| 403 | } |
| 404 | |
| 405 | func (d decoder) parseString(b []byte) ([]byte, []byte, Kind, error) { |
| 406 | if len(b) < 2 { |
| 407 | return nil, b[len(b):], Undefined, unexpectedEOF(b) |
| 408 | } |
| 409 | if b[0] != '"' { |
| 410 | return nil, b, Undefined, syntaxError(b, "expected '\"' at the beginning of a string value") |
| 411 | } |
| 412 | |
| 413 | var n int |
| 414 | if len(b) >= 9 { |
| 415 | // This is an optimization for short strings. We read 8/16 bytes, |
| 416 | // and XOR each with 0x22 (") so that these bytes (and only |
| 417 | // these bytes) are now zero. We use the hasless(u,1) trick |
| 418 | // from https://graphics.stanford.edu/~seander/bithacks.html#ZeroInWord |
| 419 | // to determine whether any bytes are zero. Finally, we CTZ |
| 420 | // to find the index of that byte. |
| 421 | const mask1 = 0x2222222222222222 |
| 422 | const mask2 = 0x0101010101010101 |
| 423 | const mask3 = 0x8080808080808080 |
| 424 | u := binary.LittleEndian.Uint64(b[1:]) ^ mask1 |
| 425 | if mask := (u - mask2) & ^u & mask3; mask != 0 { |
| 426 | n = bits.TrailingZeros64(mask)/8 + 2 |
| 427 | goto found |
| 428 | } |
| 429 | if len(b) >= 17 { |
| 430 | u = binary.LittleEndian.Uint64(b[9:]) ^ mask1 |
| 431 | if mask := (u - mask2) & ^u & mask3; mask != 0 { |
| 432 | n = bits.TrailingZeros64(mask)/8 + 10 |
| 433 | goto found |
| 434 | } |
| 435 | } |
| 436 | } |
| 437 | n = bytes.IndexByte(b[1:], '"') + 2 |
| 438 | if n <= 1 { |
| 439 | return nil, b[len(b):], Undefined, syntaxError(b, "missing '\"' at the end of a string value") |
| 440 | } |
| 441 | found: |
| 442 | if (d.flags.has(noBackslash) || bytes.IndexByte(b[1:n], '\\') < 0) && |
| 443 | (d.flags.has(validAsciiPrint) || ascii.ValidPrint(b[1:n])) { |
| 444 | return b[:n], b[n:], Unescaped, nil |
| 445 | } |
| 446 | |
| 447 | for i := 1; i < len(b); i++ { |
| 448 | switch b[i] { |
| 449 | case '\\': |
| 450 | if i++; i < len(b) { |
| 451 | switch b[i] { |
| 452 | case '"', '\\', '/', 'n', 'r', 't', 'f', 'b': |
| 453 | case 'u': |
| 454 | _, n, err := d.parseUnicode(b[i+1:]) |
| 455 | if err != nil { |
| 456 | return nil, b[i+1+n:], Undefined, err |
| 457 | } |
| 458 | i += n |
| 459 | default: |
| 460 | return nil, b, Undefined, syntaxError(b, "invalid character '%c' in string escape code", b[i]) |
| 461 | } |
| 462 | } |