src/basic/utf8.c

   1 /* SPDX-License-Identifier: LGPL-2.0-or-later */
   2
   3 /* Parts of this file are based on the GLIB utf8 validation functions. The original copyright follows.
   4  *
   5  * gutf8.c - Operations on UTF-8 strings.
   6  * Copyright (C) 1999 Tom Tromey
   7  * Copyright (C) 2000 Red Hat, Inc.
   8  */
   9
  10 #include "alloc-util.h"
  11 #include "gunicode.h"
  12 #include "hexdecoct.h"
  13 #include "string-util.h"
  14 #include "utf8.h"
  15
  16 bool unichar_is_valid(char32_t ch) {
  17
  18         if (ch >= 0x110000) /* End of unicode space */
  19                 return false;
  20         if ((ch & 0xFFFFF800) == 0xD800) /* Reserved area for UTF-16 */
  21                 return false;
  22         if ((ch >= 0xFDD0) && (ch <= 0xFDEF)) /* Reserved */
  23                 return false;
  24         if ((ch & 0xFFFE) == 0xFFFE) /* BOM (Byte Order Mark) */
  25                 return false;
  26
  27         return true;
  28 }
  29
  30 static bool unichar_is_control(char32_t ch) {
  31
  32         /*
  33           0 to ' '-1 is the C0 range.
  34           DEL=0x7F, and DEL+1 to 0x9F is C1 range.
  35           '\t' is in C0 range, but more or less harmless and commonly used.
  36         */
  37
  38         return (ch < ' ' && !IN_SET(ch, '\t', '\n')) ||
  39                 (0x7F <= ch && ch <= 0x9F);
  40 }
  41
  42 /* count of characters used to encode one unicode char */
  43 static size_t utf8_encoded_expected_len(uint8_t c) {
  44         if (c < 0x80)
  45                 return 1;
  46         if ((c & 0xe0) == 0xc0)
  47                 return 2;
  48         if ((c & 0xf0) == 0xe0)
  49                 return 3;
  50         if ((c & 0xf8) == 0xf0)
  51                 return 4;
  52         if ((c & 0xfc) == 0xf8)
  53                 return 5;
  54         if ((c & 0xfe) == 0xfc)
  55                 return 6;
  56
  57         return 0;
  58 }
  59
  60 /* decode one unicode char */
  61 int utf8_encoded_to_unichar(const char *str, char32_t *ret_unichar) {
  62         char32_t unichar;
  63         size_t len;
  64
  65         assert(str);
  66         assert(ret_unichar);
  67
  68         len = utf8_encoded_expected_len(str[0]);
  69
  70         switch (len) {
  71         case 1:
  72                 *ret_unichar = (char32_t)str[0];
  73                 return 1;
  74         case 2:
  75                 unichar = str[0] & 0x1f;
  76                 break;
  77         case 3:
  78                 unichar = (char32_t)str[0] & 0x0f;
  79                 break;
  80         case 4:
  81                 unichar = (char32_t)str[0] & 0x07;
  82                 break;
  83         case 5:
  84                 unichar = (char32_t)str[0] & 0x03;
  85                 break;
  86         case 6:
  87                 unichar = (char32_t)str[0] & 0x01;
  88                 break;
  89         default:
  90                 return -EINVAL;
  91         }
  92
  93         for (size_t i = 1; i < len; i++) {
  94                 if (((char32_t)str[i] & 0xc0) != 0x80)
  95                         return -EINVAL;
  96
  97                 unichar <<= 6;
  98                 unichar |= (char32_t)str[i] & 0x3f;
  99         }
 100
 101         *ret_unichar = unichar;
 102         return len;
 103 }
 104
 105 bool utf8_is_printable_newline(const char* str, size_t length, bool allow_newline) {
 106         assert(str);
 107
 108         for (const char *p = str; length > 0;) {
 109                 int encoded_len;
 110                 char32_t val;
 111
 112                 encoded_len = utf8_encoded_valid_unichar(p, length);
 113                 if (encoded_len < 0)
 114                         return false;
 115                 assert(encoded_len > 0 && (size_t) encoded_len <= length);
 116
 117                 if (utf8_encoded_to_unichar(p, &val) < 0 ||
 118                     unichar_is_control(val) ||
 119                     (!allow_newline && val == '\n'))
 120                         return false;
 121
 122                 length -= encoded_len;
 123                 p += encoded_len;
 124         }
 125
 126         return true;
 127 }
 128
 129 char* utf8_is_valid_n(const char *str, size_t len_bytes) {
 130         /* Check if the string is composed of valid utf8 characters. If length len_bytes is given, stop after
 131          * len_bytes. Otherwise, stop at NUL. */
 132
 133         assert(str);
 134
 135         for (size_t i = 0; len_bytes != SIZE_MAX ? i < len_bytes : str[i] != '\0'; ) {
 136                 int len;
 137
 138                 if (_unlikely_(str[i] == '\0'))
