1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080
//! This crate provides an encoder/decoder for Reed-Solomon erasure code. //! //! Please note that erasure coding means errors are not directly detected or corrected, //! but missing data pieces (shards) can be reconstructed given that //! the configuration provides high enough redundancy. //! //! You will have to implement error detection separately (e.g. via checksums) //! and simply leave out the corrupted shards when attempting to reconstruct //! the missing data. #![allow(dead_code)] #[cfg(test)] #[macro_use] extern crate quickcheck; #[cfg(test)] extern crate rand; extern crate smallvec; #[cfg(feature = "simd-accel")] extern crate libc; use std::iter::{self, FromIterator}; use std::sync::Arc; use smallvec::SmallVec; #[macro_use] mod macros; mod errors; mod inversion_tree; mod matrix; #[cfg(test)] mod tests; pub mod galois_8; pub mod galois_16; pub use crate::errors::Error; pub use crate::errors::SBSError; use crate::inversion_tree::InversionTree; use crate::matrix::Matrix; /// A finite field to perform encoding over. pub trait Field: Sized { /// The order of the field. This is a limit on the number of shards /// in an encoding. const ORDER: usize; /// The representational type of the field. type Elem: Default + Clone + Copy + PartialEq + std::fmt::Debug; /// Add two elements together. fn add(a: Self::Elem, b: Self::Elem) -> Self::Elem; /// Multiply two elements together. fn mul(a: Self::Elem, b: Self::Elem) -> Self::Elem; /// Divide a by b. Panics is b is zero. fn div(a: Self::Elem, b: Self::Elem) -> Self::Elem; /// Raise `a` to the n'th power. fn exp(a: Self::Elem, n: usize) -> Self::Elem; /// The "zero" element or additive identity. fn zero() -> Self::Elem; /// The "one" element or multiplicative identity. fn one() -> Self::Elem; /// Yield the nth element of the field. Panics if n >= ORDER. /// Assignment is arbitrary but must be unique to `n`. fn nth(n: usize) -> Self::Elem; /// Multiply a slice of elements by another. Writes into the output slice. /// /// # Panics /// Panics if the output slice does not have equal length to the input. fn mul_slice(elem: Self::Elem, input: &[Self::Elem], out: &mut [Self::Elem]) { assert_eq!(input.len(), out.len()); for (i, o) in input.iter().zip(out) { *o = Self::mul(elem.clone(), i.clone()) } } /// Multiply a slice of elements by another, adding each result to the corresponding value in /// `out`. /// /// # Panics /// Panics if the output slice does not have equal length to the input. fn mul_slice_add(elem: Self::Elem, input: &[Self::Elem], out: &mut [Self::Elem]) { assert_eq!(input.len(), out.len()); for (i, o) in input.iter().zip(out) { *o = Self::add(o.clone(), Self::mul(elem.clone(), i.clone())) } } } /// Something which might hold a shard. /// /// This trait is used in reconstruction, where some of the shards /// may be unknown. pub trait ReconstructShard<F: Field> { /// The size of the shard data; `None` if empty. fn len(&self) -> Option<usize>; /// Get a mutable reference to the shard data, returning `None` if uninitialized. fn get(&mut self) -> Option<&mut [F::Elem]>; /// Get a mutable reference to the shard data, initializing it to the /// given length if it was `None`. Returns an error if initialization fails. fn get_or_initialize(&mut self, len: usize) -> Result<&mut [F::Elem], Result<&mut [F::Elem], Error>>; } impl<F: Field, T: AsRef<[F::Elem]> + AsMut<[F::Elem]> + FromIterator<F::Elem>> ReconstructShard<F> for Option<T> { fn len(&self) -> Option<usize> { self.as_ref().map(|x| x.as_ref().len()) } fn get(&mut self) -> Option<&mut [F::Elem]> { self.as_mut().map(|x| x.as_mut()) } fn get_or_initialize(&mut self, len: usize) -> Result<&mut [F::Elem], Result<&mut [F::Elem], Error>> { let is_some = self.is_some(); let x = self .get_or_insert_with(|| iter::repeat(F::zero()).take(len).collect()) .as_mut(); if is_some { Ok(x) } else { Err(Ok(x)) } } } impl<F: