces.
                     65:  *
                     66:  * Multiple input files are formatted one at a time, so that a file
                     67:  * never ends in the middle of a line.
                     68:  *
                     69:  * There's an alternative mode of operation, invoked by giving
                     70:  * the `-c' option. In that case we just center every line,
                     71:  * and most of the other options are ignored. This should
                     72:  * really be in a separate program, but we must stay compatible
                     73:  * with old `fmt'.
                     74:  *
                     75:  * QUERY: Should `-m' also try to do the right thing with quoted text?
                     76:  * QUERY: `-b' to treat backslashed whitespace as old `fmt' does?
                     77:  * QUERY: Option meaning `never join lines'?
                     78:  * QUERY: Option meaning `split in mid-word to avoid overlong lines'?
                     79:  * (Those last two might not be useful, since we have `fold'.)
                     80:  *
                     81:  * Differences from old `fmt':
                     82:  *
                     83:  *   - We have many more options. Options that aren't understood
                     84:  *     generate a lengthy usage message, rather than being
                     85:  *     treated as filenames.
                     86:  *   - Even with `-m', our handling of message headers is
                     87:  *     significantly different. (And much better.)
                     88:  *   - We don't treat `\ ' as non-word-breaking.
                     89:  *   - Downward changes of indentation start new paragraphs
                     90:  *     for us, as well as upward. (I think old `fmt' behaves
                     91:  *     in the way it does in order to allow indented paragraphs,
                     92:  *     but this is a broken way of making indented paragraphs
                     93:  *     behave right.)
                     94:  *   - Given the choice of going over or under |goal_length|
                     95:  *     by the same amount, we go over; old `fmt' goes under.
                     96:  *   - We treat `?' as ending a sentence, and not `:'. Old `fmt'
                     97:  *     does the reverse.
                     98:  *   - We return approved return codes. Old `fmt' returns
                     99:  *     1 for some errors, and *the number of unopenable files*
                    100:  *     when that was all that went wrong.
                    101:  *   - We have fewer crashes and more helpful error messages.
                    102:  *   - We don't turn spaces into tabs at starts of lines unless
                    103:  *     specifically requested.
                    104:  *   - New `fmt' is somewhat smaller and slightly faster than
                    105:  *     old `fmt'.
                    106:  *
                    107:  * Bugs:
                    108:  *
                    109:  *   None known. There probably are some, though.
                    110:  *
                    111:  * Portability:
                    112:  *
                    113:  *   I believe this code to be pretty portable. It does require
                    114:  *   that you have `getopt'. If you need to include "getopt.h"
                    115:  *   for this (e.g., if your system didn't come with `getopt'
                    116:  *   and you installed it yourself) then you should arrange for
                    117:  *   NEED_getopt_h to be #defined.
                    118:  *
                    119:  *   Everything here should work OK even on nasty 16-bit
                    120:  *   machines and nice 64-bit ones. However, it's only really
                    121:  *   been tested on my FreeBSD machine. Your mileage may vary.
                    122:  */
                    123:
                    124: /* Copyright (c) 1997 Gareth McCaughan. All rights reserved.
                    125:  *
                    126:  * Redistribution and use of this code, in source or binary forms,
                    127:  * with or without modification, are permitted subject to the following
                    128:  * conditions:
                    129:  *
                    130:  *  - Redistribution of source code must retain the above copyright
1.1       deraadt   131:  *    notice, this list of conditions and the following disclaimer.
1.10      millert   132:  *
                    133:  *  - If you distribute modified source code it must also include
                    134:  *    a notice saying that it has been modified, and giving a brief
                    135:  *    description of what changes have been made.
                    136:  *
                    137:  * Disclaimer: I am not responsible for the results of using this code.
                    138:  *             If it formats your hard disc, sends obscene messages to
                    139:  *             your boss and kills your children then that's your problem
                    140:  *             not mine. I give absolutely no warranty of any sort as to
                    141:  *             what the program will do, and absolutely refuse to be held
                    142:  *             liable for any consequences of your using it.
                    143:  *             Thank you. Have a nice day.
                    144:  */
                    145:
                    146: /* RCS change log:
                    147:  * Revision 1.5  1998/03/02 18:02:21  gjm11
                    148:  * Minor changes for portability.
                    149:  *
                    150:  * Revision 1.4  1997/10/01 11:51:28  gjm11
                    151:  * Repair broken indented-paragraph handling.
                    152:  * Add mail message header stuff.
                    153:  * Improve comments and layout.
