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         assert(str);
 183
 184         r = utf8_encoded_to_unichar(str, &c);
 185         if (r < 0)
 186                 return r;
 187
 188         return unichar_console_width(c);
 189 }
 190
 191 int unichar_console_width(char32_t c) {
 192         if (c == '\t')
 193                 return 8; /* Assume a tab width of 8 */
 194
 195         /* TODO: we should detect combining characters */
 196
 197         return unichar_iswide(c) ? 2 : 1;
 198 }
 199
 200 char* utf8_escape_non_printable_full(const char *str, size_t console_width, bool force_ellipsis) {
 201         char *p, *s, *prev_s;
 202         size_t n = 0; /* estimated print width */
 203
 204         assert(str);
 205
 206         if (console_width == 0)
 207                 return strdup("");
 208
 209         size_t body = strlen(str) * 4;
 210         p = s = prev_s = malloc(body + STRLEN("…") + 1);
 211         if (!p)
 212                 return NULL;
 213
 214         for (;;) {
 215                 int len;
 216                 char *saved_s = s;
 217
 218                 if (!*str) { /* done! */
 219                         if (force_ellipsis)
 220                                 goto truncation;
 221                         else
 222                                 goto finish;
 223                 }
 224
 225                 len = utf8_encoded_valid_unichar(str, SIZE_MAX);
 226                 if (len > 0) {
 227                         if (utf8_is_printable(str, len)) {
 228                                 int w;
 229
 230                                 w = utf8_char_console_width(str);
 231                                 assert(w >= 0);
 232                                 if (n + w > console_width)
 233                                         goto truncation;
 234
 235                                 s = mempcpy(s, str, len);
 236                                 str += len;
 237                                 n += w;
 238
 239                         } else {
 240                                 for (; len > 0; len--) {
 241                                         if (n + 4 > console_width)
 242                                                 goto truncation;
 243
 244                                         *(s++) = '\\';
 245                                         *(s++) = 'x';
 246                                         *(s++) = hexchar((int) *str >> 4);
 247                                         *(s++) = hexchar((int) *str);
 248
 249                                         str += 1;
 250                                         n += 4;
 251                                 }
 252                         }
 253                 } else {
 254                         if (n + 1 > console_width)
 255                                 goto truncation;
 256
 257                         s = mempcpy(s, UTF8_REPLACEMENT_CHARACTER, strlen(UTF8_REPLACEMENT_CHARACTER));
 258                         str += 1;
 259                         n += 1;
 260                 }
 261
 262                 prev_s = saved_s;
 263         }
 264
 265  truncation:
 266         /* Try to go back one if we don't have enough space for the ellipsis */
 267         if (n + 1 > console_width)
 268                 s = prev_s;
 269
 270         s = mempcpy(s, "…", strlen("…"));
 271
 272  finish:
 273         *s = '\0';
 274         return str_realloc(p);
 275 }
 276
 277 char* ascii_is_valid_n(const char *str, size_t len) {
 278         /* Check whether the string consists of valid ASCII bytes, i.e values between 1 and 127, inclusive.
 279          * Stops at len, or NUL byte if len is SIZE_MAX. */
 280
 281         assert(str);
 282
 283         for (size_t i = 0; len != SIZE_MAX ? i < len : str[i] != '\0'; i++)
 284                 if ((unsigned char) str[i] >= 128 || str[i] == '\0')
 285                         return NULL;
 286
 287         return (char*) str;
 288 }
 289
 290 int utf8_to_ascii(const char *str, char replacement_char, char **ret) {
 291         /* Convert to a string that has only ASCII chars, replacing anything that is not ASCII
 292          * by replacement_char. */
 293
 294         assert(str);
 295         assert(ret);
 296
 297         _cleanup_free_ char *ans = new(char, strlen(str) + 1);
 298         if (!ans)
 299                 return -ENOMEM;
 300
 301         char *q = ans;
 302
 303         for (const char *p = str; *p; q++) {
 304                 int l;
 305
 306                 l = utf8_encoded_valid_unichar(p, SIZE_MAX);
 307                 if (l < 0)  /* Non-UTF-8, let's not even try to propagate the garbage */
 308                         return l;
 309
 310                 if (l == 1)
 311                         *q = *p;
 312                 else
 313                         /* non-ASCII, we need to replace it */
 314                         *q = replacement_char;
 315
 316                 p += l;
 317         }
 318         *q = '\0';
 319
 320         *ret = TAKE_PTR(ans);
 321         return 0;
 322 }
 323
 324 /**
 325  * utf8_encode_unichar() - Encode single UCS-4 character as UTF-8
 326  * @out_utf8: output buffer of at least 4 bytes or NULL
 327  * @g: UCS-4 character to encode
 328  *
 329  * This encodes a single UCS-4 character as UTF-8 and writes it into @out_utf8.
 330  * The length of the character is returned. It is not zero-terminated! If the
 331  * output buffer is NULL, only the length is returned.
 332  *
 333  * Returns: The length in bytes that the UTF-8 representation does or would
 334  *          occupy.