Field, T: AsRef<[F::Elem]> + AsMut<[F::Elem]>> ReconstructShard<F> for (T, bool) { fn len(&self) -> Option<usize> { if !self.1 { None } else { Some(self.0.as_ref().len()) } } fn get(&mut self) -> Option<&mut [F::Elem]> { if !self.1 { None } else { Some(self.0.as_mut()) } } fn get_or_initialize(&mut self, len: usize) -> Result<&mut [F::Elem], Result<&mut [F::Elem], Error>> { let x = self.0.as_mut(); if x.len() == len { if self.1 { Ok(x) } else { Err(Ok(x)) } } else { Err(Err(Error::IncorrectShardSize)) } } } /// Parameters for parallelism. #[derive(PartialEq, Debug, Clone, Copy)] pub struct ParallelParam { /// Number of bytes to split the slices into for computations /// which can be done in parallel. /// /// Default is 32768. pub bytes_per_encode: usize, } impl ParallelParam { /// Create a new `ParallelParam` with the given split arity. pub fn new(bytes_per_encode: usize) -> ParallelParam { ParallelParam { bytes_per_encode } } } impl Default for ParallelParam { fn default() -> Self { ParallelParam::new(32768) } } /// Bookkeeper for shard by shard encoding. /// /// This is useful for avoiding incorrect use of /// `encode_single` and `encode_single_sep` /// /// # Use cases /// /// Shard by shard encoding is useful for streamed data encoding /// where you do not have all the needed data shards immediately, /// but you want to spread out the encoding workload rather than /// doing the encoding after everything is ready. /// /// A concrete example would be network packets encoding, /// where encoding packet by packet as you receive them may be more efficient /// than waiting for N packets then encode them all at once. /// /// # Example /// /// ``` /// # #[macro_use] extern crate erase; /// # use erase::*; /// # fn main () { /// use erase::galois_8::Field; /// let r: ReedSolomon<Field> = ReedSolomon::new(3, 2).unwrap(); /// /// let mut sbs = ShardByShard::new(&r); /// /// let mut shards = shards!([0u8, 1, 2, 3, 4], /// [5, 6, 7, 8, 9], /// // say we don't have the 3rd data shard yet /// // and we want to fill it in later /// [0, 0, 0, 0, 0], /// [0, 0, 0, 0, 0], /// [0, 0, 0, 0, 0]); /// /// // encode 1st and 2nd data shard /// sbs.encode(&mut shards).unwrap(); /// sbs.encode(&mut shards).unwrap(); /// /// // fill in 3rd data shard /// shards[2][0] = 10.into(); /// shards[2][1] = 11.into(); /// shards[2][2] = 12.into(); /// shards[2][3] = 13.into(); /// shards[2][4] = 14.into(); /// /// // now do the encoding /// sbs.encode(&mut shards).unwrap(); /// /// assert!(r.verify(&shards).unwrap()); /// # } /// ``` #[derive(PartialEq, Debug)] pub struct ShardByShard<'a, F: 'a + Field> { codec: &'a ReedSolomon<F>, cur_input: usize, } impl<'a, F: 'a + Field> ShardByShard<'a, F> { /// Creates a new instance of the bookkeeping struct. pub fn new(codec: &'a ReedSolomon<F>) -> ShardByShard<'a, F> { ShardByShard { codec, cur_input: 0, } } /// Checks if the parity shards are ready to use. pub fn parity_ready(&self) -> bool { self.cur_input == self.codec.data_shard_count } /// Resets the bookkeeping data. /// /// You should call this when you have added and encoded /// all data shards, and have finished using the parity shards. /// /// Returns `SBSError::LeftoverShards` when there are shards encoded /// but parity shards are not ready to use. pub fn reset(&mut self) -> Result<(), SBSError> { if self.cur_input > 0 && !self.parity_ready() { return Err(SBSError::LeftoverShards); } self.cur_input = 0; Ok(()) } /// Resets the bookkeeping data without checking. pub fn reset_force(&mut self) { self.cur_input = 0; } /// Returns the current input shard index. pub fn cur_input_index(&self) -> usize { self.cur_input } fn return_ok_and_incre_cur_input(&mut self) -> Result<(), SBSError> { self.cur_input += 1; Ok(()) } fn sbs_encode_checks<U: AsRef<[F::Elem]> + AsMut<[F::Elem]>>( &mut self, slices: &mut [U], ) -> Result<(), SBSError> { let internal_checks = |codec: &ReedSolomon<F>, data: &mut [U]| { check_piece_count!