                    154:  * Make usable with non-BSD systems.
                    155:  * Add revision display to usage message.
                    156:  *
                    157:  * Revision 1.3  1997/09/30 16:24:47  gjm11
                    158:  * Add copyright notice, rcsid string and log message.
                    159:  *
                    160:  * Revision 1.2  1997/09/30 16:13:39  gjm11
                    161:  * Add options: -d <chars>, -l <width>, -p, -s, -t <width>, -h .
                    162:  * Parse options with `getopt'. Clean up code generally.
                    163:  * Make comments more accurate.
                    164:  *
                    165:  * Revision 1.1  1997/09/30 11:29:57  gjm11
                    166:  * Initial revision
1.1       deraadt   167:  */
                    168:
                    169: #ifndef lint
1.10      millert   170: static const char rcsid[] =
1.16    ! pjanzen   171:   "$OpenBSD: fmt.c,v 1.15 1999/12/30 18:00:42 provos Exp $";
1.10      millert   172: static const char copyright[] =
                    173:   "Copyright (c) 1997 Gareth McCaughan. All rights reserved.\n";
1.1       deraadt   174: #endif /* not lint */
                    175:
1.10      millert   176: /* Cater for BSD and non-BSD systems.
                    177:  * I hate the C preprocessor.
                    178:  */
                    179:
                    180: #undef HAVE_errx
                    181: #undef HAVE_sysexits
                    182:
                    183: #ifdef unix
                    184: # include <sys/param.h>
                    185: # ifdef BSD
                    186: #  define HAVE_errx
                    187: #  if BSD >= 199306
                    188: #   define HAVE_sysexits
                    189: #  endif
                    190: # endif
                    191: #endif
                    192:
                    193: #ifdef HAVE_errx
                    194: # include <err.h>
                    195: #else
                    196: # define errx(rc,str) { fprintf(stderr,"fmt: %s\n",str); exit(rc); }
                    197: #endif
                    198:
                    199: #ifdef HAVE_sysexits
                    200: # include <sysexits.h>
1.4       millert   201: #else
1.10      millert   202: # define EX_USAGE 1
                    203: # define EX_NOINPUT 1
                    204: # define EX_SOFTWARE 1
                    205: # define EX_OSERR 1
1.1       deraadt   206: #endif
                    207:
1.4       millert   208: #include <ctype.h>
1.1       deraadt   209: #include <stdio.h>
                    210: #include <stdlib.h>
                    211: #include <string.h>
                    212:
1.10      millert   213: #ifdef NEED_getopt_h
                    214: # include "getopt.h"
                    215: #endif
                    216:
                    217: /* Something that, we hope, will never be a genuine line length,
                    218:  * indentation etc.
1.1       deraadt   219:  */
1.10      millert   220: #define SILLY ((size_t)-1)
1.1       deraadt   221:
1.10      millert   222: /* I used to use |strtoul| for this, but (1) not all systems have it
                    223:  * and (2) it's probably better to use |strtol| to detect negative
                    224:  * numbers better.
                    225:  * If |fussyp==0| then we don't complain about non-numbers
                    226:  * (returning 0 instead), but we do complain about bad numbers.