 139                         return NULL; /* embedded NUL */
 140
 141                 len = utf8_encoded_valid_unichar(str + i,
 142                                                  len_bytes != SIZE_MAX ? len_bytes - i : SIZE_MAX);
 143                 if (_unlikely_(len < 0))
 144                         return NULL; /* invalid character */
 145
 146                 i += len;
 147         }
 148
 149         return (char*) str;
 150 }
 151
 152 char* utf8_escape_invalid(const char *str) {
 153         char *p, *s;
 154
 155         assert(str);
 156
 157         p = s = malloc(strlen(str) * 4 + 1);
 158         if (!p)
 159                 return NULL;
 160
 161         while (*str) {
 162                 int len;
 163
 164                 len = utf8_encoded_valid_unichar(str, SIZE_MAX);
 165                 if (len > 0) {
 166                         s = mempcpy(s, str, len);
 167                         str += len;
 168                 } else {
 169                         s = stpcpy(s, UTF8_REPLACEMENT_CHARACTER);
 170                         str += 1;
 171                 }
 172         }
 173
 174         *s = '\0';
 175         return str_realloc(p);
 176 }
 177
 178 int utf8_char_console_width(const char *str) {
 179         char32_t c;
 180         int r;
 181
 182         r = utf8_encoded_to_unichar(str, &c);
 183         if (r < 0)
 184                 return r;
 185
 186         if (c == '\t')
 187                 return 8; /* Assume a tab width of 8 */
 188
 189         /* TODO: we should detect combining characters */
 190
 191         return unichar_iswide(c) ? 2 : 1;
 192 }
 193
 194 char* utf8_escape_non_printable_full(const char *str, size_t console_width, bool force_ellipsis) {
 195         char *p, *s, *prev_s;
 196         size_t n = 0; /* estimated print width */
 197
 198         assert(str);
 199
 200         if (console_width == 0)
 201                 return strdup("");
 202
 203         p = s = prev_s = malloc(strlen(str) * 4 + 1);
 204         if (!p)
 205                 return NULL;
 206
 207         for (;;) {
 208                 int len;
 209                 char *saved_s = s;
 210
 211                 if (!*str) { /* done! */
 212                         if (force_ellipsis)
 213                                 goto truncation;
 214                         else
 215                                 goto finish;
 216                 }
 217
 218                 len = utf8_encoded_valid_unichar(str, SIZE_MAX);
 219                 if (len > 0) {
 220                         if (utf8_is_printable(str, len)) {
 221                                 int w;
 222
 223                                 w = utf8_char_console_width(str);
 224                                 assert(w >= 0);
 225                                 if (n + w > console_width)
 226                                         goto truncation;
 227
 228                                 s = mempcpy(s, str, len);
 229                                 str += len;
 230                                 n += w;
 231
 232                         } else {
 233                                 for (; len > 0; len--) {
 234                                         if (n + 4 > console_width)
 235                                                 goto truncation;
 236
 237                                         *(s++) = '\\';
 238                                         *(s++) = 'x';
 239                                         *(s++) = hexchar((int) *str >> 4);
 240                                         *(s++) = hexchar((int) *str);
 241
 242                                         str += 1;
 243                                         n += 4;
 244                                 }
 245                         }
 246                 } else {
 247                         if (n + 1 > console_width)
 248                                 goto truncation;
 249
 250                         s = mempcpy(s, UTF8_REPLACEMENT_CHARACTER, strlen(UTF8_REPLACEMENT_CHARACTER));
 251                         str += 1;
 252                         n += 1;
 253                 }
 254
 255                 prev_s = saved_s;
 256         }
 257
 258  truncation:
 259         /* Try to go back one if we don't have enough space for the ellipsis */
 260         if (n + 1 > console_width)
 261                 s = prev_s;
 262
 263         s = mempcpy(s, "…", strlen("…"));
 264
 265  finish:
 266         *s = '\0';
 267         return str_realloc(p);
 268 }
 269
 270 char* ascii_is_valid_n(const char *str, size_t len) {
 271         /* Check whether the string consists of valid ASCII bytes, i.e values between 1 and 127, inclusive.