(all => codec, data); check_slices!(multi => data); Ok(()) }; if self.parity_ready() { return Err(SBSError::TooManyCalls); } match internal_checks(self.codec, slices) { Ok(()) => Ok(()), Err(e) => Err(SBSError::RSError(e)), } } fn sbs_encode_sep_checks<T: AsRef<[F::Elem]>, U: AsRef<[F::Elem]> + AsMut<[F::Elem]>>( &mut self, data: &[T], parity: &mut [U], ) -> Result<(), SBSError> { let internal_checks = |codec: &ReedSolomon<F>, data: &[T], parity: &mut [U]| { check_piece_count!(data => codec, data); check_piece_count!(parity => codec, parity); check_slices!(multi => data, multi => parity); Ok(()) }; if self.parity_ready() { return Err(SBSError::TooManyCalls); } match internal_checks(self.codec, data, parity) { Ok(()) => Ok(()), Err(e) => Err(SBSError::RSError(e)), } } /// Constructs the parity shards partially using the current input data shard. /// /// Returns `SBSError::TooManyCalls` when all input data shards /// have already been filled in via `encode` pub fn encode<T, U>(&mut self, mut shards: T) -> Result<(), SBSError> where T: AsRef<[U]> + AsMut<[U]>, U: AsRef<[F::Elem]> + AsMut<[F::Elem]>, { let shards = shards.as_mut(); self.sbs_encode_checks(shards)?; self.codec.encode_single(self.cur_input, shards).unwrap(); self.return_ok_and_incre_cur_input() } /// Constructs the parity shards partially using the current input data shard. /// /// Returns `SBSError::TooManyCalls` when all input data shards /// have already been filled in via `encode` pub fn encode_sep<T: AsRef<[F::Elem]>, U: AsRef<[F::Elem]> + AsMut<[F::Elem]>>( &mut self, data: &[T], parity: &mut [U], ) -> Result<(), SBSError> { self.sbs_encode_sep_checks(data, parity)?; self.codec .encode_single_sep(self.cur_input, data[self.cur_input].as_ref(), parity) .unwrap(); self.return_ok_and_incre_cur_input() } } /// Reed-Solomon erasure code encoder/decoder. /// /// # Common error handling /// /// ## For `encode`, `encode_shards`, `verify`, `verify_shards`, `reconstruct`, `reconstruct_data`, `reconstruct_shards`, `reconstruct_data_shards` /// /// Return `Error::TooFewShards` or `Error::TooManyShards` /// when the number of provided shards /// does not match the codec's one. /// /// Return `Error::EmptyShard` when the first shard provided is /// of zero length. /// /// Return `Error::IncorrectShardSize` when the provided shards /// are of different lengths. /// /// ## For `reconstruct`, `reconstruct_data`, `reconstruct_shards`, `reconstruct_data_shards` /// /// Return `Error::TooFewShardsPresent` when there are not /// enough shards for reconstruction. /// /// Return `Error::InvalidShardFlags` when the number of flags does not match /// the total number of shards. /// /// # Variants of encoding methods /// /// ## `sep` /// /// Methods ending in `_sep` takes an immutable reference to data shards, /// and a mutable reference to parity shards. /// /// They are useful as they do not need to borrow the data shards mutably, /// and other work that only needs read-only access to data shards can be done /// in parallel/concurrently during the encoding. /// /// Following is a table of all the `sep` variants /// /// | not `sep` | `sep` | /// | --- | --- | /// | `encode_single` | `encode_single_sep` | /// | `encode` | `encode_sep` | /// /// The `sep` variants do similar checks on the provided data shards and /// parity shards. /// /// Return `Error::TooFewDataShards`, `Error::TooManyDataShards`, /// `Error::TooFewParityShards`, or `Error::TooManyParityShards` when applicable. /// /// ## `single` /// /// Methods containing `single` facilitate shard by shard encoding, where /// the parity shards are partially constructed using one data shard at a time. /// See `ShardByShard` struct for more details on how shard by shard encoding /// can be useful. /// /// They are prone to **misuse**, and it is recommended to use the `ShardByShard` /// bookkeeping struct instead for shard by