| 1 |
//! bzip2 low-level block APIs |
| 2 |
|
| 3 |
use std::mem; |
| 4 |
|
| 5 |
use tinyvec::ArrayVec; |
| 6 |
|
| 7 |
pub use self::error::BlockError; |
| 8 |
use crate::bitreader::BitReader; |
| 9 |
#[cfg(target_pointer_width = "64")] |
| 10 |
use crate::bitreader::CachedBitReader; |
| 11 |
use crate::crc::Hasher; |
| 12 |
use crate::header::Header; |
| 13 |
use crate::huffman::HuffmanTree; |
| 14 |
use crate::move_to_front::MoveToFrontDecoder; |
| 15 |
|
| 16 |
mod bwt; |
| 17 |
mod error; |
| 18 |
|
| 19 |
const BLOCK_MAGIC: u64 = 0x314159265359; |
| 20 |
const FINAL_MAGIC: u64 = 0x177245385090; |
| 21 |
|
| 22 |
pub(crate) struct Block { |
| 23 |
header: Header, |
| 24 |
|
| 25 |
tt: Vec<u32>, |
| 26 |
pre_rle_used: u32, |
| 27 |
t_pos: u32, |
| 28 |
last_byte: i16, |
| 29 |
byte_repeats: u8, |
| 30 |
repeats: u8, |
| 31 |
|
| 32 |
hasher: Hasher, |
| 33 |
expected_crc: u32, |
| 34 |
|
| 35 |
state: State, |
| 36 |
} |
| 37 |
|
| 38 |
enum State { |
| 39 |
ReadyForRead, |
| 40 |
Reading, |
| 41 |
NotReady, |
| 42 |
} |
| 43 |
|
| 44 |
impl Block { |
| 45 |
pub fn new(header: Header) -> Self { |
| 46 |
let max_blocksize = header.max_blocksize(); |
| 47 |
|
| 48 |
Self { |
| 49 |
header, |
| 50 |
|
| 51 |
tt: Vec::with_capacity(max_blocksize as usize), |
| 52 |
pre_rle_used: 0, |
| 53 |
t_pos: 0, |
| 54 |
last_byte: -1, |
| 55 |
byte_repeats: 0, |
| 56 |
repeats: 0, |
| 57 |
|
| 58 |
hasher: Hasher::new(), |
| 59 |
expected_crc: 0, |
| 60 |
|
| 61 |
state: State::NotReady, |
| 62 |
} |
| 63 |
} |
| 64 |
|
| 65 |
pub fn is_ready_for_read(&self) -> bool { |
| 66 |
match self.state { |
| 67 |
State::ReadyForRead => true, |
| 68 |
State::Reading | State::NotReady => false, |
| 69 |
} |
| 70 |
} |
| 71 |
|
| 72 |
pub fn is_reading(&self) -> bool { |
| 73 |
match self.state { |
| 74 |
State::Reading => true, |
| 75 |
State::ReadyForRead | State::NotReady => false, |
| 76 |
} |
| 77 |
} |
| 78 |
|
| 79 |
pub fn is_not_ready(&self) -> bool { |
| 80 |
match self.state { |
| 81 |
State::NotReady => true, |
| 82 |
State::ReadyForRead | State::Reading => false, |
| 83 |
} |
| 84 |
} |
| 85 |
|
| 86 |
pub fn set_ready_for_read(&mut self) { |
| 87 |
self.state = State::ReadyForRead; |
| 88 |
} |
| 89 |
|
| 90 |
pub fn read( |
| 91 |
&mut self, |
| 92 |
reader: &mut BitReader<'_>, |
| 93 |
out: &mut [u8], |
| 94 |
) -> Result<usize, BlockError> { |
| 95 |
match &self.state { |
| 96 |
State::ReadyForRead => { |
| 97 |
let magic = reader |
| 98 |
.read_u64(48) |
| 99 |
.ok_or_else(|| BlockError::new("next magic truncated"))?; |
| 100 |
match magic { |
| 101 |
BLOCK_MAGIC => { |
| 102 |
self.read_block(reader)?; |
| 103 |
self.state = State::Reading; |
| 104 |
|
| 105 |
self.read(reader, out) |
| 106 |
} |
| 107 |
FINAL_MAGIC => { |
| 108 |
let _crc = reader |
| 109 |
.read_u32(32) |
| 110 |
.ok_or_else(|| BlockError::new("whole stream crc truncated"))?; |
| 111 |
|
| 112 |
// TODO: check whole stream crc |
| 113 |
|
| 114 |
self.state = State::NotReady; |
| 115 |
Ok(0) |
| 116 |
} |
| 117 |
_ => { |
| 118 |