 272          * Stops at len, or NUL byte if len is SIZE_MAX. */
 273
 274         assert(str);
 275
 276         for (size_t i = 0; len != SIZE_MAX ? i < len : str[i] != '\0'; i++)
 277                 if ((unsigned char) str[i] >= 128 || str[i] == '\0')
 278                         return NULL;
 279
 280         return (char*) str;
 281 }
 282
 283 int utf8_to_ascii(const char *str, char replacement_char, char **ret) {
 284         /* Convert to a string that has only ASCII chars, replacing anything that is not ASCII
 285          * by replacement_char. */
 286
 287         assert(str);
 288         assert(ret);
 289
 290         _cleanup_free_ char *ans = new(char, strlen(str) + 1);
 291         if (!ans)
 292                 return -ENOMEM;
 293
 294         char *q = ans;
 295
 296         for (const char *p = str; *p; q++) {
 297                 int l;
 298
 299                 l = utf8_encoded_valid_unichar(p, SIZE_MAX);
 300                 if (l < 0)  /* Non-UTF-8, let's not even try to propagate the garbage */
 301                         return l;
 302
 303                 if (l == 1)
 304                         *q = *p;
 305                 else
 306                         /* non-ASCII, we need to replace it */
 307                         *q = replacement_char;
 308
 309                 p += l;
 310         }
 311         *q = '\0';
 312
 313         *ret = TAKE_PTR(ans);
 314         return 0;
 315 }
 316
 317 /**
 318  * utf8_encode_unichar() - Encode single UCS-4 character as UTF-8
 319  * @out_utf8: output buffer of at least 4 bytes or NULL
 320  * @g: UCS-4 character to encode
 321  *
 322  * This encodes a single UCS-4 character as UTF-8 and writes it into @out_utf8.
 323  * The length of the character is returned. It is not zero-terminated! If the
 324  * output buffer is NULL, only the length is returned.
 325  *
 326  * Returns: The length in bytes that the UTF-8 representation does or would
 327  *          occupy.
 328  */
 329 size_t utf8_encode_unichar(char *out_utf8, char32_t g) {
 330
 331         if (g < (1 << 7)) {
 332                 if (out_utf8)
 333                         out_utf8[0] = g & 0x7f;
 334                 return 1;
 335         } else if (g < (1 << 11)) {
 336                 if (out_utf8) {
 337                         out_utf8[0] = 0xc0 | ((g >> 6) & 0x1f);
 338                         out_utf8[1] = 0x80 | (g & 0x3f);
 339                 }
 340                 return 2;
 341         } else if (g < (1 << 16)) {
 342                 if (out_utf8) {
 343                         out_utf8[0] = 0xe0 | ((g >> 12) & 0x0f);
 344                         out_utf8[1] = 0x80 | ((g >> 6) & 0x3f);
 345                         out_utf8[2] = 0x80 | (g & 0x3f);
 346                 }
 347                 return 3;
 348         } else if (g < (1 << 21)) {
 349                 if (out_utf8) {
 350                         out_utf8[0] = 0xf0 | ((g >> 18) & 0x07);
 351                         out_utf8[1] = 0x80 | ((g >> 12) & 0x3f);
 352                         out_utf8[2] = 0x80 | ((g >> 6) & 0x3f);
 353                         out_utf8[3] = 0x80 | (g & 0x3f);
 354                 }
 355                 return 4;
 356         }
 357
 358         return 0;
 359 }
 360
 361 char* utf16_to_utf8(const char16_t *s, size_t length /* bytes! */) {
 362         const uint8_t *f;
 363         char *r, *t;
 364
 365         if (length == 0)
 366                 return new0(char, 1);
 367
 368         assert(s);
 369
 370         if (length == SIZE_MAX) {
 371                 length = char16_strlen(s);
 372
 373                 if (length > SIZE_MAX/2)
 374                         return NULL; /* overflow */
 375
 376                 length *= 2;
 377         }
 378
 379         /* Input length is in bytes, i.e. the shortest possible character takes 2 bytes. Each unicode character may
 380          * take up to 4 bytes in UTF-8. Let's also account for a trailing NUL byte. */