shard encoding. /// /// The ones that are also `sep` are **ESPECIALLY** prone to **misuse**. /// Only use them when you actually need the flexibility. /// /// Following is a table of all the shard by shard variants /// /// | all shards at once | shard by shard | /// | --- | --- | /// | `encode` | `encode_single` | /// | `encode_sep` | `encode_single_sep` | /// /// The `single` variants do similar checks on the provided data shards and parity shards, /// and also do index check on `i_data`. /// /// Return `Error::InvalidIndex` if `i_data >= data_shard_count`. /// /// # Encoding behaviour /// ## For `encode` /// /// You do not need to clear the parity shards beforehand, as the methods /// will overwrite them completely. /// /// ## For `encode_single`, `encode_single_sep` /// /// Calling them with `i_data` being `0` will overwrite the parity shards /// completely. If you are using the methods correctly, then you do not need /// to clear the parity shards beforehand. /// /// # Variants of verifying methods /// /// `verify` allocate sa buffer on the heap of the same size /// as the parity shards, and encode the input once using the buffer to store /// the computed parity shards, then check if the provided parity shards /// match the computed ones. /// /// `verify_with_buffer`, allows you to provide /// the buffer to avoid making heap allocation(s) for the buffer in every call. /// /// The `with_buffer` variants also guarantee that the buffer contains the correct /// parity shards if the result is `Ok(_)` (i.e. it does not matter whether the /// verification passed or not, as long as the result is not an error, the buffer /// will contain the correct parity shards after the call). /// /// Following is a table of all the `with_buffer` variants /// /// | not `with_buffer` | `with_buffer` | /// | --- | --- | /// | `verify` | `verify_with_buffer` | /// /// The `with_buffer` variants also check the dimensions of the buffer and return /// `Error::TooFewBufferShards`, `Error::TooManyBufferShards`, `Error::EmptyShard`, /// or `Error::IncorrectShardSize` when applicable. /// #[derive(Debug)] pub struct ReedSolomon<F: Field> { data_shard_count: usize, parity_shard_count: usize, total_shard_count: usize, matrix: Matrix<F>, tree: InversionTree<F>, } impl<F: Field> Clone for ReedSolomon<F> { fn clone(&self) -> ReedSolomon<F> { ReedSolomon::new( self.data_shard_count, self.parity_shard_count, ) .expect("basic checks already passed as precondition of existence of self") } } impl<F: Field> PartialEq for ReedSolomon<F> { fn eq(&self, rhs: &ReedSolomon<F>) -> bool { self.data_shard_count == rhs.data_shard_count && self.parity_shard_count == rhs.parity_shard_count } } impl<F: Field> ReedSolomon<F> { // AUDIT // // Error detection responsibilities // // Terminologies and symbols: // X =A, B, C=> Y: X delegates error checking responsibilities A, B, C to Y // X:= A, B, C: X needs to handle responsibilities A, B, C // // Encode methods // // `encode_single`:= // - check index `i_data` within range [0, data shard count) // - check length of `slices` matches total shard count exactly // - check consistency of length of individual slices // `encode_single_sep`:= // - check index `i_data` within range [0, data shard count) // - check length of `parity` matches parity shard count exactly // - check consistency of length of individual parity slices // - check length of `single_data` matches length of first parity slice // `encode`:= // - check length of `slices` matches total shard count exactly // - check consistency of length of individual slices // `encode_sep`:= // - check length of `data` matches data shard count exactly // - check length of `parity` matches parity shard count exactly // - check consistency of length of individual data slices // - check consistency of length of individual parity slices // - check length of first parity slice matches length of first data slice // // Verify