Move ZSTD_transferSequences_wBlockDelim's delimiter scan, literal copying, SeqDef storage, validation, and repcode bookkeeping into the Rust compression statistics module. Keep ZSTD_CCtx and block-state ownership in C through a narrow scalar/pointer shim, with ABI layout assertions for ZSTD_SequencePosition. Test Plan: - cargo test --manifest-path rust/Cargo.toml --no-default-features --features compression - cargo clippy --manifest-path rust/Cargo.toml --no-default-features --features compression --benches --tests - make -B -C lib -j2 lib - make -C tests -j2 test-zstream
3841 lines
128 KiB
Rust
3841 lines
128 KiB
Rust
#![allow(non_camel_case_types)]
|
|
#![allow(non_snake_case)]
|
|
|
|
//! Sequence statistics and seqStore entropy compression.
|
|
//!
|
|
//! This module translates the block entropy-statistics layer of
|
|
//! `zstd_compress.c`: `ZSTD_seqToCodes()`, `ZSTD_buildSequencesStatistics()`,
|
|
//! `ZSTD_entropyCompressSeqStore*()`, `ZSTD_buildBlockEntropyStats*()`, and
|
|
//! `ZSTD_copyBlockSequences()`. The `ZSTD_CCtx` and `ZSTD_CCtx_params`
|
|
//! layouts stay private to C: the C shims extract the sequence store, the
|
|
//! entropy-table leaves, and the two parameter scalars these paths read
|
|
//! (the compression strategy and the literals-compression switch). Block
|
|
//! dispatch and block-splitting recursion remain in C; the pure block-size
|
|
//! estimator is owned here.
|
|
|
|
use crate::bits::ZSTD_highbit32;
|
|
use crate::common::{
|
|
DEFAULT_MAX_OFF, LL_BITS, LL_DEFAULT_NORM, LL_DEFAULT_NORM_LOG, LL_FSE_LOG, LONGNBSEQ, MAX_LL,
|
|
MAX_ML, MAX_OFF, MAX_SEQ, MINMATCH, ML_BITS, ML_DEFAULT_NORM, ML_DEFAULT_NORM_LOG, ML_FSE_LOG,
|
|
OFF_FSE_LOG, OF_DEFAULT_NORM, OF_DEFAULT_NORM_LOG, REP_START_VALUE, ZSTD_MAX_FSE_HEADERS_SIZE,
|
|
ZSTD_MAX_HUF_HEADER_SIZE, ZSTD_REP_NUM,
|
|
};
|
|
use crate::errors::{ERR_isError, ZstdErrorCode, ERROR};
|
|
use crate::hist::{HIST_countFast_wksp, HIST_count_wksp};
|
|
use crate::huf_compress::{
|
|
HUF_buildCTable_wksp, HUF_estimateCompressedSize, HUF_optimalTableLog, HUF_validateCTable,
|
|
HUF_writeCTable_wksp,
|
|
};
|
|
use crate::mem::{MEM_32bits, MEM_readST, MEM_writeLE16};
|
|
use crate::zstd_compress_literals::{
|
|
copy_huf_tables, min_gain, ZSTD_compressLiterals, ZSTD_hufCTables_t,
|
|
};
|
|
use crate::zstd_compress_sequences::{
|
|
SeqDef, ZSTD_buildCTable, ZSTD_crossEntropyCost, ZSTD_encodeSequences, ZSTD_fseBitCost,
|
|
ZSTD_selectEncodingType,
|
|
};
|
|
use std::ffi::c_void;
|
|
use std::mem::{size_of, size_of_val};
|
|
use std::os::raw::{c_int, c_uint};
|
|
use std::ptr;
|
|
|
|
const SET_BASIC: c_int = 0;
|
|
const SET_RLE: c_int = 1;
|
|
const SET_COMPRESSED: c_int = 2;
|
|
const SET_REPEAT: c_int = 3;
|
|
|
|
const ZSTD_PS_AUTO: c_int = 0;
|
|
const ZSTD_PS_DISABLE: c_int = 2;
|
|
|
|
const FSE_REPEAT_NONE: c_int = 0;
|
|
|
|
const HUF_REPEAT_NONE: c_int = 0;
|
|
const HUF_REPEAT_CHECK: c_int = 1;
|
|
const HUF_REPEAT_VALID: c_int = 2;
|
|
|
|
const HUF_FLAGS_OPTIMAL_DEPTH: c_int = 1 << 1;
|
|
const HUF_OPTIMAL_DEPTH_THRESHOLD: c_int = 8; /* ZSTD_btultra */
|
|
|
|
const ZSTD_DEFAULT_DISALLOWED: c_int = 0;
|
|
const ZSTD_DEFAULT_ALLOWED: c_int = 1;
|
|
|
|
const HUF_SYMBOLVALUE_MAX: c_uint = 255;
|
|
const LIT_HUF_LOG: c_uint = 11;
|
|
const COMPRESS_LITERALS_SIZE_MIN: usize = 63;
|
|
const SUSPECT_UNCOMPRESSIBLE_LITERAL_RATIO: usize = 20;
|
|
const ZSTD_BLOCKSIZE_MAX: usize = 1 << 17;
|
|
|
|
const STREAM_ACCUMULATOR_MIN_32: u32 = 25;
|
|
const STREAM_ACCUMULATOR_MIN_64: u32 = 57;
|
|
|
|
const LL_DELTA_CODE: u32 = 19;
|
|
const ML_DELTA_CODE: u32 = 36;
|
|
|
|
pub(crate) const ZSTD_LLT_LITERAL_LENGTH: c_int = 1;
|
|
pub(crate) const ZSTD_LLT_MATCH_LENGTH: c_int = 2;
|
|
|
|
/// `FSE_CTABLE_SIZE_U32(OffFSELog, MaxOff)` from the C headers.
|
|
pub const OFF_CTABLE_SIZE: usize = 1 + (1 << (OFF_FSE_LOG - 1)) + ((MAX_OFF + 1) * 2);
|
|
/// `FSE_CTABLE_SIZE_U32(MLFSELog, MaxML)` from the C headers.
|
|
pub const ML_CTABLE_SIZE: usize = 1 + (1 << (ML_FSE_LOG - 1)) + ((MAX_ML + 1) * 2);
|
|
/// `FSE_CTABLE_SIZE_U32(LLFSELog, MaxLL)` from the C headers.
|
|
pub const LL_CTABLE_SIZE: usize = 1 + (1 << (LL_FSE_LOG - 1)) + ((MAX_LL + 1) * 2);
|
|
|
|
/// ABI-compatible `SeqStore_t` leaf layout from `zstd_compress_internal.h`.
|
|
///
|
|
/// The C context itself remains opaque; its shims pass the `seqStore` member
|
|
/// directly, so this small, stable hot-path structure is the only sequence
|
|
/// storage representation crossing into Rust.
|
|
#[repr(C)]
|
|
pub struct SeqStore_t {
|
|
pub sequencesStart: *mut SeqDef,
|
|
pub sequences: *mut SeqDef,
|
|
pub litStart: *mut u8,
|
|
pub lit: *mut u8,
|
|
pub llCode: *mut u8,
|
|
pub mlCode: *mut u8,
|
|
pub ofCode: *mut u8,
|
|
pub maxNbSeq: usize,
|
|
pub maxNbLit: usize,
|
|
pub longLengthType: c_int,
|
|
pub longLengthPos: u32,
|
|
}
|
|
|
|
/// C's `ZSTD_fseCTables_t`. The table element type is `FSE_CTable`, an
|
|
/// `unsigned`, and the table lengths are the header macros expanded above.
|
|
#[repr(C)]
|
|
pub struct ZSTD_fseCTables_t {
|
|
pub offcodeCTable: [u32; OFF_CTABLE_SIZE],
|
|
pub matchlengthCTable: [u32; ML_CTABLE_SIZE],
|
|
pub litlengthCTable: [u32; LL_CTABLE_SIZE],
|
|
pub offcode_repeatMode: c_int,
|
|
pub matchlength_repeatMode: c_int,
|
|
pub litlength_repeatMode: c_int,
|
|
}
|
|
|
|
/// C's `ZSTD_entropyCTables_t`.
|
|
#[repr(C)]
|
|
pub struct ZSTD_entropyCTables_t {
|
|
pub huf: ZSTD_hufCTables_t,
|
|
pub fse: ZSTD_fseCTables_t,
|
|
}
|
|
|
|
/// C's `ZSTD_compressedBlockState_t`.
|
|
#[repr(C)]
|
|
pub struct ZSTD_compressedBlockState_t {
|
|
pub entropy: ZSTD_entropyCTables_t,
|
|
pub rep: [u32; ZSTD_REP_NUM],
|
|
}
|
|
|
|
/// C's `ZSTD_hufCTablesMetadata_t`.
|
|
#[repr(C)]
|
|
pub struct ZSTD_hufCTablesMetadata_t {
|
|
pub hType: c_int,
|
|
pub hufDesBuffer: [u8; ZSTD_MAX_HUF_HEADER_SIZE],
|
|
pub hufDesSize: usize,
|
|
}
|
|
|
|
/// C's `ZSTD_fseCTablesMetadata_t`.
|
|
#[repr(C)]
|
|
pub struct ZSTD_fseCTablesMetadata_t {
|
|
pub llType: c_int,
|
|
pub ofType: c_int,
|
|
pub mlType: c_int,
|
|
pub fseTablesBuffer: [u8; ZSTD_MAX_FSE_HEADERS_SIZE],
|
|
pub fseTablesSize: usize,
|
|
pub lastCountSize: usize,
|
|
}
|
|
|
|
/// C's `ZSTD_entropyCTablesMetadata_t`.
|
|
#[repr(C)]
|
|
pub struct ZSTD_entropyCTablesMetadata_t {
|
|
pub hufMetadata: ZSTD_hufCTablesMetadata_t,
|
|
pub fseMetadata: ZSTD_fseCTablesMetadata_t,
|
|
}
|
|
|
|
/// Public `ZSTD_Sequence` from `zstd.h`.
|
|
#[repr(C)]
|
|
#[derive(Clone, Copy)]
|
|
pub struct ZSTD_Sequence {
|
|
pub offset: u32,
|
|
pub litLength: u32,
|
|
pub matchLength: u32,
|
|
pub rep: u32,
|
|
}
|
|
|
|
/// ABI-compatible `ZSTD_SequencePosition` from `zstd_compress_internal.h`.
|
|
#[repr(C)]
|
|
#[derive(Clone, Copy, Debug, Default)]
|
|
pub struct ZSTD_SequencePosition {
|
|
pub idx: u32,
|
|
pub posInSequence: u32,
|
|
pub posInSrc: usize,
|
|
}
|
|
|
|
/// ABI-compatible `BlockSummary` from `zstd_compress_internal.h`.
|
|
#[repr(C)]
|
|
#[derive(Clone, Copy, Debug, Default)]
|
|
pub struct BlockSummary {
|
|
pub nbSequences: usize,
|
|
pub blockSize: usize,
|
|
pub litSize: usize,
|
|
}
|
|
|
|
/// Converts public sequences to the internal no-repcodes `SeqDef` format.
|
|
///
|
|
/// The return value is a side-band marker: zero means that every length fits
|
|
/// in its `u16` field; otherwise it is one plus the match-length sequence
|
|
/// index, or `nb_sequences + 1` plus the literal-length sequence index. This
|
|
/// is the Rust leaf for C's `convertSequences_noRepcodes()`.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_convertSequencesNoRepcodes(
|
|
dst_seqs: *mut SeqDef,
|
|
in_seqs: *const ZSTD_Sequence,
|
|
nb_sequences: usize,
|
|
) -> usize {
|
|
if nb_sequences == 0 {
|
|
return 0;
|
|
}
|
|
|
|
let dst_seqs = unsafe { std::slice::from_raw_parts_mut(dst_seqs, nb_sequences) };
|
|
let in_seqs = unsafe { std::slice::from_raw_parts(in_seqs, nb_sequences) };
|
|
let mut long_length = 0usize;
|
|
|
|
for (index, input) in in_seqs.iter().enumerate() {
|
|
dst_seqs[index] = SeqDef {
|
|
offBase: input.offset.wrapping_add(ZSTD_REP_NUM as u32),
|
|
litLength: input.litLength as u16,
|
|
mlBase: input.matchLength.wrapping_sub(MINMATCH as u32) as u16,
|
|
};
|
|
|
|
if input.matchLength > 65535 + MINMATCH as u32 {
|
|
debug_assert_eq!(long_length, 0);
|
|
long_length = index + 1;
|
|
}
|
|
if input.litLength > 65535 {
|
|
debug_assert_eq!(long_length, 0);
|
|
long_length = index + nb_sequences + 1;
|
|
}
|
|
}
|
|
|
|
long_length
|
|
}
|
|
|
|
/// Finds the first explicit block delimiter and totals the sequences before
|
|
/// and including it. This is the Rust leaf for C's
|
|
/// `ZSTD_get1BlockSummary()`; the delimiter is identified by zero match
|
|
/// length and contributes its literal length to the totals.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_get1BlockSummary(
|
|
seqs: *const ZSTD_Sequence,
|
|
nb_seqs: usize,
|
|
) -> BlockSummary {
|
|
let invalid = || BlockSummary {
|
|
nbSequences: ERROR(ZstdErrorCode::ExternalSequencesInvalid),
|
|
blockSize: 0,
|
|
litSize: 0,
|
|
};
|
|
|
|
if nb_seqs == 0 || seqs.is_null() {
|
|
return invalid();
|
|
}
|
|
|
|
let seqs = unsafe { std::slice::from_raw_parts(seqs, nb_seqs) };
|
|
let mut total_match_size = 0usize;
|
|
let mut lit_size = 0usize;
|
|
|
|
for (index, sequence) in seqs.iter().enumerate() {
|
|
total_match_size = total_match_size.wrapping_add(sequence.matchLength as usize);
|
|
lit_size = lit_size.wrapping_add(sequence.litLength as usize);
|
|
if sequence.matchLength == 0 {
|
|
debug_assert_eq!(sequence.offset, 0);
|
|
return BlockSummary {
|
|
nbSequences: index + 1,
|
|
blockSize: lit_size.wrapping_add(total_match_size),
|
|
litSize: lit_size,
|
|
};
|
|
}
|
|
}
|
|
|
|
invalid()
|
|
}
|
|
|
|
/// Estimates the literal-section size used by the block splitter.