 381         if (length > (SIZE_MAX - 1) / 2)
 382                 return NULL; /* overflow */
 383
 384         r = new(char, length * 2 + 1);
 385         if (!r)
 386                 return NULL;
 387
 388         f = (const uint8_t*) s;
 389         t = r;
 390
 391         while (f + 1 < (const uint8_t*) s + length) {
 392                 char16_t w1, w2;
 393
 394                 /* see RFC 2781 section 2.2 */
 395
 396                 w1 = f[1] << 8 | f[0];
 397                 f += 2;
 398
 399                 if (!utf16_is_surrogate(w1)) {
 400                         t += utf8_encode_unichar(t, w1);
 401                         continue;
 402                 }
 403
 404                 if (utf16_is_trailing_surrogate(w1))
 405                         continue; /* spurious trailing surrogate, ignore */
 406
 407                 if (f + 1 >= (const uint8_t*) s + length)
 408                         break;
 409
 410                 w2 = f[1] << 8 | f[0];
 411                 f += 2;
 412
 413                 if (!utf16_is_trailing_surrogate(w2)) {
 414                         f -= 2;
 415                         continue; /* surrogate missing its trailing surrogate, ignore */
 416                 }
 417
 418                 t += utf8_encode_unichar(t, utf16_surrogate_pair_to_unichar(w1, w2));
 419         }
 420
 421         *t = 0;
 422         return r;
 423 }
 424
 425 size_t utf16_encode_unichar(char16_t *out, char32_t c) {
 426
 427         /* Note that this encodes as little-endian. */
 428
 429         switch (c) {
 430
 431         case 0 ... 0xd7ffU:
 432         case 0xe000U ... 0xffffU:
 433                 out[0] = htole16(c);
 434                 return 1;
 435
 436         case 0x10000U ... 0x10ffffU:
 437                 c -= 0x10000U;
 438                 out[0] = htole16((c >> 10) + 0xd800U);
 439                 out[1] = htole16((c & 0x3ffU) + 0xdc00U);
 440                 return 2;
 441
 442         default: /* A surrogate (invalid) */
 443                 return 0;
 444         }
 445 }
 446
 447 char16_t *utf8_to_utf16(const char *s, size_t length) {
 448         char16_t *n, *p;
 449         int r;
 450
 451         if (length == 0)
 452                 return new0(char16_t, 1);
 453
 454         assert(s);
 455
 456         if (length == SIZE_MAX)
 457                 length = strlen(s);
 458
 459         if (length > SIZE_MAX - 1)
 460                 return NULL; /* overflow */
 461
 462         n = new(char16_t, length + 1);
 463         if (!n)
 464                 return NULL;
 465
 466         p = n;
 467
 468         for (size_t i = 0; i < length;) {
 469                 char32_t unichar;
 470                 size_t e;
 471
 472                 e = utf8_encoded_expected_len(s[i]);
 473                 if (e <= 1) /* Invalid and single byte characters are copied as they are */
 474                         goto copy;
 475
 476                 if (i + e > length) /* sequence longer than input buffer, then copy as-is */
 477                         goto copy;
 478
 479                 r = utf8_encoded_to_unichar(s + i, &unichar);
 480                 if (r < 0) /* sequence invalid, then copy as-is */
 481                         goto copy;
 482
 483                 p += utf16_encode_unichar(p, unichar);
 484                 i += e;
 485                 continue;
 486
 487         copy:
 488                 *(p++) = htole16(s[i++]);
 489         }
 490
 491         *p = 0;
 492         return n;
 493 }
 494
 495 size_t char16_strlen(const char16_t *s) {
 496         size_t n = 0;
 497
 498         assert(s);
 499
 500         while (*s != 0)
 501                 n++, s++;
 502
 503         return n;
 504 }
 505
 506 size_t char16_strsize(const char16_t *s) {
 507         POINTER_MAY_BE_NULL(s);
 508
 509         return s ? (char16_strlen(s) + 1) * sizeof(*s) : 0;
 510 }
 511
