src/basic/string-util.c

   1 /* SPDX-License-Identifier: LGPL-2.1+ */
   2
   3 #include <errno.h>
   4 #include <stdarg.h>
   5 #include <stdint.h>
   6 #include <stdio.h>
   7 #include <stdlib.h>
   8 #include <string.h>
   9
  10 #include "alloc-util.h"
  11 #include "escape.h"
  12 #include "fileio.h"
  13 #include "gunicode.h"
  14 #include "locale-util.h"
  15 #include "macro.h"
  16 #include "memory-util.h"
  17 #include "string-util.h"
  18 #include "terminal-util.h"
  19 #include "utf8.h"
  20 #include "util.h"
  21
  22 int strcmp_ptr(const char *a, const char *b) {
  23
  24         /* Like strcmp(), but tries to make sense of NULL pointers */
  25         if (a && b)
  26                 return strcmp(a, b);
  27
  28         if (!a && b)
  29                 return -1;
  30
  31         if (a && !b)
  32                 return 1;
  33
  34         return 0;
  35 }
  36
  37 char* endswith(const char *s, const char *postfix) {
  38         size_t sl, pl;
  39
  40         assert(s);
  41         assert(postfix);
  42
  43         sl = strlen(s);
  44         pl = strlen(postfix);
  45
  46         if (pl == 0)
  47                 return (char*) s + sl;
  48
  49         if (sl < pl)
  50                 return NULL;
  51
  52         if (memcmp(s + sl - pl, postfix, pl) != 0)
  53                 return NULL;
  54
  55         return (char*) s + sl - pl;
  56 }
  57
  58 char* endswith_no_case(const char *s, const char *postfix) {
  59         size_t sl, pl;
  60
  61         assert(s);
  62         assert(postfix);
  63
  64         sl = strlen(s);
  65         pl = strlen(postfix);
  66
  67         if (pl == 0)
  68                 return (char*) s + sl;
  69
  70         if (sl < pl)
  71                 return NULL;
  72
  73         if (strcasecmp(s + sl - pl, postfix) != 0)
  74                 return NULL;
  75
  76         return (char*) s + sl - pl;
  77 }
  78
  79 char* first_word(const char *s, const char *word) {
  80         size_t sl, wl;
  81         const char *p;
  82
  83         assert(s);
  84         assert(word);
  85
  86         /* Checks if the string starts with the specified word, either
  87          * followed by NUL or by whitespace. Returns a pointer to the
  88          * NUL or the first character after the whitespace. */
  89
  90         sl = strlen(s);
  91         wl = strlen(word);
  92
  93         if (sl < wl)
  94                 return NULL;
  95
  96         if (wl == 0)
  97                 return (char*) s;
  98
  99         if (memcmp(s, word, wl) != 0)
 100                 return NULL;
 101
 102         p = s + wl;
 103         if (*p == 0)
 104                 return (char*) p;
 105
 106         if (!strchr(WHITESPACE, *p))
 107                 return NULL;
 108
 109         p += strspn(p, WHITESPACE);
 110         return (char*) p;
 111 }
 112
 113 static size_t strcspn_escaped(const char *s, const char *reject) {
 114         bool escaped = false;
 115         int n;
 116
 117         for (n=0; s[n]; n++) {
 118                 if (escaped)
 119                         escaped = false;
 120                 else if (s[n] == '\\')
 121                         escaped = true;
 122                 else if (strchr(reject, s[n]))
 123                         break;
 124         }
 125
 126         /* if s ends in \, return index of previous char */
 127         return n - escaped;
 128 }
 129
 130 /* Split a string into words. */
 131 const char* split(const char **state, size_t *l, const char *separator, SplitFlags flags) {
 132         const char *current;
 133
 134         current = *state;
 135
 136         if (!*current) {
 137                 assert(**state == '\0');
 138                 return NULL;
 139         }
 140
 141         current += strspn(current, separator);
 142         if (!*current) {