methods // // `verify`:= // - check length of `slices` matches total shard count exactly // - check consistency of length of individual slices // // Generates buffer then passes control to verify_with_buffer // // `verify_with_buffer`:= // - check length of `slices` matches total shard count exactly // - check length of `buffer` matches parity shard count exactly // - check consistency of length of individual slices // - check consistency of length of individual slices in buffer // - check length of first slice in buffer matches length of first slice // // Reconstruct methods // // `reconstruct` =ALL=> `reconstruct_internal` // `reconstruct_data`=ALL=> `reconstruct_internal` // `reconstruct_internal`:= // - check length of `slices` matches total shard count exactly // - check consistency of length of individual slices // - check length of `slice_present` matches length of `slices` fn get_parity_rows(&self) -> SmallVec<[&[F::Elem]; 32]> { let mut parity_rows = SmallVec::with_capacity(self.parity_shard_count); let matrix = &self.matrix; for i in self.data_shard_count..self.total_shard_count { parity_rows.push(matrix.get_row(i)); } parity_rows } fn build_matrix(data_shards: usize, total_shards: usize) -> Matrix<F> { let vandermonde = Matrix::vandermonde(total_shards, data_shards); let top = vandermonde.sub_matrix(0, 0, data_shards, data_shards); vandermonde.multiply(&top.invert().unwrap()) } /// Creates a new instance of Reed-Solomon erasure code encoder/decoder. /// /// Returns `Error::TooFewDataShards` if `data_shards == 0`. /// /// Returns `Error::TooFewParityShards` if `parity_shards == 0`. /// /// Returns `Error::TooManyShards` if `data_shards + parity_shards > F::ORDER`. pub fn new(data_shards: usize, parity_shards: usize) -> Result<ReedSolomon<F>, Error> { if data_shards == 0 { return Err(Error::TooFewDataShards); } if parity_shards == 0 { return Err(Error::TooFewParityShards); } if data_shards + parity_shards > F::ORDER { return Err(Error::TooManyShards); } let total_shards = data_shards + parity_shards; let matrix = Self::build_matrix(data_shards, total_shards); Ok(ReedSolomon { data_shard_count: data_shards, parity_shard_count: parity_shards, total_shard_count: total_shards, matrix, tree: InversionTree::new(data_shards, parity_shards), }) } pub fn data_shard_count(&self) -> usize { self.data_shard_count } pub fn parity_shard_count(&self) -> usize { self.parity_shard_count } pub fn total_shard_count(&self) -> usize { self.total_shard_count } fn code_some_slices<T: AsRef<[F::Elem]>, U: AsMut<[F::Elem]>>( &self, matrix_rows: &[&[F::Elem]], inputs: &[T], outputs: &mut [U], ) { for i_input in 0..self.data_shard_count { self.code_single_slice(matrix_rows, i_input, inputs[i_input].as_ref(), outputs); } } fn code_single_slice<U: AsMut<[F::Elem]>>( &self, matrix_rows: &[&[F::Elem]], i_input: usize, input: &[F::Elem], outputs: &mut [U], ) { outputs.iter_mut().enumerate().for_each(|(i_row, output)| { let matrix_row_to_use = matrix_rows[i_row][i_input]; let output = output.as_mut(); if i_input == 0 { F::mul_slice(matrix_row_to_use, input, output); } else { F::mul_slice_add(matrix_row_to_use, input, output); } }) } fn check_some_slices_with_buffer<T, U>( &self, matrix_rows: &[&[F::Elem]], inputs: &[T], to_check: &[T], buffer: &mut [U], ) -> bool where T: AsRef<[F::Elem]>, U: AsRef<[F::Elem]> + AsMut<[F::Elem]>, { self.code_some_slices(matrix_rows, inputs, buffer); let at_least_one_mismatch_present = buffer .iter_mut() .enumerate() .map(|(i, expected_parity_shard)| expected_parity_shard.as_ref() == to_check[i].as_ref() ) .any(|x| !x); // find the first false (some slice is different from the expected one) !at_least_one_mismatch_present } /// Constructs the parity shards partially using only the data shard /// indexed by `i_data`. /// /// The slots where the parity shards sit at will be overwritten. /// /// # Warning /// /// You must apply this method on the data shards in strict sequential order (0..data shard count), /// otherwise the parity shards will be incorrect. /// /// It is recommended to use the `ShardByShard` bookkeeping struct instead of this method directly. pub fn encode_single<T, U>(&self, i_data: usize, mut shards: T) -> Result<(), Error> where T: AsRef<[U]> + AsMut<[U]>, U: AsRef<[F::Elem]> + AsMut<[F::Elem]>, { let slices = shards.as_mut(); check_slice_index!