                    227:  */
                    228: size_t get_positive(const char *s, const char *err_mess, int fussyP) {
                    229:   char *t;
                    230:   long result = strtol(s,&t,0);
                    231:   if (*t) { if (fussyP) goto Lose; else return 0; }
                    232:   if (result<=0) { Lose: errx(EX_USAGE, err_mess); }
                    233:   return (size_t) result;
1.1       deraadt   234: }
                    235:
1.10      millert   236: /* Just for the sake of linguistic purity: */
                    237:
                    238: #ifdef BRITISH
                    239: # define CENTER "centre"
                    240: #else
                    241: # define CENTER "center"
                    242: #endif
                    243:
                    244: /* Global variables */
                    245:
                    246: static int centerP=0;          /* Try to center lines? */
                    247: static size_t goal_length=0;   /* Target length for output lines */
                    248: static size_t max_length=0;    /* Maximum length for output lines */
                    249: static int coalesce_spaces_P=0;        /* Coalesce multiple whitespace -> ' ' ? */
                    250: static int allow_indented_paragraphs=0;        /* Can first line have diff. ind.? */
                    251: static int tab_width=8;                /* Number of spaces per tab stop */
                    252: static int output_tab_width=0; /* Ditto, when squashing leading spaces */
                    253: static char *sentence_enders=".?!";    /* Double-space after these */
                    254: static int grok_mail_headers=0;        /* treat embedded mail headers magically? */
                    255:
                    256: static int n_errors=0;         /* Number of failed files. Return on exit. */
                    257: static char *output_buffer=0;  /* Output line will be built here */
                    258: static size_t x;               /* Horizontal position in output line */
                    259: static size_t x0;              /* Ditto, ignoring leading whitespace */
                    260: static size_t pending_spaces;  /* Spaces to add before next word */
                    261: static int output_in_paragraph=0;      /* Any of current para written out yet? */
                    262:
                    263: /* Prototypes */
                    264:
                    265: static void process_named_file (const char *);
                    266: static void     process_stream (FILE *, const char *);
                    267: static size_t    indent_length (const char *, size_t);
                    268: static int     might_be_header (const char *);
                    269: static void      new_paragraph (size_t, size_t);
                    270: static void        output_word (size_t, size_t, const char *, size_t, size_t);
                    271: static void      output_indent (size_t);
                    272: static void      center_stream (FILE *, const char *);
                    273: static char *         get_line (FILE *, size_t *);
                    274: static void *         xrealloc (void *, size_t);
                    275:
                    276: #define XMALLOC(x) xrealloc(0,x)
                    277:
                    278: /* Here is perhaps the right place to mention that this code is
                    279:  * all in top-down order. Hence, |main| comes first.
1.1       deraadt   280:  */
1.4       millert   281: int
1.10      millert   282: main(int argc, char *argv[]) {
                    283:   int ch;                      /* used for |getopt| processing */
                    284:
                    285:   /* 1. Grok parameters. */
                    286:
1.16    ! pjanzen   287:   while ((ch = getopt(argc, argv, "0123456789cd:hl:mpst:w:")) != -1)
        !           288:   switch(ch) {
1.10      millert   289:     case 'c':
                    290:       centerP = 1;
                    291:       continue;
                    292:     case 'd':
                    293:       sentence_enders = XMALLOC(strlen(optarg)+1);
1.14      deraadt   294:       strcpy(sentence_enders, optarg); /* ok */
1.10      millert   295:       continue;
                    296:     case 'l':
                    297:       output_tab_width
                    298:         = get_positive(optarg, "output tab width must be positive", 1);
                    299:       continue;
                    300:     case 'm':
                    301:       grok_mail_headers = 1;
                    302:       continue;
                    303:     case 'p':
                    304:       allow_indented_paragraphs = 1;
                    305:       continue;
                    306:     case 's':
                    307:       coalesce_spaces_P = 1;
                    308:       continue;
                    309:     case 't':
                    310:       tab_width = get_positive(optarg, "tab width must be positive", 1);
                    311:       continue;
1.16    ! pjanzen   312:     case 'w':
        !           313:       goal_length = get_positive(optarg, "width must be positive", 1);
        !           314:       max_length = goal_length;
        !           315:       continue;
        !           316:     case '0': case '1': case '2': case '3': case '4': case '5':
        !           317:     case '6': case '7': case '8': case '9':
        !           318:     /* XXX  this is not a stylistically approved use of getopt() */
        !           319:       if (goal_length==0) {
        !           320:         char *p;
        !           321:         p = argv[optind - 1];
        !           322:         if (p[0] == '-' && p[1] == ch && !p[2])
        !           323:              goal_length = get_positive(++p, "width must be nonzero", 1);
        !           324:         else
        !           325:              goal_length = get_positive(argv[optind]+1,
        !           326:                  "width must be nonzero", 1);
        !           327:         max_length = goal_length;
        !           328:       }
        !           329:       continue;
1.10      millert   330:     case 'h': default:
                    331:       fprintf(stderr,
1.16    ! pjanzen   332: "Usage:   fmt [-cmps] [-d chars] [-l num] [-t num]\n"
        !           333: "             [-w width | -width | goal [maximum]] [file ...]\n"
1.10      millert   334: "Options: -c     " CENTER " each line instead of formatting\n"
                    335: "         -d <chars> double-space after <chars> at line end\n"