 143                 *state = current;
 144                 return NULL;
 145         }
 146
 147         if (flags & SPLIT_QUOTES && strchr("\'\"", *current)) {
 148                 char quotechars[2] = {*current, '\0'};
 149
 150                 *l = strcspn_escaped(current + 1, quotechars);
 151                 if (current[*l + 1] == '\0' || current[*l + 1] != quotechars[0] ||
 152                     (current[*l + 2] && !strchr(separator, current[*l + 2]))) {
 153                         /* right quote missing or garbage at the end */
 154                         if (flags & SPLIT_RELAX) {
 155                                 *state = current + *l + 1 + (current[*l + 1] != '\0');
 156                                 return current + 1;
 157                         }
 158                         *state = current;
 159                         return NULL;
 160                 }
 161                 *state = current++ + *l + 2;
 162         } else if (flags & SPLIT_QUOTES) {
 163                 *l = strcspn_escaped(current, separator);
 164                 if (current[*l] && !strchr(separator, current[*l]) && !(flags & SPLIT_RELAX)) {
 165                         /* unfinished escape */
 166                         *state = current;
 167                         return NULL;
 168                 }
 169                 *state = current + *l;
 170         } else {
 171                 *l = strcspn(current, separator);
 172                 *state = current + *l;
 173         }
 174
 175         return current;
 176 }
 177
 178 char *strnappend(const char *s, const char *suffix, size_t b) {
 179         size_t a;
 180         char *r;
 181
 182         if (!s && !suffix)
 183                 return strdup("");
 184
 185         if (!s)
 186                 return strndup(suffix, b);
 187
 188         if (!suffix)
 189                 return strdup(s);
 190
 191         assert(s);
 192         assert(suffix);
 193
 194         a = strlen(s);
 195         if (b > ((size_t) -1) - a)
 196                 return NULL;
 197
 198         r = new(char, a+b+1);
 199         if (!r)
 200                 return NULL;
 201
 202         memcpy(r, s, a);
 203         memcpy(r+a, suffix, b);
 204         r[a+b] = 0;
 205
 206         return r;
 207 }
 208
 209 char *strappend(const char *s, const char *suffix) {
 210         return strnappend(s, suffix, strlen_ptr(suffix));
 211 }
 212
 213 char *strjoin_real(const char *x, ...) {
 214         va_list ap;
 215         size_t l;
 216         char *r, *p;
 217
 218         va_start(ap, x);
 219
 220         if (x) {
 221                 l = strlen(x);
 222
 223                 for (;;) {
 224                         const char *t;
 225                         size_t n;
 226
 227                         t = va_arg(ap, const char *);
 228                         if (!t)
 229                                 break;
 230
 231                         n = strlen(t);
 232                         if (n > ((size_t) -1) - l) {
 233                                 va_end(ap);
 234                                 return NULL;
 235                         }
 236
 237                         l += n;
 238                 }
 239         } else
 240                 l = 0;
 241
 242         va_end(ap);
 243
 244         r = new(char, l+1);
 245         if (!r)
 246                 return NULL;
 247
 248         if (x) {
 249                 p = stpcpy(r, x);
 250
 251                 va_start(ap, x);
 252
 253                 for (;;) {
 254                         const char *t;
 255
 256                         t = va_arg(ap, const char *);
 257                         if (!t)
 258                                 break;
 259
 260                         p = stpcpy(p, t);
 261                 }
 262
 263                 va_end(ap);
 264         } else
 265                 r[0] = 0;