(data => self, i_data); check_piece_count!(all=> self, slices); check_slices!(multi => slices); // Get the slice of output buffers. let (mut_input, output) = slices.split_at_mut(self.data_shard_count); let input = mut_input[i_data].as_ref(); self.encode_single_sep(i_data, input, output) } /// Constructs the parity shards partially using only the data shard provided. /// /// The data shard must match the index `i_data`. /// /// The slots where the parity shards sit at will be overwritten. /// /// # Warning /// /// You must apply this method on the data shards in strict sequential order (0..data shard count), /// otherwise the parity shards will be incorrect. /// /// It is recommended to use the `ShardByShard` bookkeeping struct instead of this method directly. pub fn encode_single_sep<U: AsRef<[F::Elem]> + AsMut<[F::Elem]>>( &self, i_data: usize, single_data: &[F::Elem], parity: &mut [U], ) -> Result<(), Error> { check_slice_index!(data => self, i_data); check_piece_count!(parity => self, parity); check_slices!(multi => parity, single => single_data); let parity_rows = self.get_parity_rows(); // Do the coding. self.code_single_slice(&parity_rows, i_data, single_data, parity); Ok(()) } /// Constructs the parity shards. /// /// The slots where the parity shards sit at will be overwritten. pub fn encode<T, U>(&self, mut shards: T) -> Result<(), Error> where T: AsRef<[U]> + AsMut<[U]>, U: AsRef<[F::Elem]> + AsMut<[F::Elem]>, { let slices: &mut [U] = shards.as_mut(); check_piece_count!(all => self, slices); check_slices!(multi => slices); // Get the slice of output buffers. let (input, output) = slices.split_at_mut(self.data_shard_count); self.encode_sep(&*input, output) } /// Constructs the parity shards using a read-only view into the /// data shards. /// /// The slots where the parity shards sit at will be overwritten. pub fn encode_sep<T: AsRef<[F::Elem]>, U: AsRef<[F::Elem]> + AsMut<[F::Elem]>>( &self, data: &[T], parity: &mut [U], ) -> Result<(), Error> { check_piece_count!(data => self, data); check_piece_count!(parity => self, parity); check_slices!(multi => data, multi => parity); let parity_rows = self.get_parity_rows(); // Do the coding. self.code_some_slices(&parity_rows, data, parity); Ok(()) } /// Checks if the parity shards are correct. /// /// This is a wrapper of `verify_with_buffer`. pub fn verify<T: AsRef<[F::Elem]>>(&self, slices: &[T]) -> Result<bool, Error> { check_piece_count!(all => self, slices); check_slices!(multi => slices); let slice_len = slices[0].as_ref().len(); let mut buffer: SmallVec<[Vec<F::Elem>; 32]> = SmallVec::with_capacity(self.parity_shard_count); for _ in 0..self.parity_shard_count { buffer.push(vec![F::zero(); slice_len]); } self.verify_with_buffer(slices, &mut buffer) } /// Checks if the parity shards are correct. pub fn verify_with_buffer<T, U>(&self, slices: &[T], buffer: &mut [U]) -> Result<bool, Error> where T: AsRef<[F::Elem]>, U: AsRef<[F::Elem]> + AsMut<[F::Elem]>, { check_piece_count!(all => self, slices); check_piece_count!(parity_buf => self, buffer); check_slices!