                    336: "         -l <n> turn each <n> spaces at start of line into a tab\n"
                    337: "         -m     try to make sure mail header lines stay separate\n"
                    338: "         -p     allow indented paragraphs\n"
                    339: "         -s     coalesce whitespace inside lines\n"
1.16    ! pjanzen   340: "         -t <n> have tabs every <n> columns\n"
        !           341: "         -w <n> set maximum width to <n>\n"
        !           342: "         goal   set target width to goal\n");
1.10      millert   343:       exit(ch=='h' ? 0 : EX_USAGE);
                    344:   }
                    345:   argc -= optind; argv += optind;
                    346:
                    347:   /* [ goal [ maximum ] ] */
                    348:
1.16    ! pjanzen   349:   if (argc>0 && goal_length==0
1.10      millert   350:       && (goal_length=get_positive(*argv,"goal length must be positive", 0))
                    351:          != 0) {
                    352:     --argc; ++argv;
                    353:     if (argc>0
1.12      millert   354:         && (max_length=get_positive(*argv,"max length must be positive", 0))
1.10      millert   355:            != 0) {
1.13      millert   356:       --argc; ++argv;
1.10      millert   357:       if (max_length<goal_length)
                    358:         errx(EX_USAGE, "max length must be >= goal length");
                    359:     }
                    360:   }
                    361:   if (goal_length==0) goal_length = 65;
                    362:   if (max_length==0) max_length = goal_length+10;
                    363:   output_buffer = XMALLOC(max_length+1);       /* really needn't be longer */
                    364:
                    365:   /* 2. Process files. */
                    366:
                    367:   if (argc>0) {
                    368:     while (argc-->0) process_named_file(*argv++);
                    369:   }
                    370:   else {
                    371:     process_stream(stdin, "standard input");
                    372:   }
                    373:
                    374:   /* We're done. */
                    375:
                    376:   return n_errors ? EX_NOINPUT : 0;
                    377:
                    378: }
                    379:
                    380: /* Process a single file, given its name.
                    381:  */
                    382: static void
                    383: process_named_file(const char *name) {
                    384:   FILE *f=fopen(name, "r");
                    385:   if (!f) { perror(name); ++n_errors; }
                    386:   else {
                    387:     process_stream(f, name);
                    388:     fclose(f);
                    389:   }
                    390: }
                    391:
                    392: /* Types of mail header continuation lines:
                    393:  */
                    394: typedef enum {
                    395:   hdr_ParagraphStart = -1,
                    396:   hdr_NonHeader      = 0,
                    397:   hdr_Header         = 1,
                    398:   hdr_Continuation   = 2
                    399: } HdrType;
                    400:
                    401: /* Process a stream. This is where the real work happens,
                    402:  * except that centering is handled separately.
                    403:  */
                    404: static void
                    405: process_stream(FILE *stream, const char *name) {
                    406:   size_t last_indent=SILLY;    /* how many spaces in last indent? */
                    407:   size_t para_line_number=0;   /* how many lines already read in this para? */
                    408:   size_t first_indent=SILLY;   /* indentation of line 0 of paragraph */
                    409:   HdrType prev_header_type=hdr_ParagraphStart;
                    410:        /* ^-- header_type of previous line; -1 at para start */
                    411:   char *line;
                    412:   size_t length;
                    413:
                    414:   if (centerP) { center_stream(stream, name); return; }
                    415:   while ((line=get_line(stream,&length)) != NULL) {
                    416:     size_t np=indent_length(line, length);
                    417:     { HdrType header_type=hdr_NonHeader;
                    418:       if (grok_mail_headers && prev_header_type!=hdr_NonHeader) {
                    419:         if (np==0 && might_be_header(line))
                    420:           header_type = hdr_Header;
                    421:         else if (np>0 && prev_header_type>hdr_NonHeader)
                    422:           header_type = hdr_Continuation;
                    423:       }
                    424:       /* We need a new paragraph if and only if:
                    425:        *   this line is blank,
                    426:        *   OR it's a mail header,
                    427:        *   OR it's not a mail header AND the last line was one,
                    428:        *   OR the indentation has changed
                    429:        *      AND the line isn't a mail header continuation line
                    430:        *      AND this isn't the second line of an indented paragraph.
                    431:        */
                    432:       if ( length==0
                    433:            || header_type==hdr_Header
                    434:            || (header_type==hdr_NonHeader && prev_header_type>hdr_NonHeader)
                    435:            || (np!=last_indent
                    436:                && header_type != hdr_Continuation
                    437:                && (!allow_indented_paragraphs || para_line_number != 1)) ) {
                    438:         new_paragraph(output_in_paragraph ? last_indent : first_indent, np);
                    439:         para_line_number = 0;
                    440:         first_indent = np;
                    441:         last_indent = np;
                    442:         if (header_type==hdr_Header) last_indent=2;    /* for cont. lines */
                    443:         if (length==0) {
                    444:           putchar('\n');
                    445:           prev_header_type=hdr_ParagraphStart;
                    446:           continue;
                    447:         }
                    448:       }
                    449:       else {
                    450:         /* If this is an indented paragraph other than a mail header
                    451:          * continuation, set |last_indent|.