|
|
unsafe fn estimate_block_size_literal(
|
|
literals: *const u8,
|
|
lit_size: usize,
|
|
huf: *const ZSTD_hufCTables_t,
|
|
huf_metadata: *const ZSTD_hufCTablesMetadata_t,
|
|
workspace: *mut c_void,
|
|
wksp_size: usize,
|
|
write_entropy: bool,
|
|
) -> usize {
|
|
let metadata = unsafe { &*huf_metadata };
|
|
let literal_header_size =
|
|
3 + usize::from(lit_size >= 1024) + usize::from(lit_size >= 16 * 1024);
|
|
let single_stream = lit_size < 256;
|
|
|
|
match metadata.hType {
|
|
SET_BASIC => lit_size,
|
|
SET_RLE => 1,
|
|
SET_COMPRESSED | SET_REPEAT => {
|
|
let mut max_symbol_value = HUF_SYMBOLVALUE_MAX;
|
|
let largest = unsafe {
|
|
HIST_count_wksp(
|
|
workspace.cast::<u32>(),
|
|
&mut max_symbol_value,
|
|
literals.cast::<c_void>(),
|
|
lit_size,
|
|
workspace,
|
|
wksp_size,
|
|
)
|
|
};
|
|
if ERR_isError(largest) {
|
|
return lit_size;
|
|
}
|
|
|
|
let mut estimate = unsafe {
|
|
HUF_estimateCompressedSize(
|
|
(*huf).CTable.as_ptr(),
|
|
workspace.cast::<u32>(),
|
|
max_symbol_value,
|
|
)
|
|
};
|
|
if write_entropy {
|
|
estimate = estimate.wrapping_add(metadata.hufDesSize);
|
|
}
|
|
if !single_stream {
|
|
estimate = estimate.wrapping_add(6);
|
|
}
|
|
estimate.wrapping_add(literal_header_size)
|
|
}
|
|
_ => {
|
|
debug_assert!(false, "invalid literal encoding type");
|
|
0
|
|
}
|
|
}
|
|
}
|
|
|
|
/// Estimates one FSE symbol stream used by the block splitter.
|
|
#[allow(clippy::too_many_arguments)]
|
|
unsafe fn estimate_block_size_symbol_type(
|
|
encoding_type: c_int,
|
|
code_table: *const u8,
|
|
nb_seq: usize,
|
|
max_code: u32,
|
|
fse_ctable: *const u32,
|
|
additional_bits: *const u8,
|
|
default_norm: *const i16,
|
|
default_norm_log: u32,
|
|
default_max: u32,
|
|
workspace: *mut c_void,
|
|
wksp_size: usize,
|
|
) -> usize {
|
|
let mut max = max_code;
|
|
unsafe {
|
|
let _ = HIST_countFast_wksp(
|
|
workspace.cast::<u32>(),
|
|
&mut max,
|
|
code_table.cast::<c_void>(),
|
|
nb_seq,
|
|
workspace,
|
|
wksp_size,
|
|
);
|
|
}
|
|
|
|
let mut estimate_in_bits = match encoding_type {
|
|
SET_BASIC => {
|
|
debug_assert!(max <= default_max);
|
|
unsafe { ZSTD_crossEntropyCost(default_norm, default_norm_log, workspace.cast(), max) }
|
|
}
|
|
SET_RLE => 0,
|
|
SET_COMPRESSED | SET_REPEAT => unsafe {
|
|
ZSTD_fseBitCost(fse_ctable, workspace.cast(), max)
|
|
},
|
|
_ => 0,
|
|
};
|
|
if ERR_isError(estimate_in_bits) {
|
|
return nb_seq.wrapping_mul(10);
|
|
}
|
|
|
|
for index in 0..nb_seq {
|
|
let code = unsafe { *code_table.add(index) };
|
|
let additional = if additional_bits.is_null() {
|
|
code as usize
|
|
} else {
|
|
unsafe { *additional_bits.add(code as usize) as usize }
|
|
};
|
|
estimate_in_bits = estimate_in_bits.wrapping_add(additional);
|
|
}
|
|
estimate_in_bits >> 3
|
|
}
|
|
|
|
/// Estimates the sequence-section size used by the block splitter.
|
|
#[allow(clippy::too_many_arguments)]
|
|
unsafe fn estimate_block_size_sequences(
|
|
of_code_table: *const u8,
|
|
ll_code_table: *const u8,
|
|
ml_code_table: *const u8,
|
|
nb_seq: usize,
|
|
fse_tables: *const ZSTD_fseCTables_t,
|
|
fse_metadata: *const ZSTD_fseCTablesMetadata_t,
|
|
workspace: *mut c_void,
|
|
wksp_size: usize,
|
|
write_entropy: bool,
|
|
) -> usize {
|
|
let metadata = unsafe { &*fse_metadata };
|
|
let tables = unsafe { &*fse_tables };
|
|
let header_size =
|
|
1 + 1 + usize::from(nb_seq >= 128) + usize::from(nb_seq >= LONGNBSEQ as usize);
|
|
let mut estimate = unsafe {
|
|
estimate_block_size_symbol_type(
|
|
metadata.ofType,
|
|
of_code_table,
|
|
nb_seq,
|
|
MAX_OFF as u32,
|
|
tables.offcodeCTable.as_ptr(),
|
|
ptr::null(),
|
|
OF_DEFAULT_NORM.as_ptr(),
|
|
OF_DEFAULT_NORM_LOG,
|
|
DEFAULT_MAX_OFF as u32,
|
|
workspace,
|
|
wksp_size,
|
|
)
|
|
};
|
|
estimate = estimate.wrapping_add(unsafe {
|
|
estimate_block_size_symbol_type(
|
|
metadata.llType,
|
|
ll_code_table,
|
|
nb_seq,
|
|
MAX_LL as u32,
|
|
tables.litlengthCTable.as_ptr(),
|
|
LL_BITS.as_ptr(),
|
|
LL_DEFAULT_NORM.as_ptr(),
|
|
LL_DEFAULT_NORM_LOG,
|
|
MAX_LL as u32,
|
|
workspace,
|
|
wksp_size,
|
|
)
|
|
});
|
|
estimate = estimate.wrapping_add(unsafe {
|
|
estimate_block_size_symbol_type(
|
|
metadata.mlType,
|
|
ml_code_table,
|
|
nb_seq,
|
|
MAX_ML as u32,
|
|
tables.matchlengthCTable.as_ptr(),
|
|
ML_BITS.as_ptr(),
|
|
ML_DEFAULT_NORM.as_ptr(),
|
|
ML_DEFAULT_NORM_LOG,
|
|
MAX_ML as u32,
|
|
workspace,
|
|
wksp_size,
|
|
)
|
|
});
|
|
if write_entropy {
|
|
estimate = estimate.wrapping_add(metadata.fseTablesSize);
|
|
}
|
|
estimate.wrapping_add(header_size)
|
|
}
|
|
|
|
/// Estimates a block's compressed size for block-split decisions. This is
|
|
/// the Rust leaf for C's `ZSTD_estimateBlockSize()` and keeps the workspace and
|
|
/// entropy-table representations on the existing C ABI.
|
|
#[allow(clippy::too_many_arguments)]
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_estimateBlockSize(
|
|
literals: *const u8,
|
|
lit_size: usize,
|
|
of_code_table: *const u8,
|
|
ll_code_table: *const u8,
|
|
ml_code_table: *const u8,
|
|
nb_seq: usize,
|
|
entropy: *const ZSTD_entropyCTables_t,
|
|
entropy_metadata: *const ZSTD_entropyCTablesMetadata_t,
|
|
workspace: *mut c_void,
|
|
wksp_size: usize,
|
|
write_lit_entropy: c_int,
|
|
write_seq_entropy: c_int,
|
|
) -> usize {
|
|
let entropy = unsafe { &*entropy };
|
|
let entropy_metadata = unsafe { &*entropy_metadata };
|
|
let literals_size = unsafe {
|
|
estimate_block_size_literal(
|
|
literals,
|
|
lit_size,
|
|
ptr::addr_of!(entropy.huf),
|
|
ptr::addr_of!(entropy_metadata.hufMetadata),
|
|
workspace,
|
|
wksp_size,
|
|
write_lit_entropy != 0,
|
|
)
|
|
};
|
|
let sequence_size = unsafe {
|
|
estimate_block_size_sequences(
|
|
of_code_table,
|
|
ll_code_table,
|
|
ml_code_table,
|
|
nb_seq,
|
|
ptr::addr_of!(entropy.fse),
|
|
ptr::addr_of!(entropy_metadata.fseMetadata),
|
|
workspace,
|
|
wksp_size,
|
|
write_seq_entropy != 0,
|
|
)
|
|
};
|
|
sequence_size.wrapping_add(literals_size).wrapping_add(3)
|
|
}
|
|
|
|
/// Converts a raw sequence offset to the stored offBase representation.
|
|
///
|
|
/// This is the Rust leaf for C's `ZSTD_finalizeOffBase()`. The repcode
|
|
/// numbering and the special `rep[0] - 1` form are part of the public sequence
|
|
/// ABI used by the external-sequence path.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_finalizeOffBase(
|
|
raw_offset: u32,
|
|
rep: *const u32,
|
|
ll0: u32,
|
|
) -> u32 {
|
|
let rep = unsafe { std::slice::from_raw_parts(rep, ZSTD_REP_NUM) };
|
|
let mut off_base = raw_offset.wrapping_add(ZSTD_REP_NUM as u32);
|
|
|
|
if ll0 == 0 && raw_offset == rep[0] {
|
|
off_base = 1;
|
|
} else if raw_offset == rep[1] {
|
|
off_base = 2u32.wrapping_sub(ll0);
|
|
} else if raw_offset == rep[2] {
|
|
off_base = 3u32.wrapping_sub(ll0);
|
|
} else if ll0 != 0 && raw_offset == rep[0].wrapping_sub(1) {
|
|
off_base = 3;
|
|
}
|
|
off_base
|
|
}
|
|
|
|
/// Validates one external sequence against the decoder window and match-size
|
|
/// rules. This is the Rust leaf for C's `ZSTD_validateSequence()`.
|
|
#[allow(clippy::too_many_arguments)]
|
|
#[no_mangle]
|
|
pub extern "C" fn ZSTD_rust_validateSequence(
|
|
off_base: u32,
|
|
match_length: u32,
|
|
min_match: u32,
|
|
pos_in_src: usize,
|
|
window_log: u32,
|
|
dict_size: usize,
|
|
use_sequence_producer: c_int,
|
|
) -> usize {
|
|
let window_size = 1u32 << window_log;
|
|
/* As long as the decoded position is within the window, the dictionary
|
|
* can extend the largest valid offset. Once it is past the window, the
|
|
* offset is limited to the window size itself. */
|
|
let offset_bound = if pos_in_src > window_size as usize {
|
|
window_size as usize
|
|
} else {
|
|
pos_in_src.wrapping_add(dict_size)
|
|
};
|
|
debug_assert!(offset_bound > 0);
|
|
|
|
let offset_bound_off_base = offset_bound.wrapping_add(ZSTD_REP_NUM);
|
|
if off_base as usize > offset_bound_off_base {
|
|
return ERROR(ZstdErrorCode::ExternalSequencesInvalid);
|
|
}
|
|
|
|
let match_len_lower_bound = if min_match == 3 || use_sequence_producer != 0 {
|
|
3
|
|
} else {
|
|
4
|
|
};
|
|
if match_length < match_len_lower_bound {
|
|
return ERROR(ZstdErrorCode::ExternalSequencesInvalid);
|
|
}
|
|
0
|
|
}
|
|
|
|
/// Stores one externally supplied sequence and copies its literal prefix into
|
|
/// the sequence store. This is the Rust equivalent of the C-only
|
|
/// `ZSTD_storeSeq()` helper used by the explicit-block-delimiter path.
|
|
unsafe fn store_external_sequence(
|
|
seq_store: &mut SeqStore_t,
|
|
src: *const u8,
|
|
source_offset: usize,
|
|
block_size: usize,
|
|
lit_length: usize,
|
|
off_base: u32,
|
|
match_length: usize,
|
|
) -> bool {
|
|
let sequence_index = unsafe { seq_store.sequences.offset_from(seq_store.sequencesStart) };
|
|
if sequence_index < 0 || sequence_index as usize >= seq_store.maxNbSeq {
|
|
return false;
|
|
}
|
|
let sequence_index = sequence_index as usize;
|
|
|
|
let sequence_size = match lit_length.checked_add(match_length) {
|
|
Some(size) => size,
|
|
None => return false,
|
|
};
|
|
let source_end = match source_offset.checked_add(sequence_size) {
|
|
Some(end) if end <= block_size => end,
|
|
_ => return false,
|
|
};
|
|
let literal_count = unsafe { seq_store.lit.offset_from(seq_store.litStart) };
|
|
if literal_count < 0
|
|
|| (literal_count as usize)
|
|
.checked_add(lit_length)
|
|
.is_none_or(|end| end > seq_store.maxNbLit)
|
|
{
|
|
return false;
|
|
}
|
|
let ml_base = match match_length.checked_sub(MINMATCH) {
|
|
Some(base) => base,
|
|
None => return false,
|
|
};
|
|
|
|
debug_assert!(source_end <= block_size);
|
|
debug_assert!(lit_length <= ZSTD_BLOCKSIZE_MAX);
|
|
debug_assert!(match_length <= ZSTD_BLOCKSIZE_MAX);
|
|
debug_assert!(seq_store.maxNbLit <= 128 << 10);
|
|
|
|
unsafe {
|
|
ptr::copy_nonoverlapping(src.add(source_offset), seq_store.lit, lit_length);
|
|
seq_store.lit = seq_store.lit.add(lit_length);
|
|
}
|
|
|
|
let sequence = unsafe { &mut *seq_store.sequences };
|
|
if lit_length > u16::MAX as usize {
|
|
debug_assert_eq!(seq_store.longLengthType, 0);
|
|
seq_store.longLengthType = ZSTD_LLT_LITERAL_LENGTH;
|
|
seq_store.longLengthPos = sequence_index as u32;
|
|
}
|
|
sequence.litLength = lit_length as u16;
|
|
sequence.offBase = off_base;
|
|
if ml_base > u16::MAX as usize {
|
|
debug_assert_eq!(seq_store.longLengthType, 0);
|
|
seq_store.longLengthType = ZSTD_LLT_MATCH_LENGTH;
|
|
seq_store.longLengthPos = sequence_index as u32;
|
|
}
|
|
sequence.mlBase = ml_base as u16;
|
|
unsafe {
|
|
seq_store.sequences = seq_store.sequences.add(1);
|
|
}
|
|
true
|
|
}
|
|
|
|
/// Transfers externally produced sequences through one explicit block
|
|
/// delimiter. The C wrapper supplies only the state extracted from
|
|
/// `ZSTD_CCtx`; the sequence-store mutation and delimiter bookkeeping live in
|
|
/// Rust. This is the Rust leaf for
|
|
/// `ZSTD_transferSequences_wBlockDelim()`.