(multi => slices, multi => buffer); let data = &slices[0..self.data_shard_count]; let to_check = &slices[self.data_shard_count..]; let parity_rows = self.get_parity_rows(); Ok(self.check_some_slices_with_buffer(&parity_rows, data, to_check, buffer)) } /// Reconstructs all shards. /// /// The shards marked not present are only overwritten when no error /// is detected. All provided shards must have the same length. /// /// This means if the method returns an `Error`, then nothing is touched. /// /// `reconstruct`, `reconstruct_data`, `reconstruct_shards`, /// `reconstruct_data_shards` share the same core code base. pub fn reconstruct<T: ReconstructShard<F>>(&self, slices: &mut [T]) -> Result<(), Error> { self.reconstruct_internal(slices, false) } /// Reconstructs only the data shards. /// /// The shards marked not present are only overwritten when no error /// is detected. All provided shards must have the same length. /// /// This means if the method returns an `Error`, then nothing is touched. /// /// `reconstruct`, `reconstruct_data`, `reconstruct_shards`, /// `reconstruct_data_shards` share the same core code base. pub fn reconstruct_data<T: ReconstructShard<F>>(&self, slices: &mut [T]) -> Result<(), Error> { self.reconstruct_internal(slices, true) } fn get_data_decode_matrix( &self, valid_indices: &[usize], invalid_indices: &[usize], ) -> Arc<Matrix<F>> { // Attempt to get the cached inverted matrix out of the tree // based on the indices of the invalid rows. match self.tree.get_inverted_matrix(&invalid_indices) { // If the inverted matrix isn't cached in the tree yet we must // construct it ourselves and insert it into the tree for the // future. In this way the inversion tree is lazily loaded. None => { // Pull out the rows of the matrix that correspond to the // shards that we have and build a square matrix. This // matrix could be used to generate the shards that we have // from the original data. let mut sub_matrix = Matrix::new(self.data_shard_count, self.data_shard_count); for (sub_matrix_row, &valid_index) in valid_indices.into_iter().enumerate() { for c in 0..self.data_shard_count { sub_matrix.set(sub_matrix_row, c, self.matrix.get(valid_index, c)); } } // Invert the matrix, so we can go from the encoded shards // back to the original data. Then pull out the row that // generates the shard that we want to decode. Note that // since this matrix maps back to the original data, it can // be used to create a data shard, but not a parity shard. let data_decode_matrix = Arc::new(sub_matrix.invert().unwrap()); // Cache the inverted matrix in the tree for future use keyed on the // indices of the invalid rows. self.tree .insert_inverted_matrix(&invalid_indices, &data_decode_matrix) .unwrap(); data_decode_matrix } Some(m) => m, } } fn reconstruct_internal<T: ReconstructShard<F>>( &self, shards: &mut [T], data_only: bool, ) -> Result<(), Error> { check_piece_count!(all => self, shards); let data_shard_count = self.data_shard_count; // Quick check: are all of the shards present? If so, there's // nothing to do. let mut number_present = 0; let mut shard_len = None; for shard in shards.iter_mut() { if let Some(len) = shard.len() { if len == 0 { return Err(Error::EmptyShard); } number_present += 1; if let Some(old_len) = shard_len { if len != old_len { // mismatch between shards. return Err(Error::IncorrectShardSize); } } shard_len = Some(len); } } if number_present == self.total_shard_count { // Cool. All of the shards are there. We don't // need to do anything. return Ok(()); } // More complete sanity check if number_present < data_shard_count { return Err(Error::TooFewShardsPresent); } let shard_len = shard_len.expect("at least one shard present; qed"); // Pull out an array holding just the shards that // correspond to the rows of the submatrix. These shards // will be the input to the decoding process that re-creates // the missing data shards. // // Also, create an array of indices of the valid rows we do have // and the invalid rows we don't have. // // The valid indices are used to construct the data decode matrix, // the invalid indices are used to key the data decode matrix // in the inversion tree. // // We only need exactly N valid indices, where N = `data_shard_count`, // as the data decode matrix is a N x N matrix, thus only needs // N valid indices for determining