 266
 267         return r;
 268 }
 269
 270 char *strstrip(char *s) {
 271         if (!s)
 272                 return NULL;
 273
 274         /* Drops trailing whitespace. Modifies the string in place. Returns pointer to first non-space character */
 275
 276         return delete_trailing_chars(skip_leading_chars(s, WHITESPACE), WHITESPACE);
 277 }
 278
 279 char *delete_chars(char *s, const char *bad) {
 280         char *f, *t;
 281
 282         /* Drops all specified bad characters, regardless where in the string */
 283
 284         if (!s)
 285                 return NULL;
 286
 287         if (!bad)
 288                 bad = WHITESPACE;
 289
 290         for (f = s, t = s; *f; f++) {
 291                 if (strchr(bad, *f))
 292                         continue;
 293
 294                 *(t++) = *f;
 295         }
 296
 297         *t = 0;
 298
 299         return s;
 300 }
 301
 302 char *delete_trailing_chars(char *s, const char *bad) {
 303         char *p, *c = s;
 304
 305         /* Drops all specified bad characters, at the end of the string */
 306
 307         if (!s)
 308                 return NULL;
 309
 310         if (!bad)
 311                 bad = WHITESPACE;
 312
 313         for (p = s; *p; p++)
 314                 if (!strchr(bad, *p))
 315                         c = p + 1;
 316
 317         *c = 0;
 318
 319         return s;
 320 }
 321
 322 char *truncate_nl(char *s) {
 323         assert(s);
 324
 325         s[strcspn(s, NEWLINE)] = 0;
 326         return s;
 327 }
 328
 329 char ascii_tolower(char x) {
 330
 331         if (x >= 'A' && x <= 'Z')
 332                 return x - 'A' + 'a';
 333
 334         return x;
 335 }
 336
 337 char ascii_toupper(char x) {
 338
 339         if (x >= 'a' && x <= 'z')
 340                 return x - 'a' + 'A';
 341
 342         return x;
 343 }
 344
 345 char *ascii_strlower(char *t) {
 346         char *p;
 347
 348         assert(t);
 349
 350         for (p = t; *p; p++)
 351                 *p = ascii_tolower(*p);
 352
 353         return t;
 354 }
 355
 356 char *ascii_strupper(char *t) {
 357         char *p;
 358
 359         assert(t);
 360
 361         for (p = t; *p; p++)
 362                 *p = ascii_toupper(*p);
 363
 364         return t;
 365 }
 366
 367 char *ascii_strlower_n(char *t, size_t n) {
 368         size_t i;
 369
 370         if (n <= 0)
 371                 return t;
 372
 373         for (i = 0; i < n; i++)
 374                 t[i] = ascii_tolower(t[i]);
 375
 376         return t;
 377 }
 378
 379 int ascii_strcasecmp_n(const char *a, const char *b, size_t n) {
 380
 381         for (; n > 0; a++, b++, n--) {
 382                 int x, y;
 383
 384                 x = (int) (uint8_t) ascii_tolower(*a);
 385                 y = (int) (uint8_t) ascii_tolower(*b);
 386
 387                 if (x != y)
 388                         return x - y;
 389         }
 390
 391         return 0;
 392 }
 393
 394 int ascii_strcasecmp_nn(const char *a, size_t n, const char *b, size_t m) {
 395         int r;
 396
 397         r = ascii_strcasecmp_n(a, b, MIN(n, m));
 398         if (r != 0)
 399                 return r;
 400
 401         return CMP(n, m);
 402 }
 403
 404 bool chars_intersect(const char *a, const char *b) {
 405         const char *p;
 406
 407         /* Returns true if any of the chars in a are in b. */
 408         for (p = a; *p; p++)
 409                 if (strchr(b, *p))
 410                         return true;
 411
 412         return false;
 413 }
 414
 415 bool string_has_cc(const char *p, const char *ok) {
 416         const char *t;
 417
 418         assert(p);
 419
 420         /*
 421          * Check if a string contains control characters. If 'ok' is
 422          * non-NULL it may be a string containing additional CCs to be
 423          * considered OK.