|
|
#[allow(clippy::too_many_arguments)]
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_transferSequencesWBlockDelim(
|
|
seq_store: *mut SeqStore_t,
|
|
seq_pos: *mut ZSTD_SequencePosition,
|
|
in_seqs: *const ZSTD_Sequence,
|
|
in_seqs_size: usize,
|
|
src: *const u8,
|
|
block_size: usize,
|
|
external_rep_search: c_int,
|
|
prev_repcodes: *const u32,
|
|
next_repcodes: *mut u32,
|
|
dict_size: u32,
|
|
validate_sequences: c_int,
|
|
min_match: u32,
|
|
window_log: u32,
|
|
use_sequence_producer: c_int,
|
|
) -> usize {
|
|
if seq_store.is_null()
|
|
|| seq_pos.is_null()
|
|
|| in_seqs.is_null()
|
|
|| src.is_null()
|
|
|| prev_repcodes.is_null()
|
|
|| next_repcodes.is_null()
|
|
{
|
|
return ERROR(ZstdErrorCode::ExternalSequencesInvalid);
|
|
}
|
|
|
|
let seq_store = unsafe { &mut *seq_store };
|
|
let seq_pos = unsafe { &mut *seq_pos };
|
|
if (seq_pos.idx as usize) > in_seqs_size {
|
|
return ERROR(ZstdErrorCode::ExternalSequencesInvalid);
|
|
}
|
|
|
|
let mut updated_repcodes = [0u32; ZSTD_REP_NUM];
|
|
unsafe {
|
|
ptr::copy_nonoverlapping(prev_repcodes, updated_repcodes.as_mut_ptr(), ZSTD_REP_NUM);
|
|
}
|
|
|
|
let start_idx = seq_pos.idx as usize;
|
|
let mut idx = start_idx;
|
|
let mut source_offset = 0usize;
|
|
|
|
while idx < in_seqs_size {
|
|
let sequence = unsafe { *in_seqs.add(idx) };
|
|
if sequence.matchLength == 0 && sequence.offset == 0 {
|
|
break;
|
|
}
|
|
|
|
let lit_length = sequence.litLength as usize;
|
|
let match_length = sequence.matchLength as usize;
|
|
let off_base = if external_rep_search == ZSTD_PS_DISABLE {
|
|
if sequence.offset == 0 {
|
|
return ERROR(ZstdErrorCode::ExternalSequencesInvalid);
|
|
}
|
|
sequence.offset.wrapping_add(ZSTD_REP_NUM as u32)
|
|
} else {
|
|
let ll0 = lit_length == 0;
|
|
let off_base = unsafe {
|
|
ZSTD_rust_finalizeOffBase(
|
|
sequence.offset,
|
|
updated_repcodes.as_ptr(),
|
|
u32::from(ll0),
|
|
)
|
|
};
|
|
update_rep(&mut updated_repcodes, off_base, ll0);
|
|
off_base
|
|
};
|
|
|
|
if validate_sequences != 0 {
|
|
seq_pos.posInSrc = seq_pos
|
|
.posInSrc
|
|
.wrapping_add(lit_length.wrapping_add(match_length));
|
|
let validation = ZSTD_rust_validateSequence(
|
|
off_base,
|
|
sequence.matchLength,
|
|
min_match,
|
|
seq_pos.posInSrc,
|
|
window_log,
|
|
dict_size as usize,
|
|
use_sequence_producer,
|
|
);
|
|
if ERR_isError(validation) {
|
|
return validation;
|
|
}
|
|
}
|
|
|
|
if idx - start_idx >= seq_store.maxNbSeq {
|
|
return ERROR(ZstdErrorCode::ExternalSequencesInvalid);
|
|
}
|
|
if !unsafe {
|
|
store_external_sequence(
|
|
seq_store,
|
|
src,
|
|
source_offset,
|
|
block_size,
|
|
lit_length,
|
|
off_base,
|
|
match_length,
|
|
)
|
|
} {
|
|
return ERROR(ZstdErrorCode::ExternalSequencesInvalid);
|
|
}
|
|
source_offset = source_offset
|
|
.checked_add(lit_length.wrapping_add(match_length))
|
|
.expect("validated external sequence source offset overflow");
|
|
idx += 1;
|
|
}
|
|
|
|
if idx == in_seqs_size {
|
|
return ERROR(ZstdErrorCode::ExternalSequencesInvalid);
|
|
}
|
|
|
|
debug_assert_ne!(external_rep_search, ZSTD_PS_AUTO);
|
|
/* If repcode search was skipped while parsing, recover the raw-offset
|
|
* history once the delimiter identifies the complete block. */
|
|
if external_rep_search == ZSTD_PS_DISABLE && idx != start_idx {
|
|
let last_seq_idx = idx - 1;
|
|
if last_seq_idx >= start_idx + 2 {
|
|
updated_repcodes[2] = unsafe { (*in_seqs.add(last_seq_idx - 2)).offset };
|
|
updated_repcodes[1] = unsafe { (*in_seqs.add(last_seq_idx - 1)).offset };
|
|
updated_repcodes[0] = unsafe { (*in_seqs.add(last_seq_idx)).offset };
|
|
} else if last_seq_idx == start_idx + 1 {
|
|
updated_repcodes[2] = updated_repcodes[0];
|
|
updated_repcodes[1] = unsafe { (*in_seqs.add(last_seq_idx - 1)).offset };
|
|
updated_repcodes[0] = unsafe { (*in_seqs.add(last_seq_idx)).offset };
|
|
} else {
|
|
debug_assert_eq!(last_seq_idx, start_idx);
|
|
updated_repcodes[2] = updated_repcodes[1];
|
|
updated_repcodes[1] = updated_repcodes[0];
|
|
updated_repcodes[0] = unsafe { (*in_seqs.add(last_seq_idx)).offset };
|
|
}
|
|
}
|
|
unsafe {
|
|
ptr::copy_nonoverlapping(updated_repcodes.as_ptr(), next_repcodes, ZSTD_REP_NUM);
|
|
}
|
|
|
|
let delimiter = unsafe { *in_seqs.add(idx) };
|
|
let delimiter_literals = delimiter.litLength as usize;
|
|
if delimiter_literals != 0 {
|
|
let delimiter_end = match source_offset.checked_add(delimiter_literals) {
|
|
Some(end) if end <= block_size => end,
|
|
_ => return ERROR(ZstdErrorCode::ExternalSequencesInvalid),
|
|
};
|
|
unsafe {
|
|
ZSTD_rust_storeLastLiterals(seq_store, src.add(source_offset), delimiter_literals);
|
|
}
|
|
source_offset = delimiter_end;
|
|
seq_pos.posInSrc = seq_pos.posInSrc.wrapping_add(delimiter_literals);
|
|
}
|
|
|
|
if source_offset != block_size {
|
|
return ERROR(ZstdErrorCode::ExternalSequencesInvalid);
|
|
}
|
|
seq_pos.idx = (idx + 1) as u32;
|
|
block_size
|
|
}
|
|
|
|
/// Finds the next explicit block delimiter and returns the represented size.
|
|
///
|
|
/// The scan is half-open at `inSeqsSize`: a delimiter at the final element is
|
|
/// included, while a missing delimiter returns the C external-sequences
|
|
/// error. The delimiter's literal length is part of the block size, but its
|
|
/// match length must be zero.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_blockSizeExplicitDelimiter(
|
|
in_seqs: *const ZSTD_Sequence,
|
|
in_seqs_size: usize,
|
|
seq_idx: u32,
|
|
) -> usize {
|
|
let mut end = false;
|
|
let mut block_size = 0usize;
|
|
let mut sequence_index = seq_idx as usize;
|
|
debug_assert!(sequence_index <= in_seqs_size);
|
|
|
|
while sequence_index < in_seqs_size {
|
|
let sequence = unsafe { *in_seqs.add(sequence_index) };
|
|
end = sequence.offset == 0;
|
|
let sequence_size = sequence.litLength.wrapping_add(sequence.matchLength) as usize;
|
|
block_size = block_size.wrapping_add(sequence_size);
|
|
if end {
|
|
if sequence.matchLength != 0 {
|
|
return ERROR(ZstdErrorCode::ExternalSequencesInvalid);
|
|
}
|
|
break;
|
|
}
|
|
sequence_index += 1;
|
|
}
|
|
|
|
if !end {
|
|
return ERROR(ZstdErrorCode::ExternalSequencesInvalid);
|
|
}
|
|
block_size
|
|
}
|
|
|
|
/// Determines the next external-sequence block size.
|
|
///
|
|
/// Mode `0` is the no-delimiter mode and returns the target size capped at the
|
|
/// remaining source. Mode `1` scans an explicit delimiter and rejects blocks
|
|
/// larger than either configured block size or the remaining frame.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_determineBlockSize(
|
|
mode: c_int,
|
|
block_size: usize,
|
|
remaining: usize,
|
|
in_seqs: *const ZSTD_Sequence,
|
|
in_seqs_size: usize,
|
|
seq_idx: u32,
|
|
) -> usize {
|
|
if mode == 0 {
|
|
return remaining.min(block_size);
|
|
}
|
|
debug_assert_eq!(mode, 1);
|
|
|
|
let explicit_block_size =
|
|
unsafe { ZSTD_rust_blockSizeExplicitDelimiter(in_seqs, in_seqs_size, seq_idx) };
|
|
if ERR_isError(explicit_block_size) {
|
|
return explicit_block_size;
|
|
}
|
|
if explicit_block_size > block_size || explicit_block_size > remaining {
|
|
return ERROR(ZstdErrorCode::ExternalSequencesInvalid);
|
|
}
|
|
explicit_block_size
|
|
}
|
|
|
|
/// Validates and post-processes sequences returned by an external sequence
|
|
/// producer. This is the Rust leaf for C's
|
|
/// `ZSTD_postProcessSequenceProducerResult()`.
|
|
///
|
|
/// A sequence with zero offset and zero match length terminates a block. If
|
|
/// the producer did not append one, this function appends a zeroed delimiter
|
|
/// when capacity permits. Error-shaped producer counts and invalid empty
|
|
/// parses return the `sequenceProducer_failed` error unchanged.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_postProcessSequenceProducerResult(
|
|
out_seqs: *mut ZSTD_Sequence,
|
|
nb_external_seqs: usize,
|
|
out_seqs_capacity: usize,
|
|
src_size: usize,
|
|
) -> usize {
|
|
if nb_external_seqs > out_seqs_capacity {
|
|
return ERROR(ZstdErrorCode::SequenceProducerFailed);
|
|
}
|
|
|
|
if nb_external_seqs == 0 && src_size > 0 {
|
|
return ERROR(ZstdErrorCode::SequenceProducerFailed);
|
|
}
|
|
|
|
if src_size == 0 {
|
|
unsafe {
|
|
ptr::write_bytes(out_seqs.cast::<u8>(), 0, size_of::<ZSTD_Sequence>());
|
|
}
|
|
return 1;
|
|
}
|
|
|
|
let last_seq = unsafe { *out_seqs.add(nb_external_seqs - 1) };
|
|
|
|
if last_seq.offset == 0 && last_seq.matchLength == 0 {
|
|
return nb_external_seqs;
|
|
}
|
|
|
|
if nb_external_seqs == out_seqs_capacity {
|
|
return ERROR(ZstdErrorCode::SequenceProducerFailed);
|
|
}
|
|
|
|
unsafe {
|
|
ptr::write_bytes(
|
|
out_seqs.add(nb_external_seqs).cast::<u8>(),
|
|
0,
|
|
size_of::<ZSTD_Sequence>(),
|
|
);
|
|
}
|
|
nb_external_seqs + 1
|
|
}
|
|
|
|
/// Resets the state that is carried from one compressed block to the next.
|
|
///
|
|
/// This is the direct Rust leaf for C's
|
|
/// `ZSTD_reset_compressedBlockState()`. The caller must provide a valid
|
|
/// mutable compressed block state.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_resetCompressedBlockState(
|
|
block_state: *mut ZSTD_compressedBlockState_t,
|
|
) {
|
|
let block_state = unsafe { &mut *block_state };
|
|
block_state.rep.copy_from_slice(&REP_START_VALUE);
|
|
block_state.entropy.huf.repeatMode = HUF_REPEAT_NONE;
|
|
block_state.entropy.fse.offcode_repeatMode = FSE_REPEAT_NONE;
|
|
block_state.entropy.fse.matchlength_repeatMode = FSE_REPEAT_NONE;
|
|
block_state.entropy.fse.litlength_repeatMode = FSE_REPEAT_NONE;
|
|
}
|
|
|
|
/// Appends the final literals of a block to its sequence store.
|
|
///
|
|
/// The source and destination ranges must be valid and non-overlapping, as
|
|
/// required by the original C `memcpy` call. `last_ll_size == 0` preserves
|
|
/// the C pointer-update behavior without changing any bytes.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_storeLastLiterals(
|
|
seq_store: *mut SeqStore_t,
|
|
anchor: *const u8,
|
|
last_ll_size: usize,
|
|
) {
|
|
let seq_store = unsafe { &mut *seq_store };
|
|
unsafe {
|
|
ptr::copy_nonoverlapping(anchor, seq_store.lit, last_ll_size);
|
|
seq_store.lit = seq_store.lit.add(last_ll_size);
|
|
}
|
|
}
|
|
|
|
/// Resets the sequence and literal cursors for a new block.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_resetSeqStore(seq_store: *mut SeqStore_t) {
|
|
let seq_store = unsafe { &mut *seq_store };
|
|
seq_store.lit = seq_store.litStart;
|
|
seq_store.sequences = seq_store.sequencesStart;
|
|
seq_store.longLengthType = 0;
|
|
}
|
|
|
|
/// Returns the sum of all literal and match lengths in an external sequence
|
|
/// buffer, without looking for a block delimiter.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_fastSequenceLengthSum(
|
|
seq_buf: *const ZSTD_Sequence,
|
|
seq_buf_size: usize,
|
|
) -> usize {
|
|
let mut match_len_sum = 0usize;
|
|
let mut lit_len_sum = 0usize;
|
|
for index in 0..seq_buf_size {
|
|
let sequence = unsafe { *seq_buf.add(index) };
|
|
lit_len_sum = lit_len_sum.wrapping_add(sequence.litLength as usize);
|
|
match_len_sum = match_len_sum.wrapping_add(sequence.matchLength as usize);
|
|
}
|
|
lit_len_sum.wrapping_add(match_len_sum)
|
|
}
|
|
|
|
/// Returns whether all bytes in the input have the same value.
