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