self.state = State::NotReady; |
| 119 |
Err(BlockError::new("bad magic value found")) |
| 120 |
} |
| 121 |
} |
| 122 |
} |
| 123 |
State::Reading => { |
| 124 |
let n = self.read_from_block(out)?; |
| 125 |
if n == 0 && !out.is_empty() { |
| 126 |
self.state = State::NotReady; |
| 127 |
} |
| 128 |
|
| 129 |
Ok(n) |
| 130 |
} |
| 131 |
State::NotReady => Err(BlockError::new("not ready")), |
| 132 |
} |
| 133 |
} |
| 134 |
|
| 135 |
pub fn read_from_block(&mut self, out: &mut [u8]) -> Result<usize, BlockError> { |
| 136 |
let mut read = 0; |
| 137 |
|
| 138 |
while (self.repeats > 0 || self.pre_rle_used < (self.tt.len() as u32)) && read < out.len() { |
| 139 |
if self.repeats > 0 { |
| 140 |
out[read] = self.last_byte as u8; |
| 141 |
read += 1; |
| 142 |
|
| 143 |
self.repeats -= 1; |
| 144 |
if self.repeats == 0 { |
| 145 |
self.last_byte = -1; |
| 146 |
} |
| 147 |
continue; |
| 148 |
} |
| 149 |
|
| 150 |
self.t_pos = self.tt[self.t_pos as usize]; |
| 151 |
let b = self.t_pos as u8; |
| 152 |
self.t_pos >>= 8; |
| 153 |
self.pre_rle_used += 1; |
| 154 |
|
| 155 |
if self.byte_repeats == 3 { |
| 156 |
self.repeats = b; |
| 157 |
self.byte_repeats = 0; |
| 158 |
continue; |
| 159 |
} |
| 160 |
|
| 161 |
if self.last_byte == i16::from(b) { |
| 162 |
self.byte_repeats += 1; |
| 163 |
} else { |
| 164 |
self.byte_repeats = 0; |
| 165 |
} |
| 166 |
self.last_byte = i16::from(b); |
| 167 |
|
| 168 |
out[read] = b; |
| 169 |
read += 1; |
| 170 |
} |
| 171 |
|
| 172 |
if read == 0 && !out.is_empty() { |
| 173 |
self.state = State::NotReady; |
| 174 |
|
| 175 |
let crc = self.hasher.finalyze(); |
| 176 |
return if self.expected_crc == crc { |
| 177 |
Ok(0) |
| 178 |
} else { |
| 179 |
Err(BlockError::new("bad crc")) |
| 180 |
}; |
| 181 |
} |
| 182 |
|
| 183 |
self.hasher.update(&out[..read]); |
| 184 |
Ok(read) |
| 185 |
} |
| 186 |
|
| 187 |
pub fn read_block(&mut self, reader: &mut BitReader<'_>) -> Result<(), BlockError> { |
| 188 |
self.hasher = Hasher::new(); |
| 189 |
self.tt.clear(); |
| 190 |
|
| 191 |
self.expected_crc = reader |
| 192 |
.read_u32(32) |
| 193 |
.ok_or_else(|| BlockError::new("crc truncated"))?; |
| 194 |
|
| 195 |
let randomised = reader |
| 196 |
.read_bool() |
| 197 |
.ok_or_else(|| BlockError::new("randomised truncated"))?; |
| 198 |
if randomised { |
| 199 |
return Err(BlockError::new("randomised expected to be 'normal'")); |
| 200 |
} |
| 201 |
|
| 202 |
let orig_ptr = reader |
| 203 |
.read_u32(24) |
| 204 |
.ok_or_else(|| BlockError::new("orig ptr truncated"))?; |
| 205 |
|
| 206 |
let mut huffman_used_symbols = ArrayVec::<[u8; 16]>::new(); |
| 207 |
let mut huffman_used_bitmaps = ArrayVec::<[u8; 256]>::new(); |
| 208 |
|
| 209 |
for i in 0..16 { |
| 210 |
if reader |
| 211 |
.read_bool() |
| 212 |
.ok_or_else(|| BlockError::new("symbol range truncated"))? |
| 213 |
{ |
| 214 |
huffman_used_symbols.push(i); |
| 215 |
} |
| 216 |
} |
| 217 |
|
| 218 |
for symbol_range in huffman_used_symbols { |
| 219 |
for symbol in 0..16 { |
| 220 |
if reader |
| 221 |
.read_bool() |
| 222 |