|
|
///
|
|
/// This keeps the original C helper's precondition that `src` points to at
|
|
/// least one readable byte, even when `length == 0`: the first byte is read
|
|
/// before the length checks.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_isRLE(src: *const u8, length: usize) -> c_int {
|
|
let value = unsafe { *src };
|
|
let value_st = (u64::from(value) * 0x0101_0101_0101_0101u64) as usize;
|
|
let unroll_size = size_of::<usize>() * 4;
|
|
let prefix_length = length & (unroll_size - 1);
|
|
|
|
if length == 1 {
|
|
return 1;
|
|
}
|
|
|
|
let mut index = 1;
|
|
while index < prefix_length {
|
|
if unsafe { *src.add(index) } != value {
|
|
return 0;
|
|
}
|
|
index += 1;
|
|
}
|
|
|
|
let mut index = prefix_length;
|
|
while index != length {
|
|
let mut unroll_offset = 0;
|
|
while unroll_offset < unroll_size {
|
|
let word = unsafe { MEM_readST(src.add(index + unroll_offset).cast()) };
|
|
if word != value_st {
|
|
return 0;
|
|
}
|
|
unroll_offset += size_of::<usize>();
|
|
}
|
|
index += unroll_size;
|
|
}
|
|
1
|
|
}
|
|
|
|
/// Heuristic used to decide whether a sequence store may describe an RLE
|
|
/// block. It intentionally only inspects the two pointer distances.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_maybeRLE(seq_store: *const SeqStore_t) -> c_int {
|
|
let seq_store = unsafe { &*seq_store };
|
|
let nb_seqs = unsafe { seq_store.sequences.offset_from(seq_store.sequencesStart) } as usize;
|
|
let nb_lits = unsafe { seq_store.lit.offset_from(seq_store.litStart) } as usize;
|
|
if nb_seqs < 4 && nb_lits < 10 {
|
|
1
|
|
} else {
|
|
0
|
|
}
|
|
}
|
|
|
|
/// Returns the literal bytes represented by a sequence store.
|
|
///
|
|
/// The stored `u16` literal length may represent one long literal through the
|
|
/// sequence store's side-band marker. This is the Rust leaf for C's
|
|
/// `ZSTD_countSeqStoreLiteralsBytes()` helper.
|
|
#[inline]
|
|
unsafe fn count_seq_store_literals_bytes(seq_store: *const SeqStore_t) -> usize {
|
|
let seq_store = unsafe { &*seq_store };
|
|
let nb_seqs = unsafe { seq_store.sequences.offset_from(seq_store.sequencesStart) } as usize;
|
|
let mut literals_bytes = 0usize;
|
|
for index in 0..nb_seqs {
|
|
let sequence = unsafe { *seq_store.sequencesStart.add(index) };
|
|
literals_bytes = literals_bytes.wrapping_add(sequence.litLength as usize);
|
|
if index == seq_store.longLengthPos as usize
|
|
&& seq_store.longLengthType == ZSTD_LLT_LITERAL_LENGTH
|
|
{
|
|
literals_bytes = literals_bytes.wrapping_add(0x10000);
|
|
}
|
|
}
|
|
literals_bytes
|
|
}
|
|
|
|
/// Returns the match bytes represented by a sequence store.
|
|
///
|
|
/// `mlBase` stores `matchLength - MINMATCH`; the side-band long-length marker
|
|
/// adds the same `0x10000` extension used by the original C helper.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_countSeqStoreLiteralsBytes(
|
|
seq_store: *const SeqStore_t,
|
|
) -> usize {
|
|
unsafe { count_seq_store_literals_bytes(seq_store) }
|
|
}
|
|
|
|
#[inline]
|
|
unsafe fn count_seq_store_match_bytes(seq_store: *const SeqStore_t) -> usize {
|
|
let seq_store = unsafe { &*seq_store };
|
|
let nb_seqs = unsafe { seq_store.sequences.offset_from(seq_store.sequencesStart) } as usize;
|
|
let mut match_bytes = 0usize;
|
|
for index in 0..nb_seqs {
|
|
let sequence = unsafe { *seq_store.sequencesStart.add(index) };
|
|
match_bytes = match_bytes.wrapping_add(sequence.mlBase as usize + MINMATCH);
|
|
if index == seq_store.longLengthPos as usize
|
|
&& seq_store.longLengthType == ZSTD_LLT_MATCH_LENGTH
|
|
{
|
|
match_bytes = match_bytes.wrapping_add(0x10000);
|
|
}
|
|
}
|
|
match_bytes
|
|
}
|
|
|
|
/// Returns the match bytes represented by a sequence store.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_countSeqStoreMatchBytes(seq_store: *const SeqStore_t) -> usize {
|
|
unsafe { count_seq_store_match_bytes(seq_store) }
|
|
}
|
|
|
|
/// Derives the sequence-store view for the half-open sequence range
|
|
/// `[start_idx, end_idx)`.
|
|
///
|
|
/// This is the Rust leaf for C's `ZSTD_deriveSeqStoreChunk()`. The original
|
|
/// store remains untouched; the result is a shallow copy whose pointers refer
|
|
/// into the original sequence, literal, and code buffers.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_deriveSeqStoreChunk(
|
|
result_seq_store: *mut SeqStore_t,
|
|
original_seq_store: *const SeqStore_t,
|
|
start_idx: usize,
|
|
end_idx: usize,
|
|
) {
|
|
let original_seq_store = unsafe { &*original_seq_store };
|
|
let result_seq_store = unsafe { &mut *result_seq_store };
|
|
*result_seq_store = unsafe { ptr::read(original_seq_store) };
|
|
|
|
if start_idx > 0 {
|
|
result_seq_store.sequences = unsafe { original_seq_store.sequencesStart.add(start_idx) };
|
|
let literals_bytes =
|
|
unsafe { count_seq_store_literals_bytes(std::ptr::addr_of!(*result_seq_store)) };
|
|
result_seq_store.litStart = unsafe { result_seq_store.litStart.add(literals_bytes) };
|
|
}
|
|
|
|
/* Move longLengthPos into the correct position if necessary. */
|
|
if original_seq_store.longLengthType != 0 {
|
|
if original_seq_store.longLengthPos as usize > end_idx
|
|
|| (original_seq_store.longLengthPos as usize) < start_idx
|
|
{
|
|
result_seq_store.longLengthType = 0;
|
|
} else {
|
|
result_seq_store.longLengthPos = original_seq_store
|
|
.longLengthPos
|
|
.wrapping_sub(start_idx as u32);
|
|
}
|
|
}
|
|
result_seq_store.sequencesStart = unsafe { original_seq_store.sequencesStart.add(start_idx) };
|
|
result_seq_store.sequences = unsafe { original_seq_store.sequencesStart.add(end_idx) };
|
|
|
|
let original_nb_sequences = unsafe {
|
|
original_seq_store
|
|
.sequences
|
|
.offset_from(original_seq_store.sequencesStart)
|
|
} as usize;
|
|
if end_idx == original_nb_sequences {
|
|
/* This accounts for possible last literals at the end of the block. */
|
|
debug_assert_eq!(result_seq_store.lit, original_seq_store.lit);
|
|
} else {
|
|
let literals_bytes =
|
|
unsafe { count_seq_store_literals_bytes(std::ptr::addr_of!(*result_seq_store)) };
|
|
result_seq_store.lit = unsafe { result_seq_store.litStart.add(literals_bytes) };
|
|
}
|
|
result_seq_store.llCode = unsafe { result_seq_store.llCode.add(start_idx) };
|
|
result_seq_store.mlCode = unsafe { result_seq_store.mlCode.add(start_idx) };
|
|
result_seq_store.ofCode = unsafe { result_seq_store.ofCode.add(start_idx) };
|
|
}
|
|
|
|
/// Resolves a stored repcode against the current raw-offset history.
|
|
///
|
|
/// This is the Rust leaf for C's `ZSTD_resolveRepcodeToRawOffset()`. The
|
|
/// caller must pass an `off_base` in the repcode range `1..=ZSTD_REP_NUM` and
|
|
/// an `ll0` value of zero or one.
|
|
#[inline]
|
|
fn resolve_repcode_to_raw_offset(rep: &[u32; ZSTD_REP_NUM], off_base: u32, ll0: u32) -> u32 {
|
|
debug_assert!((1..=ZSTD_REP_NUM as u32).contains(&off_base));
|
|
let adjusted_rep_code = off_base.wrapping_sub(1).wrapping_add(ll0);
|
|
if adjusted_rep_code == ZSTD_REP_NUM as u32 {
|
|
debug_assert_ne!(ll0, 0);
|
|
rep[0].wrapping_sub(1)
|
|
} else {
|
|
rep[adjusted_rep_code as usize]
|
|
}
|
|
}
|
|
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_resolveRepcodeToRawOffset(
|
|
rep: *const u32,
|
|
off_base: u32,
|
|
ll0: u32,
|
|
) -> u32 {
|
|
let rep = unsafe { &*rep.cast::<[u32; ZSTD_REP_NUM]>() };
|
|
resolve_repcode_to_raw_offset(rep, off_base, ll0)
|
|
}
|
|
|
|
/// Reconciles decompression and compression repcode histories for a sequence
|
|
/// store. The two histories remain owned by the C caller; only their scalar
|
|
/// arrays and the ABI-compatible sequence store cross the language boundary.
|
|
///
|
|
/// A repcode mismatch is materialized as a full offset in the sequence store,
|
|
/// while the compression history continues to follow the original offBase.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_seqStore_resolveOffCodes(
|
|
d_rep: *mut u32,
|
|
c_rep: *mut u32,
|
|
seq_store: *const SeqStore_t,
|
|
nb_seq: u32,
|
|
) {
|
|
let d_rep = unsafe { &mut *d_rep.cast::<[u32; ZSTD_REP_NUM]>() };
|
|
let c_rep = unsafe { &mut *c_rep.cast::<[u32; ZSTD_REP_NUM]>() };
|
|
let seq_store = unsafe { &*seq_store };
|
|
let long_lit_len_idx = if seq_store.longLengthType == ZSTD_LLT_LITERAL_LENGTH {
|
|
seq_store.longLengthPos
|
|
} else {
|
|
nb_seq
|
|
};
|
|
|
|
for index in 0..nb_seq {
|
|
let seq = unsafe { &mut *seq_store.sequencesStart.add(index as usize) };
|
|
let ll0 = u32::from(seq.litLength == 0 && index != long_lit_len_idx);
|
|
let off_base = seq.offBase;
|
|
debug_assert!(off_base > 0);
|
|
|
|
if (1..=ZSTD_REP_NUM as u32).contains(&off_base) {
|
|
let d_raw_offset = resolve_repcode_to_raw_offset(d_rep, off_base, ll0);
|
|
let c_raw_offset = resolve_repcode_to_raw_offset(c_rep, off_base, ll0);
|
|
/* Keep the decoder's simulated history aligned with the raw
|
|
* offset represented by the compression-side history. */
|
|
if d_raw_offset != c_raw_offset {
|
|
seq.offBase = c_raw_offset.wrapping_add(ZSTD_REP_NUM as u32);
|
|
}
|
|
}
|
|
|
|
/* The decoder sees the corrected sequence; compression follows the
|
|
* original sequence exactly. */
|
|
update_rep(d_rep, seq.offBase, ll0 != 0);
|
|
update_rep(c_rep, off_base, ll0 != 0);
|
|
}
|
|
}
|
|
|
|
/// C's `SeqCollector` leaf from `zstd_compress_internal.h`.
|
|
#[repr(C)]
|
|
pub struct SeqCollector {
|
|
pub collectSequences: c_int,
|
|
pub seqStart: *mut ZSTD_Sequence,
|
|
pub seqIndex: usize,
|
|
pub maxSequences: usize,
|
|
}
|
|
|
|
/// Result of `build_sequences_statistics()`; mirrors C's
|
|
/// `ZSTD_symbolEncodingTypeStats_t`, which no longer crosses the boundary.
|
|
pub(crate) struct SymbolEncodingTypeStats {
|
|
pub ll_type: u32,
|
|
pub off_type: u32,
|
|
pub ml_type: u32,
|
|
pub size: usize,
|
|
/// Accounts for a bug in zstd 1.3.4; see
|
|
/// `entropy_compress_seq_store_internal()`.
|
|
pub last_count_size: usize,
|
|
pub long_offsets: c_int,
|
|
}
|
|
|
|
/// C's `ZSTD_updateRep()` repeat-offset history update.