 335  */
 336 size_t utf8_encode_unichar(char *out_utf8, char32_t g) {
 337
 338         if (g < (1 << 7)) {
 339                 if (out_utf8)
 340                         out_utf8[0] = g & 0x7f;
 341                 return 1;
 342         } else if (g < (1 << 11)) {
 343                 if (out_utf8) {
 344                         out_utf8[0] = 0xc0 | ((g >> 6) & 0x1f);
 345                         out_utf8[1] = 0x80 | (g & 0x3f);
 346                 }
 347                 return 2;
 348         } else if (g < (1 << 16)) {
 349                 if (out_utf8) {
 350                         out_utf8[0] = 0xe0 | ((g >> 12) & 0x0f);
 351                         out_utf8[1] = 0x80 | ((g >> 6) & 0x3f);
 352                         out_utf8[2] = 0x80 | (g & 0x3f);
 353                 }
 354                 return 3;
 355         } else if (g < (1 << 21)) {
 356                 if (out_utf8) {
 357                         out_utf8[0] = 0xf0 | ((g >> 18) & 0x07);
 358                         out_utf8[1] = 0x80 | ((g >> 12) & 0x3f);
 359                         out_utf8[2] = 0x80 | ((g >> 6) & 0x3f);
 360                         out_utf8[3] = 0x80 | (g & 0x3f);
 361                 }
 362                 return 4;
 363         }
 364
 365         return 0;
 366 }
 367
 368 char* utf16_to_utf8(const char16_t *s, size_t length /* bytes! */) {
 369         const uint8_t *f;
 370         char *r, *t;
 371
 372         if (length == 0)
 373                 return new0(char, 1);
 374
 375         assert(s);
 376
 377         if (length == SIZE_MAX) {
 378                 length = char16_strlen(s);
 379
 380                 if (length > SIZE_MAX/2)
 381                         return NULL; /* overflow */
 382
 383                 length *= 2;
 384         }
 385
 386         /* Input length is in bytes, i.e. the shortest possible character takes 2 bytes. Each unicode character may
 387          * take up to 4 bytes in UTF-8. Let's also account for a trailing NUL byte. */
 388         if (length > (SIZE_MAX - 1) / 2)
 389                 return NULL; /* overflow */
 390
 391         r = new(char, length * 2 + 1);
 392         if (!r)
 393                 return NULL;
 394
 395         f = (const uint8_t*) s;
 396         t = r;
 397
 398         while (f + 1 < (const uint8_t*) s + length) {
 399                 char16_t w1, w2;
 400
 401                 /* see RFC 2781 section 2.2 */
 402
 403                 w1 = f[1] << 8 | f[0];
 404                 f += 2;
 405
 406                 if (!utf16_is_surrogate(w1)) {
 407                         t += utf8_encode_unichar(t, w1);
 408                         continue;
 409                 }
 410
 411                 if (utf16_is_trailing_surrogate(w1))
 412                         continue; /* spurious trailing surrogate, ignore */
 413
 414                 if (f + 1 >= (const uint8_t*) s + length)
 415                         break;
 416
 417                 w2 = f[1] << 8 | f[0];
 418                 f += 2;
 419
 420                 if (!utf16_is_trailing_surrogate(w2)) {
 421                         f -= 2;
 422                         continue; /* surrogate missing its trailing surrogate, ignore */
 423                 }
 424
 425                 t += utf8_encode_unichar(t, utf16_surrogate_pair_to_unichar(w1, w2));
 426         }
 427
 428         *t = 0;
 429         return r;
 430 }
 431
 432 size_t utf16_encode_unichar(char16_t *out, char32_t c) {
 433         assert(out);
 434
 435         /* Note that this encodes as little-endian. */
 436
 437         switch (c) {
 438
 439         case 0 ... 0xd7ffU:
 440         case 0xe000U ... 0xffffU:
 441                 out[0] = htole16(c);
 442                 return 1;
 443
 444         case 0x10000U ... 0x10ffffU:
 445                 c -= 0x10000U;
 446                 out[0] = htole16((c >> 10) + 0xd800U);
 447                 out[1] = htole16((c & 0x3ffU) + 0xdc00U);
 448                 return 2;
 449
 450         default: /* A surrogate (invalid) */
 451                 return 0;
 452         }
 453 }
 454
 455 char16_t *utf8_to_utf16(const char *s, size_t length) {
 456         char16_t *n, *p;
 457         int r;
 458
 459         if (length == 0)
 460                 return new0(char16_t, 1);
 461
 462         assert(s);
 463
 464         if (length == SIZE_MAX)
 465                 length = strlen(s);
 466
 467         if (length > SIZE_MAX - 1)
 468                 return NULL; /* overflow */
 469
 470         n = new(char16_t, length + 1);
 471         if (!n)
 472                 return NULL;
 473
 474         p = n;
 475
 476         for (size_t i = 0; i < length;) {
 477                 char32_t unichar;
 478                 size_t e;
 479
 480                 e = utf8_encoded_expected_len(s[i]);
 481                 if (e <= 1) /* Invalid and single byte characters are copied as they are */
 482                         goto copy;
 483