 424          */
 425
 426         for (t = p; *t; t++) {
 427                 if (ok && strchr(ok, *t))
 428                         continue;
 429
 430                 if (*t > 0 && *t < ' ')
 431                         return true;
 432
 433                 if (*t == 127)
 434                         return true;
 435         }
 436
 437         return false;
 438 }
 439
 440 static int write_ellipsis(char *buf, bool unicode) {
 441         if (unicode || is_locale_utf8()) {
 442                 buf[0] = 0xe2; /* tri-dot ellipsis: … */
 443                 buf[1] = 0x80;
 444                 buf[2] = 0xa6;
 445         } else {
 446                 buf[0] = '.';
 447                 buf[1] = '.';
 448                 buf[2] = '.';
 449         }
 450
 451         return 3;
 452 }
 453
 454 static char *ascii_ellipsize_mem(const char *s, size_t old_length, size_t new_length, unsigned percent) {
 455         size_t x, need_space, suffix_len;
 456         char *t;
 457
 458         assert(s);
 459         assert(percent <= 100);
 460         assert(new_length != (size_t) -1);
 461
 462         if (old_length <= new_length)
 463                 return strndup(s, old_length);
 464
 465         /* Special case short ellipsations */
 466         switch (new_length) {
 467
 468         case 0:
 469                 return strdup("");
 470
 471         case 1:
 472                 if (is_locale_utf8())
 473                         return strdup("…");
 474                 else
 475                         return strdup(".");
 476
 477         case 2:
 478                 if (!is_locale_utf8())
 479                         return strdup("..");
 480
 481                 break;
 482
 483         default:
 484                 break;
 485         }
 486
 487         /* Calculate how much space the ellipsis will take up. If we are in UTF-8 mode we only need space for one
 488          * character ("…"), otherwise for three characters ("..."). Note that in both cases we need 3 bytes of storage,
 489          * either for the UTF-8 encoded character or for three ASCII characters. */
 490         need_space = is_locale_utf8() ? 1 : 3;
 491
 492         t = new(char, new_length+3);
 493         if (!t)
 494                 return NULL;
 495
 496         assert(new_length >= need_space);
 497
 498         x = ((new_length - need_space) * percent + 50) / 100;
 499         assert(x <= new_length - need_space);
 500
 501         memcpy(t, s, x);
 502         write_ellipsis(t + x, false);
 503         suffix_len = new_length - x - need_space;
 504         memcpy(t + x + 3, s + old_length - suffix_len, suffix_len);
 505         *(t + x + 3 + suffix_len) = '\0';
 506
 507         return t;
 508 }
 509
 510 char *ellipsize_mem(const char *s, size_t old_length, size_t new_length, unsigned percent) {
 511         size_t x, k, len, len2;
 512         const char *i, *j;
 513         char *e;
 514         int r;
 515
 516         /* Note that 'old_length' refers to bytes in the string, while 'new_length' refers to character cells taken up
 517          * on screen. This distinction doesn't matter for ASCII strings, but it does matter for non-ASCII UTF-8
 518          * strings.
 519          *
 520          * Ellipsation is done in a locale-dependent way:
 521          * 1. If the string passed in is fully ASCII and the current locale is not UTF-8, three dots are used ("...")
 522          * 2. Otherwise, a unicode ellipsis is used ("…")
 523          *
 524          * In other words: you'll get a unicode ellipsis as soon as either the string contains non-ASCII characters or
 525          * the current locale is UTF-8.