|
|
#[inline]
|
|
pub(crate) fn update_rep(reps: &mut [u32; 3], off_base: u32, literal_length_is_zero: bool) {
|
|
if off_base > 3 {
|
|
reps[2] = reps[1];
|
|
reps[1] = reps[0];
|
|
reps[0] = off_base - 3;
|
|
return;
|
|
}
|
|
let rep_code = off_base - 1 + u32::from(literal_length_is_zero);
|
|
if rep_code == 0 {
|
|
return;
|
|
}
|
|
let current_offset = if rep_code == 3 {
|
|
reps[0].wrapping_sub(1)
|
|
} else {
|
|
reps[rep_code as usize]
|
|
};
|
|
reps[2] = if rep_code >= 2 { reps[1] } else { reps[2] };
|
|
reps[1] = reps[0];
|
|
reps[0] = current_offset;
|
|
}
|
|
|
|
#[inline]
|
|
fn ll_code(lit_length: u32) -> u32 {
|
|
const LL_CODE: [u8; 64] = [
|
|
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 16, 17, 17, 18, 18, 19, 19, 20,
|
|
20, 20, 20, 21, 21, 21, 21, 22, 22, 22, 22, 22, 22, 22, 22, 23, 23, 23, 23, 23, 23, 23, 23,
|
|
24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24, 24,
|
|
];
|
|
if lit_length > 63 {
|
|
ZSTD_highbit32(lit_length) + LL_DELTA_CODE
|
|
} else {
|
|
LL_CODE[lit_length as usize] as u32
|
|
}
|
|
}
|
|
|
|
#[inline]
|
|
fn ml_code(ml_base: u32) -> u32 {
|
|
const ML_CODE: [u8; 128] = [
|
|
0, 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, 32, 33, 33, 34, 34, 35, 35, 36, 36, 36, 36, 37, 37, 37, 37,
|
|
38, 38, 38, 38, 38, 38, 38, 38, 39, 39, 39, 39, 39, 39, 39, 39, 40, 40, 40, 40, 40, 40, 40,
|
|
40, 40, 40, 40, 40, 40, 40, 40, 40, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,
|
|
41, 41, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,
|
|
42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,
|
|
];
|
|
if ml_base > 127 {
|
|
ZSTD_highbit32(ml_base) + ML_DELTA_CODE
|
|
} else {
|
|
ML_CODE[ml_base as usize] as u32
|
|
}
|
|
}
|
|
|
|
#[inline]
|
|
unsafe fn copy_fse_tables(dst: *mut ZSTD_fseCTables_t, src: *const ZSTD_fseCTables_t) {
|
|
unsafe {
|
|
ptr::copy_nonoverlapping(
|
|
src.cast::<u8>(),
|
|
dst.cast::<u8>(),
|
|
size_of::<ZSTD_fseCTables_t>(),
|
|
)
|
|
};
|
|
}
|
|
|
|
/// Converts the stored lengths and distances into symbol codes.
|
|
///
|
|
/// Retains its original exported name: the C dictionary builder and the
|
|
/// decodecorpus test tool link against it directly.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_seqToCodes(seq_store: *const SeqStore_t) -> c_int {
|
|
let store = unsafe { &*seq_store };
|
|
let sequences = store.sequencesStart;
|
|
let nb_seq = unsafe { store.sequences.offset_from(sequences) } as usize;
|
|
let mut long_offsets = 0;
|
|
debug_assert!(nb_seq <= store.maxNbSeq);
|
|
|
|
for index in 0..nb_seq {
|
|
let sequence = unsafe { sequences.add(index).read() };
|
|
let of_code = ZSTD_highbit32(sequence.offBase);
|
|
unsafe {
|
|
*store.llCode.add(index) = ll_code(sequence.litLength as u32) as u8;
|
|
*store.ofCode.add(index) = of_code as u8;
|
|
*store.mlCode.add(index) = ml_code(sequence.mlBase as u32) as u8;
|
|
}
|
|
debug_assert!(MEM_32bits() || of_code < STREAM_ACCUMULATOR_MIN_64);
|
|
if MEM_32bits() && of_code >= STREAM_ACCUMULATOR_MIN_32 {
|
|
long_offsets = 1;
|
|
}
|
|
}
|
|
if store.longLengthType == ZSTD_LLT_LITERAL_LENGTH {
|
|
unsafe { *store.llCode.add(store.longLengthPos as usize) = MAX_LL as u8 };
|
|
}
|
|
if store.longLengthType == ZSTD_LLT_MATCH_LENGTH {
|
|
unsafe { *store.mlCode.add(store.longLengthPos as usize) = MAX_ML as u8 };
|
|
}
|
|
long_offsets
|
|
}
|
|
|
|
/// C's `ZSTD_buildSequencesStatistics()`.
|
|
///
|
|
/// Returns the statistics with a zstd error code in the `size` field on
|
|
/// failure, and updates `next_entropy` as a side effect. `nb_seq` must be
|
|
/// greater than zero, and the entropy workspace must hold at least
|
|
/// `ENTROPY_WORKSPACE_SIZE - (MaxSeq + 1) * sizeof(U32)` bytes.
|
|
#[allow(clippy::too_many_arguments)]
|
|
pub(crate) unsafe fn build_sequences_statistics(
|
|
seq_store: *const SeqStore_t,
|
|
nb_seq: usize,
|
|
prev_entropy: *const ZSTD_fseCTables_t,
|
|
next_entropy: *mut ZSTD_fseCTables_t,
|
|
dst: *mut u8,
|
|
dst_capacity: usize,
|
|
strategy: c_int,
|
|
count_workspace: *mut c_uint,
|
|
entropy_workspace: *mut c_void,
|
|
entropy_wksp_size: usize,
|
|
) -> SymbolEncodingTypeStats {
|
|
let store = unsafe { &*seq_store };
|
|
let mut stats = SymbolEncodingTypeStats {
|
|
ll_type: 0,
|
|
off_type: 0,
|
|
ml_type: 0,
|
|
size: 0,
|
|
last_count_size: 0,
|
|
long_offsets: 0,
|
|
};
|
|
let mut written = 0usize;
|
|
|
|
/* convert length/distances into codes */
|
|
stats.long_offsets = unsafe { ZSTD_seqToCodes(seq_store) };
|
|
debug_assert!(nb_seq != 0); /* ZSTD_selectEncodingType() divides by nb_seq */
|
|
|
|
/* build CTable for Literal Lengths */
|
|
{
|
|
let mut max = MAX_LL as c_uint;
|
|
let most_frequent = unsafe {
|
|
HIST_countFast_wksp(
|
|
count_workspace,
|
|
&mut max,
|
|
store.llCode.cast::<c_void>(),
|
|
nb_seq,
|
|
entropy_workspace,
|
|
entropy_wksp_size,
|
|
)
|
|
}; /* can't fail */
|
|
unsafe { (*next_entropy).litlength_repeatMode = (*prev_entropy).litlength_repeatMode };
|
|
stats.ll_type = unsafe {
|
|
ZSTD_selectEncodingType(
|
|
ptr::addr_of_mut!((*next_entropy).litlength_repeatMode),
|
|
count_workspace,
|
|
max,
|
|
most_frequent,
|
|
nb_seq,
|
|
LL_FSE_LOG as c_uint,
|
|
ptr::addr_of!((*prev_entropy).litlengthCTable).cast::<u32>(),
|
|
LL_DEFAULT_NORM.as_ptr(),
|
|
LL_DEFAULT_NORM_LOG,
|
|
ZSTD_DEFAULT_ALLOWED,
|
|
strategy,
|
|
)
|
|
} as u32;
|
|
let count_size = unsafe {
|
|
ZSTD_buildCTable(
|
|
dst.add(written).cast::<c_void>(),
|
|
dst_capacity - written,
|
|
ptr::addr_of_mut!((*next_entropy).litlengthCTable).cast::<u32>(),
|
|
LL_FSE_LOG as u32,
|
|
stats.ll_type as c_int,
|
|
count_workspace,
|
|
max,
|
|
store.llCode,
|
|
nb_seq,
|
|
LL_DEFAULT_NORM.as_ptr(),
|
|
LL_DEFAULT_NORM_LOG,
|
|
MAX_LL as u32,
|
|
ptr::addr_of!((*prev_entropy).litlengthCTable).cast::<u32>(),
|
|
size_of::<[u32; LL_CTABLE_SIZE]>(),
|
|
entropy_workspace,
|
|
entropy_wksp_size,
|
|
)
|
|
};
|
|
if ERR_isError(count_size) {
|
|
stats.size = count_size;
|
|
return stats;
|
|
}
|
|
if stats.ll_type == SET_COMPRESSED as u32 {
|
|
stats.last_count_size = count_size;
|
|
}
|
|
written += count_size;
|
|
}
|
|
/* build CTable for Offsets */
|
|
{
|
|
let mut max = MAX_OFF as c_uint;
|
|
let most_frequent = unsafe {
|
|
HIST_countFast_wksp(
|
|
count_workspace,
|
|
&mut max,
|
|
store.ofCode.cast::<c_void>(),
|
|
nb_seq,
|
|
entropy_workspace,
|
|
entropy_wksp_size,
|
|
)
|
|
}; /* can't fail */
|
|
/* We can only use the basic table if max <= DefaultMaxOff, otherwise the offsets are too large */
|
|
let default_policy = if max <= DEFAULT_MAX_OFF as c_uint {
|
|
ZSTD_DEFAULT_ALLOWED
|
|
} else {
|
|
ZSTD_DEFAULT_DISALLOWED
|
|
};
|
|
unsafe { (*next_entropy).offcode_repeatMode = (*prev_entropy).offcode_repeatMode };
|
|
stats.off_type = unsafe {
|
|
ZSTD_selectEncodingType(
|
|
ptr::addr_of_mut!((*next_entropy).offcode_repeatMode),
|
|
count_workspace,
|
|
max,
|
|
most_frequent,
|
|
nb_seq,
|
|
OFF_FSE_LOG as c_uint,
|
|
ptr::addr_of!((*prev_entropy).offcodeCTable).cast::<u32>(),
|
|
OF_DEFAULT_NORM.as_ptr(),
|
|
OF_DEFAULT_NORM_LOG,
|
|
default_policy,
|
|
strategy,
|
|
)
|
|
} as u32;
|
|
let count_size = unsafe {
|
|
ZSTD_buildCTable(
|
|
dst.add(written).cast::<c_void>(),
|
|
dst_capacity - written,
|
|
ptr::addr_of_mut!((*next_entropy).offcodeCTable).cast::<u32>(),
|
|
OFF_FSE_LOG as u32,
|
|
stats.off_type as c_int,
|
|
count_workspace,
|
|
max,
|
|
store.ofCode,
|
|
nb_seq,
|
|
OF_DEFAULT_NORM.as_ptr(),
|
|
OF_DEFAULT_NORM_LOG,
|
|
DEFAULT_MAX_OFF as u32,
|
|
ptr::addr_of!((*prev_entropy).offcodeCTable).cast::<u32>(),
|
|
size_of::<[u32; OFF_CTABLE_SIZE]>(),
|
|
entropy_workspace,
|
|
entropy_wksp_size,
|
|
)
|
|
};
|
|
if ERR_isError(count_size) {
|
|
stats.size = count_size;
|
|
return stats;
|
|
}
|
|
if stats.off_type == SET_COMPRESSED as u32 {
|
|
stats.last_count_size = count_size;
|
|
}
|
|
written += count_size;
|
|
}
|
|
/* build CTable for MatchLengths */
|
|
{
|
|
let mut max = MAX_ML as c_uint;
|
|
let most_frequent = unsafe {
|
|
HIST_countFast_wksp(
|
|
count_workspace,
|
|
&mut max,
|
|
store.mlCode.cast::<c_void>(),
|
|
nb_seq,
|
|
entropy_workspace,
|
|
entropy_wksp_size,
|
|
)
|
|
}; /* can't fail */
|
|
unsafe { (*next_entropy).matchlength_repeatMode = (*prev_entropy).matchlength_repeatMode };
|
|
stats.ml_type = unsafe {
|
|
ZSTD_selectEncodingType(
|
|
ptr::addr_of_mut!((*next_entropy).matchlength_repeatMode),
|
|
count_workspace,
|
|
max,
|
|
most_frequent,
|
|
nb_seq,
|
|
ML_FSE_LOG as c_uint,
|
|
ptr::addr_of!((*prev_entropy).matchlengthCTable).cast::<u32>(),
|
|
ML_DEFAULT_NORM.as_ptr(),
|
|
ML_DEFAULT_NORM_LOG,
|
|
ZSTD_DEFAULT_ALLOWED,
|
|
strategy,
|
|
)
|
|
} as u32;
|
|
let count_size = unsafe {
|
|
ZSTD_buildCTable(
|
|
dst.add(written).cast::<c_void>(),
|
|
dst_capacity - written,
|
|
ptr::addr_of_mut!((*next_entropy).matchlengthCTable).cast::<u32>(),
|
|
ML_FSE_LOG as u32,
|
|
stats.ml_type as c_int,
|
|
count_workspace,
|
|
max,
|
|
store.mlCode,
|
|
nb_seq,
|
|
ML_DEFAULT_NORM.as_ptr(),
|
|
ML_DEFAULT_NORM_LOG,
|
|
MAX_ML as u32,
|
|
ptr::addr_of!((*prev_entropy).matchlengthCTable).cast::<u32>(),
|
|
size_of::<[u32; ML_CTABLE_SIZE]>(),
|
|
entropy_workspace,
|
|
entropy_wksp_size,
|
|
)
|
|
};
|
|
if ERR_isError(count_size) {
|
|
stats.size = count_size;
|
|
return stats;
|
|
}
|
|
if stats.ml_type == SET_COMPRESSED as u32 {
|
|
stats.last_count_size = count_size;
|
|
}
|
|
written += count_size;
|
|
}
|
|
stats.size = written;
|
|
stats
|
|
}
|
|
|
|
/// C's `ZSTD_entropyCompressSeqStore_internal()`: compresses literals and
|
|
/// sequences into one compressed-block body.