 512 /* expected size used to encode one unicode char */
 513 static int utf8_unichar_to_encoded_len(char32_t unichar) {
 514
 515         if (unichar < 0x80)
 516                 return 1;
 517         if (unichar < 0x800)
 518                 return 2;
 519         if (unichar < 0x10000)
 520                 return 3;
 521         if (unichar < 0x200000)
 522                 return 4;
 523         if (unichar < 0x4000000)
 524                 return 5;
 525
 526         return 6;
 527 }
 528
 529 /* validate one encoded unicode char and return its length */
 530 int utf8_encoded_valid_unichar(const char *str, size_t length /* bytes */) {
 531         char32_t unichar;
 532         size_t len;
 533         int r;
 534
 535         assert(str);
 536         assert(length > 0);
 537
 538         /* We read until NUL, at most length bytes. SIZE_MAX may be used to disable the length check. */
 539
 540         len = utf8_encoded_expected_len(str[0]);
 541         if (len == 0)
 542                 return -EINVAL;
 543
 544         /* Do we have a truncated multi-byte character? */
 545         if (len > length)
 546                 return -EINVAL;
 547
 548         /* ascii is valid */
 549         if (len == 1)
 550                 return 1;
 551
 552         /* check if expected encoded chars are available */
 553         for (size_t i = 0; i < len; i++)
 554                 if ((str[i] & 0x80) != 0x80)
 555                         return -EINVAL;
 556
 557         r = utf8_encoded_to_unichar(str, &unichar);
 558         if (r < 0)
 559                 return r;
 560
 561         /* check if encoded length matches encoded value */
 562         if (utf8_unichar_to_encoded_len(unichar) != (int) len)
 563                 return -EINVAL;
 564
 565         /* check if value has valid range */
 566         if (!unichar_is_valid(unichar))
 567                 return -EINVAL;
 568
 569         return (int) len;
 570 }
 571
 572 size_t utf8_n_codepoints(const char *str) {
 573         size_t n = 0;
 574
 575         assert(str);
 576
 577         /* Returns the number of UTF-8 codepoints in this string, or SIZE_MAX if the string is not valid UTF-8. */
 578
 579         while (*str != 0) {
 580                 int k;
 581
 582                 k = utf8_encoded_valid_unichar(str, SIZE_MAX);
 583                 if (k < 0)
 584                         return SIZE_MAX;
 585
 586                 str += k;
 587                 n++;
 588         }
 589
 590         return n;
 591 }
 592
 593 size_t utf8_console_width(const char *str) {
 594         POINTER_MAY_BE_NULL(str);
 595
 596         if (isempty(str))
 597                 return 0;
 598
 599         /* Returns the approximate width a string will take on screen when printed on a character cell
 600          * terminal/console. */
 601
 602         size_t n = 0;
 603         while (*str) {
 604                 int w;
 605
 606                 w = utf8_char_console_width(str);
 607                 if (w < 0)
 608                         return SIZE_MAX;
 609
 610                 n += w;
 611                 str = utf8_next_char(str);
 612         }
 613
 614         return n;
 615 }
 616
 617 size_t utf8_last_length(const char *s, size_t n) {
 618         int r;
 619
 620         assert(s);
 621
 622         if (n == SIZE_MAX)
 623                 n = strlen(s);
 624
 625         /* Determines length in bytes of last UTF-8 codepoint in string. If the string is empty, returns
 626          * zero. Treats invalid UTF-8 codepoints as 1 sized ones. */
 627
 628         for (size_t last = 0;;) {
 629                 if (n == 0)
 630                         return last;
 631
 632                 r = utf8_encoded_valid_unichar(s, n);
 633                 if (r <= 0)
 634                         r = 1; /* treat invalid UTF-8 as byte-wide */
 635
 636                 s += r;
 637                 n -= r;
 638                 last = r;
 639         }
 640 }