.ok_or_else(|| BlockError::new("symbol range truncated"))? |
| 223 |
{ |
| 224 |
huffman_used_bitmaps.push(symbol_range * 16 + symbol); |
| 225 |
} |
| 226 |
} |
| 227 |
} |
| 228 |
|
| 229 |
if huffman_used_bitmaps.is_empty() { |
| 230 |
return Err(BlockError::new("huffman no symbols in input")); |
| 231 |
} |
| 232 |
|
| 233 |
let huffman_groups = reader |
| 234 |
.read_u8(3) |
| 235 |
.ok_or_else(|| BlockError::new("huffmann groups truncated"))?; |
| 236 |
if huffman_groups < 2 || huffman_groups > 6 { |
| 237 |
return Err(BlockError::new("invalid number of huffman trees")); |
| 238 |
} |
| 239 |
|
| 240 |
let selectors_used = reader |
| 241 |
.read_u16(15) |
| 242 |
.ok_or_else(|| BlockError::new("selectors used truncated"))?; |
| 243 |
|
| 244 |
#[cfg(feature = "nightly")] |
| 245 |
let mut selectors_list = ArrayVec::<[u8; 18001]>::new(); |
| 246 |
#[cfg(feature = "nightly")] |
| 247 |
selectors_list.set_len(usize::from(selectors_used)); |
| 248 |
#[cfg(not(feature = "nightly"))] |
| 249 |
let mut selectors_list = vec![0u8; usize::from(selectors_used)]; |
| 250 |
|
| 251 |
let mut move_to_front_decoder = MoveToFrontDecoder::new(); |
| 252 |
for selector_item in selectors_list.iter_mut().rev() { |
| 253 |
let mut trees = 0; |
| 254 |
|
| 255 |
while reader |
| 256 |
.read_bool() |
| 257 |
.ok_or_else(|| BlockError::new("selector truncated"))? |
| 258 |
{ |
| 259 |
trees += 1; |
| 260 |
|
| 261 |
if trees >= huffman_groups { |
| 262 |
return Err(BlockError::new("tree index too large")); |
| 263 |
} |
| 264 |
} |
| 265 |
|
| 266 |
*selector_item = move_to_front_decoder.decode(trees); |
| 267 |
} |
| 268 |
|
| 269 |
// limited in lenght of huffman_used_symbols |
| 270 |
let mut symbols = [0u8; 256]; |
| 271 |
symbols[..huffman_used_bitmaps.len()].copy_from_slice(&huffman_used_bitmaps); |
| 272 |
let mut move_to_front_decoder_2 = MoveToFrontDecoder::new_from_symbols(symbols); |
| 273 |
|
| 274 |
let mut huffman_trees = ArrayVec::<[HuffmanTree; 6]>::new(); |
| 275 |
|
| 276 |
let mut lengths = ArrayVec::<[u8; crate::LEN_258]>::new(); |
| 277 |
lengths.set_len(huffman_used_bitmaps.len() + 2); |
| 278 |
|
| 279 |
for _ in 0..huffman_groups { |
| 280 |
let mut length = reader |
| 281 |
.read_u8(5) |
| 282 |
.ok_or_else(|| BlockError::new("huffman group length truncated"))?; |
| 283 |
|
| 284 |
for length_item in &mut *lengths { |
| 285 |
loop { |
| 286 |
if length < 1 || length > 20 { |
| 287 |
return Err(BlockError::new("huffman length out of range")); |
| 288 |
} |
| 289 |
|
| 290 |
if !reader |
| 291 |
.read_bool() |
| 292 |
.ok_or_else(|| BlockError::new("length bit1 truncated"))? |
| 293 |
{ |
| 294 |
break; |
| 295 |
} |
| 296 |
|
| 297 |
if reader |
| 298 |
.read_bool() |
| 299 |
.ok_or_else(|| BlockError::new("length bit2 truncated"))? |
| 300 |
{ |
| 301 |
length -= 1; |
| 302 |
} else { |
| 303 |
length += 1; |
| 304 |
} |
| 305 |
} |
| 306 |
|
| 307 |
*length_item = length; |
| 308 |
} |
| 309 |
|
| 310 |
let tree = HuffmanTree::new(&lengths).map_err(BlockError::new)?; |
| 311 |
huffman_trees.push(tree); |
| 312 |
} |
| 313 |