|
|
#[allow(clippy::too_many_arguments)]
|
|
unsafe fn entropy_compress_seq_store_internal(
|
|
dst: *mut c_void,
|
|
dst_capacity: usize,
|
|
literals: *const c_void,
|
|
lit_size: usize,
|
|
seq_store: *const SeqStore_t,
|
|
prev_entropy: *const ZSTD_entropyCTables_t,
|
|
next_entropy: *mut ZSTD_entropyCTables_t,
|
|
strategy: c_int,
|
|
disable_literal_compression: c_int,
|
|
entropy_workspace: *mut c_void,
|
|
entropy_wksp_size: usize,
|
|
bmi2: c_int,
|
|
) -> usize {
|
|
let count = entropy_workspace.cast::<c_uint>();
|
|
let entropy_workspace = unsafe { count.add(MAX_SEQ + 1) }.cast::<c_void>();
|
|
let entropy_wksp_size = entropy_wksp_size.wrapping_sub((MAX_SEQ + 1) * size_of::<c_uint>());
|
|
let store = unsafe { &*seq_store };
|
|
let sequences = store.sequencesStart;
|
|
let nb_seq = unsafe { store.sequences.offset_from(sequences) } as usize;
|
|
let ostart = dst.cast::<u8>();
|
|
let mut written = 0usize;
|
|
let last_count_size;
|
|
let long_offsets;
|
|
|
|
/* Compress literals */
|
|
{
|
|
/* Base suspicion of uncompressibility on ratio of literals to sequences */
|
|
let suspect_uncompressible =
|
|
c_int::from(nb_seq == 0 || lit_size / nb_seq >= SUSPECT_UNCOMPRESSIBLE_LITERAL_RATIO);
|
|
let c_size = unsafe {
|
|
ZSTD_compressLiterals(
|
|
dst,
|
|
dst_capacity,
|
|
literals,
|
|
lit_size,
|
|
entropy_workspace,
|
|
entropy_wksp_size,
|
|
ptr::addr_of!((*prev_entropy).huf),
|
|
ptr::addr_of_mut!((*next_entropy).huf),
|
|
strategy,
|
|
disable_literal_compression,
|
|
suspect_uncompressible,
|
|
bmi2,
|
|
)
|
|
};
|
|
if ERR_isError(c_size) {
|
|
return c_size;
|
|
}
|
|
debug_assert!(c_size <= dst_capacity);
|
|
written += c_size;
|
|
}
|
|
|
|
/* Sequences Header */
|
|
if (dst_capacity as isize).wrapping_sub(written as isize) < 3 /* max nbSeq Size */ + 1
|
|
/* seqHead */
|
|
{
|
|
return ERROR(ZstdErrorCode::DstSizeTooSmall);
|
|
}
|
|
unsafe {
|
|
if nb_seq < 128 {
|
|
*ostart.add(written) = nb_seq as u8;
|
|
written += 1;
|
|
} else if nb_seq < LONGNBSEQ as usize {
|
|
*ostart.add(written) = ((nb_seq >> 8) + 0x80) as u8;
|
|
*ostart.add(written + 1) = nb_seq as u8;
|
|
written += 2;
|
|
} else {
|
|
*ostart.add(written) = 0xFF;
|
|
MEM_writeLE16(
|
|
ostart.add(written + 1).cast::<c_void>(),
|
|
(nb_seq - LONGNBSEQ as usize) as u16,
|
|
);
|
|
written += 3;
|
|
}
|
|
}
|
|
if nb_seq == 0 {
|
|
/* Copy the old tables over as if we repeated them */
|
|
unsafe {
|
|
copy_fse_tables(
|
|
ptr::addr_of_mut!((*next_entropy).fse),
|
|
ptr::addr_of!((*prev_entropy).fse),
|
|
)
|
|
};
|
|
return written;
|
|
}
|
|
{
|
|
let seq_head = written;
|
|
written += 1;
|
|
/* build stats for sequences */
|
|
let stats = unsafe {
|
|
build_sequences_statistics(
|
|
seq_store,
|
|
nb_seq,
|
|
ptr::addr_of!((*prev_entropy).fse),
|
|
ptr::addr_of_mut!((*next_entropy).fse),
|
|
ostart.add(written),
|
|
dst_capacity - written,
|
|
strategy,
|
|
count,
|
|
entropy_workspace,
|
|
entropy_wksp_size,
|
|
)
|
|
};
|
|
if ERR_isError(stats.size) {
|
|
return stats.size;
|
|
}
|
|
unsafe {
|
|
*ostart.add(seq_head) =
|
|
((stats.ll_type << 6) + (stats.off_type << 4) + (stats.ml_type << 2)) as u8;
|
|
}
|
|
last_count_size = stats.last_count_size;
|
|
written += stats.size;
|
|
long_offsets = stats.long_offsets;
|
|
}
|
|
|
|
{
|
|
let bitstream_size = unsafe {
|
|
ZSTD_encodeSequences(
|
|
ostart.add(written).cast::<c_void>(),
|
|
dst_capacity - written,
|
|
ptr::addr_of!((*next_entropy).fse.matchlengthCTable).cast::<u32>(),
|
|
store.mlCode,
|
|
ptr::addr_of!((*next_entropy).fse.offcodeCTable).cast::<u32>(),
|
|
store.ofCode,
|
|
ptr::addr_of!((*next_entropy).fse.litlengthCTable).cast::<u32>(),
|
|
store.llCode,
|
|
sequences,
|
|
nb_seq,
|
|
long_offsets,
|
|
bmi2,
|
|
)
|
|
};
|
|
if ERR_isError(bitstream_size) {
|
|
return bitstream_size;
|
|
}
|
|
written += bitstream_size;
|
|
/* zstd versions <= 1.3.4 mistakenly report corruption when
|
|
* FSE_readNCount() receives a buffer < 4 bytes.
|
|
* Fixed by https://github.com/facebook/zstd/pull/1146.
|
|
* This can happen when the last set_compressed table present is 2
|
|
* bytes and the bitstream is only one byte.
|
|
* In this exceedingly rare case, we will simply emit an uncompressed
|
|
* block, since it isn't worth optimizing.
|
|
*/
|
|
if last_count_size != 0 && last_count_size + bitstream_size < 4 {
|
|
/* last_count_size >= 2 && bitstream_size > 0 ==> last_count_size == 3 */
|
|
debug_assert!(last_count_size + bitstream_size == 3);
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
written
|
|
}
|
|
|
|
/// C's `ZSTD_entropyCompressSeqStore_wExtLitBuffer()`: adds the
|
|
/// incompressibility fallbacks around the internal entropy pass.
|
|
#[allow(clippy::too_many_arguments)]
|
|
unsafe fn entropy_compress_seq_store_w_ext_lit_buffer(
|
|
dst: *mut c_void,
|
|
dst_capacity: usize,
|
|
literals: *const c_void,
|
|
lit_size: usize,
|
|
block_size: usize,
|
|
seq_store: *const SeqStore_t,
|
|
prev_entropy: *const ZSTD_entropyCTables_t,
|
|
next_entropy: *mut ZSTD_entropyCTables_t,
|
|
strategy: c_int,
|
|
disable_literal_compression: c_int,
|
|
entropy_workspace: *mut c_void,
|
|
entropy_wksp_size: usize,
|
|
bmi2: c_int,
|
|
) -> usize {
|
|
let c_size = unsafe {
|
|
entropy_compress_seq_store_internal(
|
|
dst,
|
|
dst_capacity,
|
|
literals,
|
|
lit_size,
|
|
seq_store,
|
|
prev_entropy,
|
|
next_entropy,
|
|
strategy,
|
|
disable_literal_compression,
|
|
entropy_workspace,
|
|
entropy_wksp_size,
|
|
bmi2,
|
|
)
|
|
};
|
|
if c_size == 0 {
|
|
return 0;
|
|
}
|
|
/* When srcSize <= dstCapacity, there is enough space to write a raw uncompressed block.
|
|
* Since we ran out of space, block must be not compressible, so fall back to raw uncompressed block.
|
|
*/
|
|
if c_size == ERROR(ZstdErrorCode::DstSizeTooSmall) && block_size <= dst_capacity {
|
|
return 0; /* block not compressed */
|
|
}
|
|
if ERR_isError(c_size) {
|
|
return c_size;
|
|
}
|
|
|
|
/* Check compressibility */
|
|
let max_c_size = block_size.wrapping_sub(min_gain(block_size, strategy));
|
|
if c_size >= max_c_size {
|
|
return 0; /* block not compressed */
|
|
}
|
|
c_size
|
|
}
|
|
|
|
/// C ABI wrapper for the external-literal-buffer entropy pass; the C shim
|
|
/// keeps the `ZSTD_entropyCompressSeqStore_internal()` name and extracts the
|
|
/// two `ZSTD_CCtx_params` scalars.
|
|
#[allow(clippy::too_many_arguments)]
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_entropyCompressSeqStore_internal(
|
|
dst: *mut c_void,
|
|
dst_capacity: usize,
|
|
literals: *const c_void,
|
|
lit_size: usize,
|
|
seq_store: *const SeqStore_t,
|
|
prev_entropy: *const ZSTD_entropyCTables_t,
|
|
next_entropy: *mut ZSTD_entropyCTables_t,
|
|
strategy: c_int,
|
|
disable_literal_compression: c_int,
|
|
entropy_workspace: *mut c_void,
|
|
entropy_wksp_size: usize,
|
|
bmi2: c_int,
|
|
) -> usize {
|
|
unsafe {
|
|
entropy_compress_seq_store_internal(
|
|
dst,
|
|
dst_capacity,
|
|
literals,
|
|
lit_size,
|
|
seq_store,
|
|
prev_entropy,
|
|
next_entropy,
|
|
strategy,
|
|
disable_literal_compression,
|
|
entropy_workspace,
|
|
entropy_wksp_size,
|
|
bmi2,
|
|
)
|
|
}
|
|
}
|
|
|
|
/// C ABI implementation behind the `ZSTD_entropyCompressSeqStore()` shim.
|
|
#[allow(clippy::too_many_arguments)]
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_entropyCompressSeqStore(
|
|
seq_store: *const SeqStore_t,
|
|
prev_entropy: *const ZSTD_entropyCTables_t,
|
|
next_entropy: *mut ZSTD_entropyCTables_t,
|
|
strategy: c_int,
|
|
disable_literal_compression: c_int,
|
|
dst: *mut c_void,
|
|
dst_capacity: usize,
|
|
src_size: usize,
|
|
entropy_workspace: *mut c_void,
|
|
entropy_wksp_size: usize,
|
|
bmi2: c_int,
|
|
) -> usize {
|
|
let store = unsafe { &*seq_store };
|
|
let lit_size = unsafe { store.lit.offset_from(store.litStart) } as usize;
|
|
unsafe {
|
|
entropy_compress_seq_store_w_ext_lit_buffer(
|
|
dst,
|
|
dst_capacity,
|
|
store.litStart.cast::<c_void>(),
|
|
lit_size,
|
|
src_size,
|
|
seq_store,
|
|
prev_entropy,
|
|
next_entropy,
|
|
strategy,
|
|
disable_literal_compression,
|
|
entropy_workspace,
|
|
entropy_wksp_size,
|
|
bmi2,
|
|
)
|
|
}
|
|
}
|
|
|
|
/// C's `ZSTD_buildBlockEntropyStats_literals()`.
|
|
///
|
|
/// Stores the literals block type and Huffman description table into
|
|
/// `huf_metadata` and returns the description size, or an error code.
|
|
#[allow(clippy::too_many_arguments)]
|
|
unsafe fn build_block_entropy_stats_literals(
|
|
src: *const c_void,
|
|
src_size: usize,
|
|
prev_huf: *const ZSTD_hufCTables_t,
|
|
next_huf: *mut ZSTD_hufCTables_t,
|
|
huf_metadata: *mut ZSTD_hufCTablesMetadata_t,
|
|
literals_compression_is_disabled: c_int,
|
|
workspace: *mut c_void,
|
|
wksp_size: usize,
|
|
huf_flags: c_int,
|
|
) -> usize {
|
|
let count_wksp = workspace.cast::<c_uint>();
|
|
let count_wksp_size = (HUF_SYMBOLVALUE_MAX as usize + 1) * size_of::<c_uint>();
|
|
let node_wksp = unsafe { workspace.cast::<u8>().add(count_wksp_size) }.cast::<c_void>();
|
|
let node_wksp_size = wksp_size.wrapping_sub(count_wksp_size);
|
|
let mut max_symbol_value = HUF_SYMBOLVALUE_MAX;
|
|
let mut huff_log = LIT_HUF_LOG;
|
|
let mut repeat = unsafe { (*prev_huf).repeatMode };
|
|
|
|
/* Prepare nextEntropy assuming reusing the existing table */
|
|
unsafe { copy_huf_tables(next_huf, prev_huf) };
|
|
|
|
if literals_compression_is_disabled != 0 {
|
|
unsafe { (*huf_metadata).hType = SET_BASIC };
|
|
return 0;
|
|
}
|
|
|
|
/* small ? don't even attempt compression (speed opt) */
|
|
{
|
|
let min_lit_size = if repeat == HUF_REPEAT_VALID {
|
|
6
|
|
} else {
|
|
COMPRESS_LITERALS_SIZE_MIN
|
|
};
|
|
if src_size <= min_lit_size {
|
|
unsafe { (*huf_metadata).hType = SET_BASIC };
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
/* Scan input and build symbol stats */
|
|
{
|
|
let largest = unsafe {
|
|
HIST_count_wksp(
|
|
count_wksp,
|
|
&mut max_symbol_value,
|
|
src,
|
|
src_size,
|
|
workspace,
|
|
wksp_size,
|
|
)
|
|
};
|
|
if ERR_isError(largest) {
|
|
return largest;
|
|
}
|
|
if largest == src_size {
|
|
/* only one literal symbol */
|
|
unsafe { (*huf_metadata).hType = SET_RLE };
|
|
return 0;
|
|
}
|
|
if largest <= (src_size >> 7) + 4 {
|
|
/* heuristic: likely not compressible */
|
|
unsafe { (*huf_metadata).hType = SET_BASIC };
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
/* Validate the previous Huffman table */
|
|
if repeat == HUF_REPEAT_CHECK
|
|
&& unsafe {
|
|
HUF_validateCTable(
|
|
ptr::addr_of!((*prev_huf).CTable).cast::<usize>(),
|
|
count_wksp,
|
|
max_symbol_value,
|
|
)
|
|
} == 0
|
|
{
|
|
repeat = HUF_REPEAT_NONE;
|
|
}
|
|
|
|
/* Build Huffman Tree */
|
|
unsafe {
|
|
ptr::write_bytes(
|
|
ptr::addr_of_mut!((*next_huf).CTable).cast::<u8>(),
|
|
0,
|
|
size_of_val(&(*next_huf).CTable),
|
|
);
|
|
}
|
|
huff_log = unsafe {
|
|
HUF_optimalTableLog(
|
|
huff_log,
|
|
src_size,
|
|
max_symbol_value,
|
|
node_wksp,
|
|
node_wksp_size,
|
|
ptr::addr_of_mut!((*next_huf).CTable).cast::<usize>(),
|
|
count_wksp,
|
|
huf_flags,
|
|
)
|
|
};
|
|
debug_assert!(huff_log <= LIT_HUF_LOG);
|
|
{
|
|
let max_bits = unsafe {
|
|
HUF_buildCTable_wksp(
|
|
ptr::addr_of_mut!((*next_huf).CTable).cast::<usize>(),
|
|
count_wksp,
|
|
max_symbol_value,
|
|
huff_log,
|
|
node_wksp,
|
|
node_wksp_size,
|
|
)
|
|
};
|
|
if ERR_isError(max_bits) {
|
|
return max_bits;
|
|
}
|
|
huff_log = max_bits as c_uint;
|
|
}
|
|
{
|
|
/* Build and write the CTable */
|
|
let new_c_size = unsafe {
|
|
HUF_estimateCompressedSize(
|
|
ptr::addr_of!((*next_huf).CTable).cast::<usize>(),
|
|
count_wksp,
|
|
max_symbol_value,
|
|
)
|
|
};
|
|
let h_size = unsafe {
|
|
HUF_writeCTable_wksp(
|
|
ptr::addr_of_mut!((*huf_metadata).hufDesBuffer).cast::<c_void>(),
|
|
size_of::<[u8; ZSTD_MAX_HUF_HEADER_SIZE]>(),
|
|
ptr::addr_of!((*next_huf).CTable).cast::<usize>(),
|
|
max_symbol_value,
|
|
huff_log,
|
|
node_wksp,
|
|
node_wksp_size,
|
|
)
|
|
};
|
|
/* Check against repeating the previous CTable */
|
|
if repeat != HUF_REPEAT_NONE {
|
|
let old_c_size = unsafe {
|
|
HUF_estimateCompressedSize(
|
|
ptr::addr_of!((*prev_huf).CTable).cast::<usize>(),
|
|
count_wksp,
|
|
max_symbol_value,
|
|
)
|
|
};
|
|
if old_c_size < src_size
|
|
&& (old_c_size <= h_size.wrapping_add(new_c_size)
|
|
|| h_size.wrapping_add(12) >= src_size)
|
|
{
|
|
unsafe {
|
|
copy_huf_tables(next_huf, prev_huf);
|
|
(*huf_metadata).hType = SET_REPEAT;
|
|
}
|
|
return 0;
|
|
}
|
|
}
|
|
if new_c_size.wrapping_add(h_size) >= src_size {
|
|
unsafe {
|
|
copy_huf_tables(next_huf, prev_huf);
|
|
(*huf_metadata).hType = SET_BASIC;
|
|
}
|
|
return 0;
|
|
}
|
|
unsafe {
|
|
(*huf_metadata).hType = SET_COMPRESSED;
|
|
(*next_huf).repeatMode = HUF_REPEAT_CHECK;
|
|
}
|
|
h_size
|
|
}
|
|
}
|
|
|
|
/// C's `ZSTD_buildDummySequencesStatistics()`: marks every encoding type as
|
|
/// basic and resets the FSE repeat modes.