                    452:          */
                    453:         if (np != last_indent && header_type != hdr_Continuation)
                    454:           last_indent=np;
                    455:       }
                    456:       prev_header_type = header_type;
                    457:     }
                    458:
                    459:     { size_t n=np;
                    460:       while (n<length) {
                    461:         /* Find word end and count spaces after it */
                    462:         size_t word_length=0, space_length=0;
                    463:         while (n+word_length < length && line[n+word_length] != ' ')
                    464:           ++word_length;
                    465:         space_length = word_length;
                    466:         while (n+space_length < length && line[n+space_length] == ' ')
                    467:           ++space_length;
                    468:         /* Send the word to the output machinery. */
                    469:         output_word(first_indent, last_indent,
                    470:                     line+n, word_length, space_length-word_length);
                    471:         n += space_length;
                    472:       }
                    473:     }
                    474:     ++para_line_number;
                    475:   }
                    476:   new_paragraph(output_in_paragraph ? last_indent : first_indent, 0);
                    477:   if (ferror(stream)) { perror(name); ++n_errors; }
                    478: }
                    479:
                    480: /* How long is the indent on this line?
                    481:  */
                    482: static size_t
                    483: indent_length(const char *line, size_t length) {
                    484:   size_t n=0;
                    485:   while (n<length && *line++ == ' ') ++n;
                    486:   return n;
                    487: }
                    488:
                    489: /* Might this line be a mail header?
                    490:  * We deem a line to be a possible header if it matches the
                    491:  * Perl regexp /^[A-Z][-A-Za-z0-9]*:\s/. This is *not* the same
                    492:  * as in RFC whatever-number-it-is; we want to be gratuitously
                    493:  * conservative to avoid mangling ordinary civilised text.
                    494:  */
                    495: static int
                    496: might_be_header(const char *line) {
                    497:   if (!isupper(*line++)) return 0;
                    498:   while (*line && (isalnum(*line) || *line=='-')) ++line;
                    499:   return (*line==':' && isspace(line[1]));
                    500: }
                    501:
                    502: /* Begin a new paragraph with an indent of |indent| spaces.
                    503:  */
                    504: static void
                    505: new_paragraph(size_t old_indent, size_t indent) {
                    506:   if (x0) {
                    507:     if (old_indent>0) output_indent(old_indent);
                    508:     fwrite(output_buffer, 1, x0, stdout);
                    509:     putchar('\n');
                    510:   }
                    511:   x=indent; x0=0; pending_spaces=0;
                    512:   output_in_paragraph = 0;
                    513: }
                    514:
                    515: /* Output spaces or tabs for leading indentation.
                    516:  */
                    517: static void
                    518: output_indent(size_t n_spaces) {
                    519:   if (output_tab_width) {
                    520:     while (n_spaces >= output_tab_width) {
                    521:       putchar('\t');
                    522:       n_spaces -= output_tab_width;
                    523:     }
                    524:   }
                    525:   while (n_spaces-- > 0) putchar(' ');
                    526: }
                    527:
                    528: /* Output a single word, or add it to the buffer.
                    529:  * indent0 and indent1 are the indents to use on the first and subsequent
                    530:  * lines of a paragraph. They'll often be the same, of course.
                    531:  */
                    532: static void
                    533: output_word(size_t indent0, size_t indent1, const char *word, size_t length, size_t spaces) {
                    534:   size_t new_x = x+pending_spaces+length;
                    535:   size_t indent = output_in_paragraph ? indent1 : indent0;
                    536:
                    537:   /* If either |spaces==0| (at end of line) or |coalesce_spaces_P|
                    538:    * (squashing internal whitespace), then add just one space;
                    539:    * except that if the last character was a sentence-ender we
                    540:    * actually add two spaces.
                    541:    */
                    542:   if (coalesce_spaces_P || spaces==0)
                    543:     spaces = strchr(sentence_enders, word[length-1]) ? 2 : 1;
                    544:
                    545:   if (new_x<=goal_length) {
                    546:     /* After adding the word we still aren't at the goal length,
                    547:      * so clearly we add it to the buffer rather than outputing it.