|
| 314 |
let selector = selectors_list |
| 315 |
.pop() |
| 316 |
.ok_or_else(|| BlockError::new("no tree selectors given"))?; |
| 317 |
let mut current_huffman_tree = huffman_trees |
| 318 |
.get(usize::from(selector)) |
| 319 |
.ok_or_else(|| BlockError::new("tree selector out of range"))?; |
| 320 |
|
| 321 |
let mut repeat = 0u32; |
| 322 |
let mut repeat_power = 0u32; |
| 323 |
let mut c = [0u32; 256]; |
| 324 |
|
| 325 |
let mut decoded = 0; |
| 326 |
#[cfg(target_pointer_width = "64")] |
| 327 |
let mut r = CachedBitReader::new(reader) |
| 328 |
.ok_or_else(|| BlockError::new("huffman bitstream truncated"))?; |
| 329 |
loop { |
| 330 |
if decoded == 50 { |
| 331 |
let selector = selectors_list.pop().ok_or_else(|| { |
| 332 |
BlockError::new("insufficient selector indices for number of symbols") |
| 333 |
})?; |
| 334 |
|
| 335 |
current_huffman_tree = huffman_trees |
| 336 |
.get(usize::from(selector)) |
| 337 |
.ok_or_else(|| BlockError::new("tree selector out of range"))?; |
| 338 |
decoded = 0; |
| 339 |
} |
| 340 |
|
| 341 |
#[cfg(target_pointer_width = "64")] |
| 342 |
let v = { |
| 343 |
let read = r.read(); |
| 344 |
let v = current_huffman_tree.decode(&mut r); |
| 345 |
|
| 346 |
if r.overflowed() { |
| 347 |
r.restore(reader, read); |
| 348 |
r.refresh(reader) |
| 349 |
.ok_or_else(|| BlockError::new("huffman bitstream truncated"))?; |
| 350 |
|
| 351 |
current_huffman_tree.decode(&mut r) |
| 352 |
} else { |
| 353 |
v |
| 354 |
} |
| 355 |
}; |
| 356 |
#[cfg(not(target_pointer_width = "64"))] |
| 357 |
let v = current_huffman_tree.decode(reader); |
| 358 |
|
| 359 |
let v = v.ok_or_else(|| BlockError::new("huffman bitstream truncated"))?; |
| 360 |
decoded += 1; |
| 361 |
|
| 362 |
if v < 2 { |
| 363 |
if repeat == 0 { |
| 364 |
repeat_power = 1; |
| 365 |
} |
| 366 |
repeat += repeat_power << v; |
| 367 |
repeat_power <<= 1; |
| 368 |
|
| 369 |
if repeat > 2 * 1024 * 1024 { |
| 370 |
return Err(BlockError::new("repeat count too large")); |
| 371 |
} |
| 372 |
continue; |
| 373 |
} |
| 374 |
|
| 375 |
let old_repeat = mem::replace(&mut repeat, 0); |
| 376 |
if old_repeat > 0 { |
| 377 |
if old_repeat > self.header.max_blocksize() - (self.tt.len() as u32) { |
| 378 |
return Err(BlockError::new("repeats past end of block")); |
| 379 |
} |
| 380 |
|
| 381 |
let b = move_to_front_decoder_2.first(); |
| 382 |
// extend self.tt with `b` repeated `old_repeat` times |
| 383 |
let new_len = self.tt.len() + old_repeat as usize; |
| 384 |
self.tt.resize(new_len, u32::from(b)); |
| 385 |
c[usize::from(b)] += old_repeat; |
| 386 |
} |
| 387 |
|
| 388 |
if usize::from(v) == (huffman_used_bitmaps.len() + 2) - 1 { |
| 389 |
break; |
| 390 |
} |
| 391 |
|
| 392 |
let b = move_to_front_decoder_2.decode((v - 1) as u8); |
| 393 |
if self.tt.len() >= self.header.max_blocksize() as usize { |
| 394 |
return Err(BlockError::new("data exceeds block size")); |
| 395 |
} |
| 396 |
|
| 397 |
self.tt.push(u32::from(b)); |
| 398 |
c[usize::from(b)] += 1; |
| 399 |
} |
| 400 |
#[cfg(target_pointer_width = "64")] |
| 401 |