the N rows to pick from // `self.matrix`. let mut sub_shards: SmallVec<[&[F::Elem]; 32]> = SmallVec::with_capacity(data_shard_count); let mut missing_data_slices: SmallVec<[&mut [F::Elem]; 32]> = SmallVec::with_capacity(self.parity_shard_count); let mut missing_parity_slices: SmallVec<[&mut [F::Elem]; 32]> = SmallVec::with_capacity(self.parity_shard_count); let mut valid_indices: SmallVec<[usize; 32]> = SmallVec::with_capacity(data_shard_count); let mut invalid_indices: SmallVec<[usize; 32]> = SmallVec::with_capacity(data_shard_count); // Separate the shards into groups for (matrix_row, shard) in shards.into_iter().enumerate() { // get or initialize the shard so we can reconstruct in-place, // but if we are only reconstructing data shard, // do not initialize if the shard is not a data shard let shard_data = if matrix_row >= data_shard_count && data_only { shard.get().ok_or(None) } else { shard.get_or_initialize(shard_len).map_err(Some) }; match shard_data { Ok(shard) => { if sub_shards.len() < data_shard_count { sub_shards.push(shard); valid_indices.push(matrix_row); } else { // Already have enough shards in `sub_shards` // as we only need N shards, where N = `data_shard_count`, // for the data decode matrix // // So nothing to do here } } Err(None) => { // the shard data is not meant to be initialized here, // but we should still note it missing. invalid_indices.push(matrix_row); } Err(Some(x)) => { // initialized missing shard data. let shard = x?; if matrix_row < data_shard_count { missing_data_slices.push(shard); } else { missing_parity_slices.push(shard); } invalid_indices.push(matrix_row); } } } let data_decode_matrix = self.get_data_decode_matrix(&valid_indices, &invalid_indices); // Re-create any data shards that were missing. // // The input to the coding is all of the shards we actually // have, and the output is the missing data shards. The computation // is done using the special decode matrix we just built. let mut matrix_rows: SmallVec<[&[F::Elem]; 32]> = SmallVec::with_capacity(self.parity_shard_count); for i_slice in invalid_indices .iter() .cloned() .take_while(|i| i < &data_shard_count) { matrix_rows.push(data_decode_matrix.get_row(i_slice)); } self.code_some_slices(&matrix_rows, &sub_shards, &mut missing_data_slices); if data_only { Ok(()) } else { // Now that we have all of the data shards intact, we can // compute any of the parity that is missing. // // The input to the coding is ALL of the data shards, including // any that we just calculated. The output is whichever of the // parity shards were missing. let mut matrix_rows: SmallVec<[&[F::Elem]; 32]> = SmallVec::with_capacity(self.parity_shard_count); let parity_rows = self.get_parity_rows(); for i_slice in invalid_indices .iter() .cloned() .skip_while(|i| i < &data_shard_count) { matrix_rows.push(parity_rows[i_slice - data_shard_count]); } { // Gather up all the data shards. // old data shards are in `sub_shards`, // new ones are in `missing_data_slices`. let mut i_old_data_slice = 0; let mut i_new_data_slice = 0; let mut all_data_slices: SmallVec<[&[F::Elem]; 32]> = SmallVec::with_capacity(data_shard_count); let mut next_maybe_good = 0; let mut push_good_up_to = move |data_slices: &mut SmallVec<_>, up_to| { // if next_maybe_good == up_to, this loop is a no-op. for _ in next_maybe_good..up_to { // push all good indices we just skipped. data_slices.push(sub_shards[i_old_data_slice]); i_old_data_slice += 1; } next_maybe_good = up_to + 1; }; for i_slice in invalid_indices .iter() .cloned() .take_while(|i| i < &data_shard_count) { push_good_up_to(&mut all_data_slices, i_slice); all_data_slices.push(missing_data_slices[i_new_data_slice]); i_new_data_slice += 1; } push_good_up_to(&mut all_data_slices, data_shard_count); // Now do the actual computation for the missing // parity shards self.code_some_slices(&matrix_rows, &all_data_slices, &mut missing_parity_slices); } Ok(()) } } }