 526          */
 527
 528         assert(s);
 529         assert(percent <= 100);
 530
 531         if (new_length == (size_t) -1)
 532                 return strndup(s, old_length);
 533
 534         if (new_length == 0)
 535                 return strdup("");
 536
 537         /* If no multibyte characters use ascii_ellipsize_mem for speed */
 538         if (ascii_is_valid_n(s, old_length))
 539                 return ascii_ellipsize_mem(s, old_length, new_length, percent);
 540
 541         x = ((new_length - 1) * percent) / 100;
 542         assert(x <= new_length - 1);
 543
 544         k = 0;
 545         for (i = s; i < s + old_length; i = utf8_next_char(i)) {
 546                 char32_t c;
 547                 int w;
 548
 549                 r = utf8_encoded_to_unichar(i, &c);
 550                 if (r < 0)
 551                         return NULL;
 552
 553                 w = unichar_iswide(c) ? 2 : 1;
 554                 if (k + w <= x)
 555                         k += w;
 556                 else
 557                         break;
 558         }
 559
 560         for (j = s + old_length; j > i; ) {
 561                 char32_t c;
 562                 int w;
 563                 const char *jj;
 564
 565                 jj = utf8_prev_char(j);
 566                 r = utf8_encoded_to_unichar(jj, &c);
 567                 if (r < 0)
 568                         return NULL;
 569
 570                 w = unichar_iswide(c) ? 2 : 1;
 571                 if (k + w <= new_length) {
 572                         k += w;
 573                         j = jj;
 574                 } else
 575                         break;
 576         }
 577         assert(i <= j);
 578
 579         /* we don't actually need to ellipsize */
 580         if (i == j)
 581                 return memdup_suffix0(s, old_length);
 582
 583         /* make space for ellipsis, if possible */
 584         if (j < s + old_length)
 585                 j = utf8_next_char(j);
 586         else if (i > s)
 587                 i = utf8_prev_char(i);
 588
 589         len = i - s;
 590         len2 = s + old_length - j;
 591         e = new(char, len + 3 + len2 + 1);
 592         if (!e)
 593                 return NULL;
 594
 595         /*
 596         printf("old_length=%zu new_length=%zu x=%zu len=%u len2=%u k=%u\n",
 597                old_length, new_length, x, len, len2, k);
 598         */
 599
 600         memcpy(e, s, len);
 601         write_ellipsis(e + len, true);
 602         memcpy(e + len + 3, j, len2);
 603         *(e + len + 3 + len2) = '\0';
 604
 605         return e;
 606 }
 607
 608 char *cellescape(char *buf, size_t len, const char *s) {
 609         /* Escape and ellipsize s into buffer buf of size len. Only non-control ASCII
 610          * characters are copied as they are, everything else is escaped. The result
 611          * is different then if escaping and ellipsization was performed in two
 612          * separate steps, because each sequence is either stored in full or skipped.
 613          *
 614          * This function should be used for logging about strings which expected to
 615          * be plain ASCII in a safe way.
 616          *
 617          * An ellipsis will be used if s is too long. It was always placed at the
 618          * very end.
 619          */
 620
 621         size_t i = 0, last_char_width[4] = {}, k = 0, j;
 622
 623         assert(len > 0); /* at least a terminating NUL */
 624
 625         for (;;) {
 626                 char four[4];
 627                 int w;
 628
 629                 if (*s == 0) /* terminating NUL detected? then we are done! */
 630                         goto done;
 631
 632                 w = cescape_char(*s, four);
 633                 if (i + w + 1 > len) /* This character doesn't fit into the buffer anymore? In that case let's
 634                                       * ellipsize at the previous location */
 635                         break;
 636
 637                 /* OK, there was space, let's add this escaped character to the buffer */
 638                 memcpy(buf + i, four, w);
 639                 i += w;
 640
 641                 /* And remember its width in the ring buffer */