 484                 if (i + e > length) /* sequence longer than input buffer, then copy as-is */
 485                         goto copy;
 486
 487                 r = utf8_encoded_to_unichar(s + i, &unichar);
 488                 if (r < 0) /* sequence invalid, then copy as-is */
 489                         goto copy;
 490
 491                 p += utf16_encode_unichar(p, unichar);
 492                 i += e;
 493                 continue;
 494
 495         copy:
 496                 *(p++) = htole16(s[i++]);
 497         }
 498
 499         *p = 0;
 500         return n;
 501 }
 502
 503 size_t char16_strlen(const char16_t *s) {
 504         size_t n = 0;
 505
 506         assert(s);
 507
 508         while (*s != 0)
 509                 n++, s++;
 510
 511         return n;
 512 }
 513
 514 size_t char16_strsize(const char16_t *s) {
 515         POINTER_MAY_BE_NULL(s);
 516
 517         return s ? (char16_strlen(s) + 1) * sizeof(*s) : 0;
 518 }
 519
 520 /* expected size used to encode one unicode char */
 521 static int utf8_unichar_to_encoded_len(char32_t unichar) {
 522
 523         if (unichar < 0x80)
 524                 return 1;
 525         if (unichar < 0x800)
 526                 return 2;
 527         if (unichar < 0x10000)
 528                 return 3;
 529         if (unichar < 0x200000)
 530                 return 4;
 531         if (unichar < 0x4000000)
 532                 return 5;
 533
 534         return 6;
 535 }
 536
 537 /* validate one encoded unicode char and return its length */
 538 int utf8_encoded_valid_unichar(const char *str, size_t length /* bytes */) {
 539         char32_t unichar;
 540         size_t len;
 541         int r;
 542
 543         assert(str);
 544         assert(length > 0);
 545
 546         /* We read until NUL, at most length bytes. SIZE_MAX may be used to disable the length check. */
 547
 548         len = utf8_encoded_expected_len(str[0]);
 549         if (len == 0)
 550                 return -EINVAL;
 551
 552         /* Do we have a truncated multi-byte character? */
 553         if (len > length)
 554                 return -EINVAL;
 555
 556         /* ascii is valid */
 557         if (len == 1)
 558                 return 1;
 559
 560         /* check if expected encoded chars are available */
 561         for (size_t i = 0; i < len; i++)
 562                 if ((str[i] & 0x80) != 0x80)
 563                         return -EINVAL;
 564
 565         r = utf8_encoded_to_unichar(str, &unichar);
 566         if (r < 0)
 567                 return r;
 568
 569         /* check if encoded length matches encoded value */
 570         if (utf8_unichar_to_encoded_len(unichar) != (int) len)
 571                 return -EINVAL;
 572
 573         /* check if value has valid range */
 574         if (!unichar_is_valid(unichar))
 575                 return -EINVAL;
 576
 577         return (int) len;
 578 }
 579
 580 size_t utf8_n_codepoints(const char *str) {
 581         size_t n = 0;
 582
 583         assert(str);
 584
 585         /* Returns the number of UTF-8 codepoints in this string, or SIZE_MAX if the string is not valid UTF-8. */
 586
 587         while (*str != 0) {
 588                 int k;
 589
 590                 k = utf8_encoded_valid_unichar(str, SIZE_MAX);
 591                 if (k < 0)
 592                         return SIZE_MAX;
 593
 594                 str += k;
 595                 n++;
 596         }
 597
 598         return n;
 599 }
 600
 601 size_t utf8_console_width(const char *str) {
 602         POINTER_MAY_BE_NULL(str);
 603
 604         if (isempty(str))
 605                 return 0;
 606
 607         /* Returns the approximate width a string will take on screen when printed on a character cell
 608          * terminal/console. */
 609
 610         size_t n = 0;
 611         while (*str) {
 612                 int w;
 613
 614                 w = utf8_char_console_width(str);
 615                 if (w < 0)
 616                         return SIZE_MAX;
 617
 618                 n += w;
 619                 str = utf8_next_char(str);
 620         }
 621
 622         return n;
 623 }
 624
 625 size_t utf8_last_length(const char *s, size_t n) {
 626         int r;
 627
 628         assert(s);
 629
 630         if (n == SIZE_MAX)
 631                 n = strlen(s);
 632
 633         /* Determines length in bytes of last UTF-8 codepoint in string. If the string is empty, returns
 634          * zero. Treats invalid UTF-8 codepoints as 1 sized ones. */
 635
 636         for (size_t last = 0;;) {
 637                 if (n == 0)
 638                         return last;
 639
 640                 r = utf8_encoded_valid_unichar(s, n);
 641                 if (r <= 0)
 642                         r = 1; /* treat invalid UTF-8 as byte-wide */
 643
 644                 s += r;
 645                 n -= r;
 646                 last = r;
 647         }
 648 }