r.restore(reader, r.read()); |
| 402 |
|
| 403 |
if (orig_ptr as usize) >= self.tt.len() { |
| 404 |
return Err(BlockError::new("orig_ptr out of bounds")); |
| 405 |
} |
| 406 |
|
| 407 |
self.pre_rle_used = 0; |
| 408 |
self.t_pos = bwt::inverse_bwt(&mut self.tt, orig_ptr as usize, c); |
| 409 |
self.last_byte = -1; |
| 410 |
self.byte_repeats = 0; |
| 411 |
self.repeats = 0; |
| 412 |
|
| 413 |
Ok(()) |
| 414 |
} |
| 415 |
} |
| 416 |
|
| 417 |
#[cfg(test)] |
| 418 |
mod tests { |
| 419 |
use std::convert::TryInto; |
| 420 |
|
| 421 |
use super::*; |
| 422 |
|
| 423 |
#[test] |
| 424 |
fn decode1() { |
| 425 |
let compressed = include_bytes!("../../tests/samplefiles/sample1.bz2"); |
| 426 |
let decompressed = include_bytes!("../../tests/samplefiles/sample1.ref"); |
| 427 |
|
| 428 |
let header = Header::parse(compressed[..4].try_into().unwrap()).unwrap(); |
| 429 |
println!("block_size: {}", header.raw_blocksize()); |
| 430 |
|
| 431 |
let compressed = &compressed[4..]; |
| 432 |
|
| 433 |
let mut bits = BitReader::new(&compressed); |
| 434 |
let mut reader = Block::new(header); |
| 435 |
|
| 436 |
let mut out = vec![0u8; decompressed.len()]; |
| 437 |
|
| 438 |
reader.set_ready_for_read(); |
| 439 |
let read = reader.read(&mut bits, &mut out).unwrap(); |
| 440 |
assert_eq!(&out[..read], decompressed.as_ref()); |
| 441 |
} |
| 442 |
|
| 443 |
#[test] |
| 444 |
fn decode2() { |
| 445 |
let compressed = include_bytes!("../../tests/samplefiles/sample2.bz2"); |
| 446 |
let decompressed = include_bytes!("../../tests/samplefiles/sample2.ref"); |
| 447 |
|
| 448 |
let header = Header::parse(compressed[..4].try_into().unwrap()).unwrap(); |
| 449 |
println!("block_size: {}", header.raw_blocksize()); |
| 450 |
|
| 451 |
let compressed = &compressed[4..]; |
| 452 |
|
| 453 |
let mut bits = BitReader::new(&compressed); |
| 454 |
let mut reader = Block::new(header); |
| 455 |
|
| 456 |
let mut out = vec![0u8; decompressed.len()]; |
| 457 |
|
| 458 |
reader.set_ready_for_read(); |
| 459 |
let read1 = reader.read(&mut bits, &mut out).unwrap(); |
| 460 |
reader.set_ready_for_read(); |
| 461 |
let read2 = reader.read(&mut bits, &mut out[read1..]).unwrap(); |
| 462 |
assert_eq!(&out[..read1 + read2], decompressed.as_ref()); |
| 463 |
} |
| 464 |
|
| 465 |
#[test] |
| 466 |
fn decode3() { |
| 467 |
let compressed = include_bytes!("../../tests/samplefiles/sample3.bz2"); |
| 468 |
let decompressed = include_bytes!("../../tests/samplefiles/sample3.ref"); |
| 469 |
|
| 470 |
let header = Header::parse(compressed[..4].try_into().unwrap()).unwrap(); |
| 471 |
println!("block_size: {}", header.raw_blocksize()); |
| 472 |
|
| 473 |
let compressed = &compressed[4..]; |
| 474 |
|
| 475 |
let mut bits = BitReader::new(&compressed); |
| 476 |
let mut reader = Block::new(header); |
| 477 |
|
| 478 |
let mut out = vec![0u8; decompressed.len()]; |
| 479 |
|
| 480 |
reader.set_ready_for_read(); |
| 481 |
let read = reader.read(&mut bits, &mut out).unwrap(); |
| 482 |
assert_eq!(&out[..read], decompressed.as_ref()); |
| 483 |
} |
| 484 |
} |