 642                 last_char_width[k] = w;
 643                 k = (k + 1) % 4;
 644
 645                 s++;
 646         }
 647
 648         /* Ellipsation is necessary. This means we might need to truncate the string again to make space for 4
 649          * characters ideally, but the buffer is shorter than that in the first place take what we can get */
 650         for (j = 0; j < ELEMENTSOF(last_char_width); j++) {
 651
 652                 if (i + 4 <= len) /* nice, we reached our space goal */
 653                         break;
 654
 655                 k = k == 0 ? 3 : k - 1;
 656                 if (last_char_width[k] == 0) /* bummer, we reached the beginning of the strings */
 657                         break;
 658
 659                 assert(i >= last_char_width[k]);
 660                 i -= last_char_width[k];
 661         }
 662
 663         if (i + 4 <= len) /* yay, enough space */
 664                 i += write_ellipsis(buf + i, false);
 665         else if (i + 3 <= len) { /* only space for ".." */
 666                 buf[i++] = '.';
 667                 buf[i++] = '.';
 668         } else if (i + 2 <= len) /* only space for a single "." */
 669                 buf[i++] = '.';
 670         else
 671                 assert(i + 1 <= len);
 672
 673  done:
 674         buf[i] = '\0';
 675         return buf;
 676 }
 677
 678 char* strshorten(char *s, size_t l) {
 679         assert(s);
 680
 681         if (strnlen(s, l+1) > l)
 682                 s[l] = 0;
 683
 684         return s;
 685 }
 686
 687 char *strreplace(const char *text, const char *old_string, const char *new_string) {
 688         size_t l, old_len, new_len, allocated = 0;
 689         char *t, *ret = NULL;
 690         const char *f;
 691
 692         assert(old_string);
 693         assert(new_string);
 694
 695         if (!text)
 696                 return NULL;
 697
 698         old_len = strlen(old_string);
 699         new_len = strlen(new_string);
 700
 701         l = strlen(text);
 702         if (!GREEDY_REALLOC(ret, allocated, l+1))
 703                 return NULL;
 704
 705         f = text;
 706         t = ret;
 707         while (*f) {
 708                 size_t d, nl;
 709
 710                 if (!startswith(f, old_string)) {
 711                         *(t++) = *(f++);
 712                         continue;
 713                 }
 714
 715                 d = t - ret;
 716                 nl = l - old_len + new_len;
 717
 718                 if (!GREEDY_REALLOC(ret, allocated, nl + 1))
 719                         return mfree(ret);
 720
 721                 l = nl;
 722                 t = ret + d;
 723
 724                 t = stpcpy(t, new_string);
 725                 f += old_len;
 726         }
 727
 728         *t = 0;
 729         return ret;
 730 }
 731
 732 static void advance_offsets(ssize_t diff, size_t offsets[static 2], size_t shift[static 2], size_t size) {
 733         if (!offsets)
 734                 return;
 735
 736         if ((size_t) diff < offsets[0])
 737                 shift[0] += size;
 738         if ((size_t) diff < offsets[1])
 739                 shift[1] += size;
 740 }
 741
 742 char *strip_tab_ansi(char **ibuf, size_t *_isz, size_t highlight[2]) {
 743         const char *i, *begin = NULL;
 744         enum {
 745                 STATE_OTHER,
 746                 STATE_ESCAPE,
 747                 STATE_CSI,
 748                 STATE_CSO,
 749         } state = STATE_OTHER;
 750         char *obuf = NULL;
 751         size_t osz = 0, isz, shift[2] = {};
 752         FILE *f;
 753
 754         assert(ibuf);
 755         assert(*ibuf);
 756
 757         /* This does three things:
 758          *
 759          * 1. Replaces TABs by 8 spaces
 760          * 2. Strips ANSI color sequences (a subset of CSI), i.e. ESC '[' … 'm' sequences
 761          * 3. Strips ANSI operating system sequences (CSO), i.e. ESC ']' … BEL sequences
 762          *
 763          * Everything else will be left as it is. In particular other ANSI sequences are left as they are, as
 764          * are any other special characters. Truncated ANSI sequences are left-as is too. This call is
 765          * supposed to suppress the most basic formatting noise, but nothing else.