|
|
unsafe fn build_dummy_sequences_statistics(
|
|
next_entropy: *mut ZSTD_fseCTables_t,
|
|
) -> SymbolEncodingTypeStats {
|
|
unsafe {
|
|
(*next_entropy).litlength_repeatMode = FSE_REPEAT_NONE;
|
|
(*next_entropy).offcode_repeatMode = FSE_REPEAT_NONE;
|
|
(*next_entropy).matchlength_repeatMode = FSE_REPEAT_NONE;
|
|
}
|
|
SymbolEncodingTypeStats {
|
|
ll_type: SET_BASIC as u32,
|
|
off_type: SET_BASIC as u32,
|
|
ml_type: SET_BASIC as u32,
|
|
size: 0,
|
|
last_count_size: 0,
|
|
long_offsets: 0,
|
|
}
|
|
}
|
|
|
|
/// C's `ZSTD_buildBlockEntropyStats_sequences()`.
|
|
#[allow(clippy::too_many_arguments)]
|
|
unsafe fn build_block_entropy_stats_sequences(
|
|
seq_store: *const SeqStore_t,
|
|
prev_entropy: *const ZSTD_fseCTables_t,
|
|
next_entropy: *mut ZSTD_fseCTables_t,
|
|
strategy: c_int,
|
|
fse_metadata: *mut ZSTD_fseCTablesMetadata_t,
|
|
workspace: *mut c_void,
|
|
wksp_size: usize,
|
|
) -> usize {
|
|
let store = unsafe { &*seq_store };
|
|
let nb_seq = unsafe { store.sequences.offset_from(store.sequencesStart) } as usize;
|
|
let count_workspace = workspace.cast::<c_uint>();
|
|
let entropy_workspace = unsafe { count_workspace.add(MAX_SEQ + 1) }.cast::<c_void>();
|
|
let entropy_workspace_size = wksp_size.wrapping_sub((MAX_SEQ + 1) * size_of::<c_uint>());
|
|
|
|
let stats = if nb_seq != 0 {
|
|
unsafe {
|
|
build_sequences_statistics(
|
|
seq_store,
|
|
nb_seq,
|
|
prev_entropy,
|
|
next_entropy,
|
|
ptr::addr_of_mut!((*fse_metadata).fseTablesBuffer).cast::<u8>(),
|
|
ZSTD_MAX_FSE_HEADERS_SIZE,
|
|
strategy,
|
|
count_workspace,
|
|
entropy_workspace,
|
|
entropy_workspace_size,
|
|
)
|
|
}
|
|
} else {
|
|
unsafe { build_dummy_sequences_statistics(next_entropy) }
|
|
};
|
|
if ERR_isError(stats.size) {
|
|
return stats.size;
|
|
}
|
|
unsafe {
|
|
(*fse_metadata).llType = stats.ll_type as c_int;
|
|
(*fse_metadata).ofType = stats.off_type as c_int;
|
|
(*fse_metadata).mlType = stats.ml_type as c_int;
|
|
(*fse_metadata).lastCountSize = stats.last_count_size;
|
|
}
|
|
stats.size
|
|
}
|
|
|
|
/// C's `ZSTD_buildBlockEntropyStats()`: builds entropy statistics for one
|
|
/// block. Requires an `ENTROPY_WORKSPACE_SIZE` workspace.
|
|
#[allow(clippy::too_many_arguments)]
|
|
pub(crate) unsafe fn build_block_entropy_stats(
|
|
seq_store: *const SeqStore_t,
|
|
prev_entropy: *const ZSTD_entropyCTables_t,
|
|
next_entropy: *mut ZSTD_entropyCTables_t,
|
|
strategy: c_int,
|
|
disable_literal_compression: c_int,
|
|
entropy_metadata: *mut ZSTD_entropyCTablesMetadata_t,
|
|
workspace: *mut c_void,
|
|
wksp_size: usize,
|
|
) -> usize {
|
|
let store = unsafe { &*seq_store };
|
|
let lit_size = unsafe { store.lit.offset_from(store.litStart) } as usize;
|
|
let huf_flags = if strategy >= HUF_OPTIMAL_DEPTH_THRESHOLD {
|
|
HUF_FLAGS_OPTIMAL_DEPTH
|
|
} else {
|
|
0
|
|
};
|
|
|
|
let huf_des_size = unsafe {
|
|
build_block_entropy_stats_literals(
|
|
store.litStart.cast::<c_void>(),
|
|
lit_size,
|
|
ptr::addr_of!((*prev_entropy).huf),
|
|
ptr::addr_of_mut!((*next_entropy).huf),
|
|
ptr::addr_of_mut!((*entropy_metadata).hufMetadata),
|
|
disable_literal_compression,
|
|
workspace,
|
|
wksp_size,
|
|
huf_flags,
|
|
)
|
|
};
|
|
unsafe { (*entropy_metadata).hufMetadata.hufDesSize = huf_des_size };
|
|
if ERR_isError(huf_des_size) {
|
|
return huf_des_size;
|
|
}
|
|
|
|
let fse_tables_size = unsafe {
|
|
build_block_entropy_stats_sequences(
|
|
seq_store,
|
|
ptr::addr_of!((*prev_entropy).fse),
|
|
ptr::addr_of_mut!((*next_entropy).fse),
|
|
strategy,
|
|
ptr::addr_of_mut!((*entropy_metadata).fseMetadata),
|
|
workspace,
|
|
wksp_size,
|
|
)
|
|
};
|
|
unsafe { (*entropy_metadata).fseMetadata.fseTablesSize = fse_tables_size };
|
|
if ERR_isError(fse_tables_size) {
|
|
return fse_tables_size;
|
|
}
|
|
0
|
|
}
|
|
|
|
/// C ABI wrapper behind the `ZSTD_buildBlockEntropyStats()` shim, which is
|
|
/// still exported from C for the block splitter's size estimation.
|
|
#[allow(clippy::too_many_arguments)]
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_buildBlockEntropyStats(
|
|
seq_store: *const SeqStore_t,
|
|
prev_entropy: *const ZSTD_entropyCTables_t,
|
|
next_entropy: *mut ZSTD_entropyCTables_t,
|
|
strategy: c_int,
|
|
disable_literal_compression: c_int,
|
|
entropy_metadata: *mut ZSTD_entropyCTablesMetadata_t,
|
|
workspace: *mut c_void,
|
|
wksp_size: usize,
|
|
) -> usize {
|
|
unsafe {
|
|
build_block_entropy_stats(
|
|
seq_store,
|
|
prev_entropy,
|
|
next_entropy,
|
|
strategy,
|
|
disable_literal_compression,
|
|
entropy_metadata,
|
|
workspace,
|
|
wksp_size,
|
|
)
|
|
}
|
|
}
|
|
|
|
/// C's `ZSTD_copyBlockSequences()`: exports a block's sequences in the public
|
|
/// `ZSTD_Sequence` format, appending the last-literals block delimiter.
|
|
#[no_mangle]
|
|
pub unsafe extern "C" fn ZSTD_rust_copyBlockSequences(
|
|
seq_collector: *mut SeqCollector,
|
|
seq_store: *const SeqStore_t,
|
|
prev_repcodes: *const u32,
|
|
) -> usize {
|
|
let collector = unsafe { &mut *seq_collector };
|
|
let store = unsafe { &*seq_store };
|
|
let in_seqs = store.sequencesStart;
|
|
let nb_in_sequences = unsafe { store.sequences.offset_from(in_seqs) } as usize;
|
|
let nb_in_literals = unsafe { store.lit.offset_from(store.litStart) } as usize;
|
|
let out_seqs = if collector.seqIndex == 0 {
|
|
collector.seqStart
|
|
} else {
|
|
unsafe { collector.seqStart.add(collector.seqIndex) }
|
|
};
|
|
let nb_out_sequences = nb_in_sequences + 1;
|
|
let mut nb_out_literals = 0usize;
|
|
|
|
/* Bounds check that we have enough space for every input sequence
|
|
* and the block delimiter
|
|
*/
|
|
debug_assert!(collector.seqIndex <= collector.maxSequences);
|
|
if nb_out_sequences > collector.maxSequences.wrapping_sub(collector.seqIndex) {
|
|
return ERROR(ZstdErrorCode::DstSizeTooSmall);
|
|
}
|
|
|
|
let mut repcodes = [0u32; ZSTD_REP_NUM];
|
|
unsafe { ptr::copy_nonoverlapping(prev_repcodes, repcodes.as_mut_ptr(), ZSTD_REP_NUM) };
|
|
for index in 0..nb_in_sequences {
|
|
let in_seq = unsafe { in_seqs.add(index).read() };
|
|
let out = unsafe { out_seqs.add(index) };
|
|
let mut lit_length = in_seq.litLength as u32;
|
|
let mut match_length = in_seq.mlBase as u32 + MINMATCH as u32;
|
|
let mut rep = 0u32;
|
|
|
|
/* Handle the possible single length >= 64K
|
|
* There can only be one because we add MINMATCH to every match length,
|
|
* and blocks are at most 128K.
|
|
*/
|
|
if index == store.longLengthPos as usize {
|
|
if store.longLengthType == ZSTD_LLT_LITERAL_LENGTH {
|
|
lit_length += 0x10000;
|
|
} else if store.longLengthType == ZSTD_LLT_MATCH_LENGTH {
|
|
match_length += 0x10000;
|
|
}
|
|
}
|
|
|
|
/* Determine the raw offset given the offBase, which may be a repcode. */
|
|
let raw_offset = if (1..=ZSTD_REP_NUM as u32).contains(&in_seq.offBase) {
|
|
let repcode = in_seq.offBase;
|
|
debug_assert!(repcode > 0);
|
|
rep = repcode;
|
|
if lit_length != 0 {
|
|
repcodes[repcode as usize - 1]
|
|
} else if repcode == 3 {
|
|
debug_assert!(repcodes[0] > 1);
|
|
repcodes[0].wrapping_sub(1)
|
|
} else {
|
|
repcodes[repcode as usize]
|
|
}
|
|
} else {
|
|
in_seq.offBase.wrapping_sub(ZSTD_REP_NUM as u32)
|
|
};
|
|
unsafe {
|
|
(*out).litLength = lit_length;
|
|
(*out).matchLength = match_length;
|
|
(*out).offset = raw_offset;
|
|
(*out).rep = rep;
|
|
}
|
|
|
|
/* Update repcode history for the sequence */
|
|
update_rep(&mut repcodes, in_seq.offBase, in_seq.litLength == 0);
|
|
|
|
nb_out_literals += lit_length as usize;
|
|
}
|
|
/* Insert last literals (if any exist) in the block as a sequence with ml == off == 0.
|
|
* If there are no last literals, then we'll emit (of: 0, ml: 0, ll: 0), which is a marker
|
|
* for the block boundary, according to the API.