                    548:      */
                    549:     memset(output_buffer+x0, ' ', pending_spaces);
                    550:     x0 += pending_spaces; x += pending_spaces;
                    551:     memcpy(output_buffer+x0, word, length);
                    552:     x0 += length; x += length;
                    553:     pending_spaces = spaces;
                    554:   }
                    555:   else {
                    556:     /* Adding the word takes us past the goal. Print the line-so-far,
                    557:      * and the word too iff either (1) the lsf is empty or (2) that
                    558:      * makes us nearer the goal but doesn't take us over the limit,
                    559:      * or (3) the word on its own takes us over the limit.
                    560:      * In case (3) we put a newline in between.
                    561:      */
                    562:     if (indent>0) output_indent(indent);
                    563:     fwrite(output_buffer, 1, x0, stdout);
                    564:     if (x0==0 || (new_x <= max_length && new_x-goal_length <= goal_length-x)) {
                    565:       printf("%*s", pending_spaces, "");
                    566:       goto write_out_word;
                    567:     }
                    568:     else {
                    569:       /* If the word takes us over the limit on its own, just
                    570:        * spit it out and don't bother buffering it.
                    571:        */
                    572:       if (indent+length > max_length) {
                    573:         putchar('\n');
                    574:         if (indent>0) output_indent(indent);
                    575: write_out_word:
                    576:         fwrite(word, 1, length, stdout);
                    577:         x0 = 0; x = indent1; pending_spaces = 0;
                    578:       }
                    579:       else {
                    580:         memcpy(output_buffer, word, length);
                    581:         x0 = length; x = length+indent1; pending_spaces = spaces;
                    582:       }
                    583:     }
                    584:     putchar('\n');
                    585:     output_in_paragraph = 1;
                    586:   }
                    587: }
                    588:
                    589: /* Process a stream, but just center its lines rather than trying to
                    590:  * format them neatly.
                    591:  */
                    592: static void
                    593: center_stream(FILE *stream, const char *name) {
                    594:   char *line;
                    595:   size_t length;
                    596:   while ((line=get_line(stream, &length)) != 0) {
                    597:     size_t l=length;
                    598:     while (l>0 && isspace(*line)) { ++line; --l; }
                    599:     length=l;
                    600:     while (l<goal_length) { putchar(' '); l+=2; }
                    601:     fwrite(line, 1, length, stdout);
                    602:     putchar('\n');
                    603:   }
                    604:   if (ferror(stream)) { perror(name); ++n_errors; }
                    605: }
                    606:
                    607: /* Get a single line from a stream. Expand tabs, strip control
                    608:  * characters and trailing whitespace, and handle backspaces.
                    609:  * Return the address of the buffer containing the line, and
                    610:  * put the length of the line in |lengthp|.
                    611:  * This can cope with arbitrarily long lines, and with lines
                    612:  * without terminating \n.
                    613:  * If there are no characters left or an error happens, we
                    614:  * return 0.
                    615:  * Don't confuse |spaces_pending| here with the global
                    616:  * |pending_spaces|.
                    617:  */
                    618: static char *
                    619: get_line(FILE *stream, size_t *lengthp) {
                    620:   static char *buf=NULL;
                    621:   static size_t length=0;
                    622:   size_t len=0;
                    623:   int ch;
                    624:   size_t spaces_pending=0;
                    625:
                    626:   if (buf==NULL) { length=100; buf=XMALLOC(length); }
                    627:   while ((ch=getc(stream)) != '\n' && ch != EOF) {
                    628:     if (ch==' ') ++spaces_pending;
1.15      provos    629:     else if (!iscntrl(ch)) {
1.10      millert   630:       while (len+spaces_pending >= length) {
                    631:         length*=2; buf=xrealloc(buf, length);
                    632:       }
                    633:       while (spaces_pending > 0) { --spaces_pending; buf[len++]=' '; }
                    634:       buf[len++] = ch;
                    635:     }
                    636:     else if (ch=='\t')
                    637:       spaces_pending += tab_width - (len+spaces_pending)%tab_width;
                    638:     else if (ch=='\b') { if (len) --len; }
                    639:   }
                    640:   *lengthp=len;
                    641:   return (len>0 || ch!=EOF) ? buf : 0;
                    642: }
                    643:
                    644: /* (Re)allocate some memory, exiting with an error if we can't.
                    645:  */
                    646: static void *
                    647: xrealloc(void *ptr, size_t nbytes) {
                    648:   void *p = realloc(ptr, nbytes);
                    649:   if (p == NULL) errx(EX_OSERR, "out of memory");
                    650:   return p;
1.1       deraadt   651: }