 766          *
 767          * Why care for CSO sequences? Well, to undo what terminal_urlify() and friends generate. */
 768
 769         isz = _isz ? *_isz : strlen(*ibuf);
 770
 771         /* Note we turn off internal locking on f for performance reasons. It's safe to do so since we
 772          * created f here and it doesn't leave our scope. */
 773         f = open_memstream_unlocked(&obuf, &osz);
 774         if (!f)
 775                 return NULL;
 776
 777         for (i = *ibuf; i < *ibuf + isz + 1; i++) {
 778
 779                 switch (state) {
 780
 781                 case STATE_OTHER:
 782                         if (i >= *ibuf + isz) /* EOT */
 783                                 break;
 784                         else if (*i == '\x1B')
 785                                 state = STATE_ESCAPE;
 786                         else if (*i == '\t') {
 787                                 fputs("        ", f);
 788                                 advance_offsets(i - *ibuf, highlight, shift, 7);
 789                         } else
 790                                 fputc(*i, f);
 791
 792                         break;
 793
 794                 case STATE_ESCAPE:
 795                         if (i >= *ibuf + isz) { /* EOT */
 796                                 fputc('\x1B', f);
 797                                 advance_offsets(i - *ibuf, highlight, shift, 1);
 798                                 break;
 799                         } else if (*i == '[') { /* ANSI CSI */
 800                                 state = STATE_CSI;
 801                                 begin = i + 1;
 802                         } else if (*i == ']') { /* ANSI CSO */
 803                                 state = STATE_CSO;
 804                                 begin = i + 1;
 805                         } else {
 806                                 fputc('\x1B', f);
 807                                 fputc(*i, f);
 808                                 advance_offsets(i - *ibuf, highlight, shift, 1);
 809                                 state = STATE_OTHER;
 810                         }
 811
 812                         break;
 813
 814                 case STATE_CSI:
 815
 816                         if (i >= *ibuf + isz || /* EOT … */
 817                             !strchr("01234567890;m", *i)) { /* … or invalid chars in sequence */
 818                                 fputc('\x1B', f);
 819                                 fputc('[', f);
 820                                 advance_offsets(i - *ibuf, highlight, shift, 2);
 821                                 state = STATE_OTHER;
 822                                 i = begin-1;
 823                         } else if (*i == 'm')
 824                                 state = STATE_OTHER;
 825
 826                         break;
 827
 828                 case STATE_CSO:
 829
 830                         if (i >= *ibuf + isz || /* EOT … */
 831                             (*i != '\a' && (uint8_t) *i < 32U) || (uint8_t) *i > 126U) { /* … or invalid chars in sequence */
 832                                 fputc('\x1B', f);
 833                                 fputc(']', f);
 834                                 advance_offsets(i - *ibuf, highlight, shift, 2);
 835                                 state = STATE_OTHER;
 836                                 i = begin-1;
 837                         } else if (*i == '\a')
 838                                 state = STATE_OTHER;
 839
 840                         break;
 841                 }
 842         }
 843
 844         if (fflush_and_check(f) < 0) {
 845                 fclose(f);
 846                 return mfree(obuf);
 847         }
 848
 849         fclose(f);
 850
 851         free(*ibuf);
 852         *ibuf = obuf;
 853
 854         if (_isz)
 855                 *_isz = osz;
 856
 857         if (highlight) {
 858                 highlight[0] += shift[0];
 859                 highlight[1] += shift[1];
 860         }
 861
 862         return obuf;
 863 }
 864
 865 char *strextend_with_separator(char **x, const char *separator, ...) {
 866         bool need_separator;
 867         size_t f, l, l_separator;
 868         char *r, *p;
 869         va_list ap;
 870
 871         assert(x);
 872
 873         l = f = strlen_ptr(*x);
 874
 875         need_separator = !isempty(*x);
 876         l_separator = strlen_ptr(separator);
 877
 878         va_start(ap, separator);
 879         for (;;) {
 880                 const char *t;
 881                 size_t n;
 882
 883                 t = va_arg(ap, const char *);
 884                 if (!t)
 885                         break;
 886
 887                 n = strlen(t);
 888
 889                 if (need_separator)
 890                         n += l_separator;
 891
 892                 if (n > ((size_t) -1) - l) {
 893                         va_end(ap);
 894                         return NULL;
 895                 }
 896
 897                 l += n;
 898                 need_separator = true;
 899         }
 900         va_end(ap);
 901
 902         need_separator = !isempty(*x);
 903
 904         r = realloc(*x, l+1);
 905         if (!r)
 906                 return NULL;
 907
 908         p = r + f;
 909
 910         va_start(ap, separator);
 911         for (;;) {
 912                 const char *t;
 913
 914                 t = va_arg(ap, const char *);
 915                 if (!t)
 916                         break;
 917
 918                 if (need_separator && separator)
 919                         p = stpcpy(p, separator);
 920
 921                 p = stpcpy(p, t);
 922
 923                 need_separator = true;
 924         }
 925         va_end(ap);
 926
 927         assert(p == r + l);
 928
 929         *p = 0;
 930         *x = r;
 931
 932         return r + l;
 933 }
 934
 935 char *strrep(const char *s, unsigned n) {
 936         size_t l;
 937         char *r, *p;
 938         unsigned i;
 939
 940         assert(s);
 941
 942         l = strlen(s);
 943         p = r = malloc(l * n + 1);
 944         if (!r)
 945                 return NULL;
 946
 947         for (i = 0; i < n; i++)
 948                 p = stpcpy(p, s);
 949
 950         *p = 0;
 951         return r;
 952 }
 953
 954 int split_pair(const char *s, const char *sep, char **l, char **r) {
 955         char *x, *a, *b;
 956
 957         assert(s);
 958         assert(sep);
 959         assert(l);
 960         assert(r);
 961
 962         if (isempty(sep))
 963                 return -EINVAL;
 964
 965         x = strstr(s, sep);
 966         if (!x)
 967                 return -EINVAL;
 968
 969         a = strndup(s, x - s);
 970         if (!a)
 971                 return -ENOMEM;
 972
 973         b = strdup(x + strlen(sep));
 974         if (!b) {
 975                 free(a);
 976                 return -ENOMEM;
 977         }
 978
 979         *l = a;
 980         *r = b;
 981
 982         return 0;
 983 }
 984
 985 int free_and_strdup(char **p, const char *s) {
 986         char *t;
 987
 988         assert(p);
 989
 990         /* Replaces a string pointer with a strdup()ed new string,
 991          * possibly freeing the old one. */
 992
 993         if (streq_ptr(*p, s))
 994                 return 0;
 995
 996         if (s) {
 997                 t = strdup(s);
 998                 if (!t)
 999                         return -ENOMEM;
1000         } else
1001                 t = NULL;
1002
1003         free(*p);
1004         *p = t;
1005
1006         return 1;
1007 }
1008
1009 int free_and_strndup(char **p, const char *s, size_t l) {
1010         char *t;
1011
1012         assert(p);
1013         assert(s || l == 0);
1014
1015         /* Replaces a string pointer with a strndup()ed new string,
1016          * freeing the old one. */
1017
1018         if (!*p && !s)
1019                 return 0;
1020
1021         if (*p && s && strneq(*p, s, l) && (l > strlen(*p) || (*p)[l] == '\0'))
1022                 return 0;
1023
1024         if (s) {
1025                 t = strndup(s, l);
1026                 if (!t)
1027                         return -ENOMEM;
1028         } else
1029                 t = NULL;
1030
1031         free_and_replace(*p, t);
1032         return 1;
1033 }
1034
1035 char* string_erase(char *x) {
1036         if (!x)
1037                 return NULL;
1038
1039         /* A delicious drop of snake-oil! To be called on memory where
1040          * we stored passphrases or so, after we used them. */
1041         explicit_bzero_safe(x, strlen(x));
1042         return x;
1043 }
1044
1045 char *string_free_erase(char *s) {
1046         return mfree(string_erase(s));
1047 }
1048
1049 bool string_is_safe(const char *p) {
1050         const char *t;
1051
1052         if (!p)
1053                 return false;
1054
1055         for (t = p; *t; t++) {
1056                 if (*t > 0 && *t < ' ') /* no control characters */
1057                         return false;
1058
1059                 if (strchr(QUOTES "\\\x7f", *t))
1060                         return false;
1061         }
1062
1063         return true;
1064 }