|
|
*/
|
|
debug_assert!(nb_in_literals >= nb_out_literals);
|
|
{
|
|
let last_ll_size = nb_in_literals.wrapping_sub(nb_out_literals);
|
|
unsafe {
|
|
let last = out_seqs.add(nb_in_sequences);
|
|
(*last).litLength = last_ll_size as u32;
|
|
(*last).matchLength = 0;
|
|
(*last).offset = 0;
|
|
}
|
|
}
|
|
collector.seqIndex += nb_out_sequences;
|
|
debug_assert!(collector.seqIndex <= collector.maxSequences);
|
|
|
|
0
|
|
}
|
|
|
|
#[cfg(test)]
|
|
mod tests {
|
|
use super::*;
|
|
use std::mem::{align_of, offset_of};
|
|
|
|
const HUF_CTABLE_SIZE_ST: usize = HUF_SYMBOLVALUE_MAX as usize + 2;
|
|
const ENTROPY_WORKSPACE_SIZE: usize = (8 << 10) + 512 + (MAX_SEQ + 2) * size_of::<u32>();
|
|
|
|
fn empty_huf_tables(repeat: c_int) -> ZSTD_hufCTables_t {
|
|
ZSTD_hufCTables_t {
|
|
CTable: [0; HUF_CTABLE_SIZE_ST],
|
|
repeatMode: repeat,
|
|
}
|
|
}
|
|
|
|
fn empty_fse_tables(repeat: c_int) -> ZSTD_fseCTables_t {
|
|
ZSTD_fseCTables_t {
|
|
offcodeCTable: [0; OFF_CTABLE_SIZE],
|
|
matchlengthCTable: [0; ML_CTABLE_SIZE],
|
|
litlengthCTable: [0; LL_CTABLE_SIZE],
|
|
offcode_repeatMode: repeat,
|
|
matchlength_repeatMode: repeat,
|
|
litlength_repeatMode: repeat,
|
|
}
|
|
}
|
|
|
|
struct TestSeqStore {
|
|
sequences: Vec<SeqDef>,
|
|
literals: Vec<u8>,
|
|
ll_codes: Vec<u8>,
|
|
ml_codes: Vec<u8>,
|
|
of_codes: Vec<u8>,
|
|
}
|
|
|
|
impl TestSeqStore {
|
|
fn new(sequences: Vec<SeqDef>, literals: Vec<u8>) -> Self {
|
|
let nb_seq = sequences.len();
|
|
TestSeqStore {
|
|
sequences,
|
|
literals,
|
|
ll_codes: vec![0; nb_seq],
|
|
ml_codes: vec![0; nb_seq],
|
|
of_codes: vec![0; nb_seq],
|
|
}
|
|
}
|
|
|
|
fn seq_store(&mut self) -> SeqStore_t {
|
|
let nb_seq = self.sequences.len();
|
|
let nb_lit = self.literals.len();
|
|
SeqStore_t {
|
|
sequencesStart: self.sequences.as_mut_ptr(),
|
|
sequences: unsafe { self.sequences.as_mut_ptr().add(nb_seq) },
|
|
litStart: self.literals.as_mut_ptr(),
|
|
lit: unsafe { self.literals.as_mut_ptr().add(nb_lit) },
|
|
llCode: self.ll_codes.as_mut_ptr(),
|
|
mlCode: self.ml_codes.as_mut_ptr(),
|
|
ofCode: self.of_codes.as_mut_ptr(),
|
|
maxNbSeq: nb_seq,
|
|
maxNbLit: nb_lit,
|
|
longLengthType: 0,
|
|
longLengthPos: 0,
|
|
}
|
|
}
|
|
}
|
|
|
|
#[test]
|
|
fn c_leaf_layouts_match_supported_abis() {
|
|
assert_eq!(size_of::<SeqDef>(), 8);
|
|
assert_eq!(align_of::<SeqDef>(), align_of::<u32>());
|
|
assert_eq!(size_of::<BlockSummary>(), 3 * size_of::<usize>());
|
|
assert_eq!(offset_of!(BlockSummary, blockSize), size_of::<usize>());
|
|
assert_eq!(offset_of!(BlockSummary, litSize), 2 * size_of::<usize>());
|
|
assert_eq!(offset_of!(SeqStore_t, sequencesStart), 0);
|
|
assert_eq!(
|
|
offset_of!(SeqStore_t, longLengthPos),
|
|
9 * size_of::<usize>() + 4
|
|
);
|
|
assert_eq!(size_of::<SeqStore_t>(), 9 * size_of::<usize>() + 8);
|
|
assert_eq!(size_of::<ZSTD_fseCTables_t>(), 3552);
|
|
assert_eq!(offset_of!(ZSTD_compressedBlockState_t, entropy), 0);
|
|
assert_eq!(
|
|
offset_of!(ZSTD_compressedBlockState_t, rep),
|
|
size_of::<ZSTD_entropyCTables_t>()
|
|
);
|
|
assert_eq!(size_of::<ZSTD_Sequence>(), 16);
|
|
assert_eq!(offset_of!(ZSTD_SequencePosition, idx), 0);
|
|
assert_eq!(
|
|
offset_of!(ZSTD_SequencePosition, posInSequence),
|
|
size_of::<u32>()
|
|
);
|
|
assert_eq!(
|
|
offset_of!(ZSTD_SequencePosition, posInSrc),
|
|
2 * size_of::<u32>()
|
|
);
|
|
assert_eq!(
|
|
size_of::<ZSTD_SequencePosition>(),
|
|
2 * size_of::<u32>() + size_of::<usize>()
|
|
);
|
|
assert_eq!(offset_of!(SeqCollector, seqStart), size_of::<usize>());
|
|
assert_eq!(offset_of!(SeqCollector, seqIndex), 2 * size_of::<usize>());
|
|
assert_eq!(size_of::<SeqCollector>(), 4 * size_of::<usize>());
|
|
if size_of::<usize>() == 8 {
|
|
assert_eq!(size_of::<ZSTD_hufCTablesMetadata_t>(), 144);
|
|
assert_eq!(size_of::<ZSTD_fseCTablesMetadata_t>(), 168);
|
|
assert_eq!(size_of::<ZSTD_entropyCTablesMetadata_t>(), 312);
|
|
assert_eq!(size_of::<ZSTD_entropyCTables_t>(), 5616);
|
|
assert_eq!(size_of::<ZSTD_compressedBlockState_t>(), 5632);
|
|
} else {
|
|
assert_eq!(size_of::<ZSTD_hufCTablesMetadata_t>(), 136);
|
|
assert_eq!(size_of::<ZSTD_fseCTablesMetadata_t>(), 156);
|
|
assert_eq!(size_of::<ZSTD_entropyCTablesMetadata_t>(), 292);
|
|
assert_eq!(size_of::<ZSTD_entropyCTables_t>(), 4584);
|
|
assert_eq!(size_of::<ZSTD_compressedBlockState_t>(), 4596);
|
|
}
|
|
}
|
|
|
|
#[test]
|
|
fn convert_sequences_no_repcodes_maps_fields_and_ignores_rep() {
|
|
let input = [
|
|
ZSTD_Sequence {
|
|
offset: 1,
|
|
litLength: 7,
|
|
matchLength: 3,
|
|
rep: 99,
|
|
},
|
|
ZSTD_Sequence {
|
|
offset: 1234,
|
|
litLength: 0x12345,
|
|
matchLength: 17,
|
|
rep: 1,
|
|
},
|
|
];
|
|
let mut output = [SeqDef::default(); 2];
|
|
|
|
let long_length = unsafe {
|
|
ZSTD_rust_convertSequencesNoRepcodes(output.as_mut_ptr(), input.as_ptr(), input.len())
|
|
};
|
|
|
|
assert_eq!(long_length, input.len() + 1 + 1);
|
|
assert_eq!(output[0].offBase, 1 + ZSTD_REP_NUM as u32);
|
|
assert_eq!(output[0].litLength, 7);
|
|
assert_eq!(output[0].mlBase, 0);
|
|
assert_eq!(output[1].offBase, 1234 + ZSTD_REP_NUM as u32);
|
|
assert_eq!(output[1].litLength, 0x2345);
|
|
assert_eq!(output[1].mlBase, 14);
|
|
}
|
|
|
|
#[test]
|
|
fn convert_sequences_no_repcodes_marks_a_long_match() {
|
|
let input = [ZSTD_Sequence {
|
|
offset: 8,
|
|
litLength: 2,
|
|
matchLength: 65535 + MINMATCH as u32 + 1,
|
|
rep: 0,
|
|
}];
|
|
let mut output = [SeqDef::default()];
|
|
|
|
let long_length = unsafe {
|
|
ZSTD_rust_convertSequencesNoRepcodes(output.as_mut_ptr(), input.as_ptr(), input.len())
|
|
};
|
|
|
|
assert_eq!(long_length, 1);
|
|
assert_eq!(output[0].offBase, 8 + ZSTD_REP_NUM as u32);
|
|
assert_eq!(output[0].litLength, 2);
|
|
assert_eq!(output[0].mlBase, 0);
|
|
}
|
|
|
|
#[test]
|
|
fn convert_sequences_no_repcodes_accepts_an_empty_range() {
|
|
let result = unsafe {
|
|
ZSTD_rust_convertSequencesNoRepcodes(std::ptr::null_mut(), std::ptr::null(), 0)
|
|
};
|
|
assert_eq!(result, 0);
|
|
}
|
|
|
|
#[test]
|
|
fn block_summary_stops_at_and_includes_the_first_delimiter() {
|
|
let sequences = [
|
|
ZSTD_Sequence {
|
|
offset: 7,
|
|
litLength: 3,
|
|
matchLength: 4,
|
|
rep: 0,
|
|
},
|
|
ZSTD_Sequence {
|
|
offset: 0,
|
|
litLength: 5,
|
|
matchLength: 0,
|
|
rep: 0,
|
|
},
|
|
ZSTD_Sequence {
|
|
offset: 11,
|
|
litLength: 99,
|
|
matchLength: 100,
|
|
rep: 0,
|
|
},
|
|
];
|
|
|
|
let summary = unsafe { ZSTD_rust_get1BlockSummary(sequences.as_ptr(), sequences.len()) };
|
|
|
|
assert_eq!(summary.nbSequences, 2);
|
|
assert_eq!(summary.blockSize, 12);
|
|
assert_eq!(summary.litSize, 8);
|
|
}
|
|
|
|
#[test]
|
|
fn block_summary_rejects_missing_or_empty_delimiters() {
|
|
let sequences = [ZSTD_Sequence {
|
|
offset: 7,
|
|
litLength: 3,
|
|
matchLength: 4,
|
|
rep: 0,
|
|
}];
|
|
let missing = unsafe { ZSTD_rust_get1BlockSummary(sequences.as_ptr(), sequences.len()) };
|
|
assert_eq!(
|
|
missing.nbSequences,
|
|
ERROR(ZstdErrorCode::ExternalSequencesInvalid)
|
|
);
|
|
|
|
let empty = unsafe { ZSTD_rust_get1BlockSummary(std::ptr::null(), 0) };
|
|
assert_eq!(
|
|
empty.nbSequences,
|
|
ERROR(ZstdErrorCode::ExternalSequencesInvalid)
|
|
);
|
|
}
|
|
|
|
#[test]
|
|
fn transfer_sequences_with_disabled_rep_search_copies_delimiter_literals() {
|
|
let input = [
|
|
ZSTD_Sequence {
|
|
offset: 5,
|
|
litLength: 2,
|
|
matchLength: 3,
|
|
rep: 0,
|
|
},
|
|
ZSTD_Sequence {
|
|
offset: 0,
|
|
litLength: 2,
|
|
matchLength: 0,
|
|
rep: 0,
|
|
},
|
|
];
|
|
let source = [1u8, 2, 3, 4, 5, 6, 7];
|
|
let mut output = [SeqDef::default(); 2];
|
|
let mut literals = [0u8; 4];
|
|
let mut seq_store = SeqStore_t {
|
|
sequencesStart: output.as_mut_ptr(),
|
|
sequences: output.as_mut_ptr(),
|
|
litStart: literals.as_mut_ptr(),
|
|
lit: literals.as_mut_ptr(),
|
|
llCode: std::ptr::null_mut(),
|
|
mlCode: std::ptr::null_mut(),
|
|
ofCode: std::ptr::null_mut(),
|
|
maxNbSeq: output.len(),
|
|
maxNbLit: literals.len(),
|
|
longLengthType: 0,
|
|
longLengthPos: 0,
|
|
};
|
|
let mut position = ZSTD_SequencePosition::default();
|
|
let previous_repcodes = [5u32, 6, 7];
|
|
let mut next_repcodes = [0u32; ZSTD_REP_NUM];
|
|
|
|
let result = unsafe {
|
|
ZSTD_rust_transferSequencesWBlockDelim(
|
|
&mut seq_store,
|
|
&mut position,
|
|
input.as_ptr(),
|
|
input.len(),
|
|
source.as_ptr(),
|
|
source.len(),
|
|
ZSTD_PS_DISABLE,
|
|
previous_repcodes.as_ptr(),
|
|
next_repcodes.as_mut_ptr(),
|
|
0,
|
|
0,
|
|
4,
|
|
10,
|
|
0,
|
|
)
|
|
};
|
|
|
|
assert_eq!(result, source.len());
|
|
assert_eq!(position.idx, 2);
|
|
assert_eq!(position.posInSrc, 2);
|
|
assert_eq!(output[0].offBase, 5 + ZSTD_REP_NUM as u32);
|
|
assert_eq!(output[0].litLength, 2);
|
|
assert_eq!(output[0].mlBase, 0);
|
|
assert_eq!(literals, [1, 2, 6, 7]);
|
|
assert_eq!(next_repcodes, [5, 5, 6]);
|
|
}
|
|
|
|
#[test]
|
|
fn transfer_sequences_with_enabled_rep_search_uses_repcode_offsets() {
|
|
let input = [
|
|
ZSTD_Sequence {
|
|
offset: 5,
|
|
litLength: 1,
|
|
matchLength: 3,
|
|
rep: 0,
|
|
},
|
|
ZSTD_Sequence {
|
|
offset: 0,
|
|
litLength: 0,
|
|
matchLength: 0,
|
|
rep: 0,
|
|
},
|
|
];
|
|
let source = [9u8, 8, 7, 6];
|
|
let mut output = [SeqDef::default(); 2];
|
|
let mut literals = [0u8; 1];
|
|
let mut seq_store = SeqStore_t {
|
|
sequencesStart: output.as_mut_ptr(),
|
|
sequences: output.as_mut_ptr(),
|
|
litStart: literals.as_mut_ptr(),
|
|
lit: literals.as_mut_ptr(),
|
|
llCode: std::ptr::null_mut(),
|
|
mlCode: std::ptr::null_mut(),
|
|
ofCode: std::ptr::null_mut(),
|
|
maxNbSeq: output.len(),
|
|
maxNbLit: literals.len(),
|
|
longLengthType: 0,
|
|
longLengthPos: 0,
|
|
};
|
|
let mut position = ZSTD_SequencePosition::default();
|
|
let previous_repcodes = [5u32, 6, 7];
|
|
let mut next_repcodes = [0u32; ZSTD_REP_NUM];
|
|
|
|
let result = unsafe {
|
|
ZSTD_rust_transferSequencesWBlockDelim(
|
|
&mut seq_store,
|
|
&mut position,
|
|
input.as_ptr(),
|
|
input.len(),
|
|
source.as_ptr(),
|
|
source.len(),
|
|
1,
|
|
previous_repcodes.as_ptr(),
|
|
next_repcodes.as_mut_ptr(),
|
|
0,
|
|
0,
|
|
4,
|
|
10,
|
|
0,
|
|
)
|
|
};
|
|
|
|
assert_eq!(result, source.len());
|
|
assert_eq!(output[0].offBase, 1);
|
|
assert_eq!(next_repcodes, previous_repcodes);
|
|
assert_eq!(literals, [9]);
|
|
}
|
|
|
|
#[test]
|
|
fn transfer_sequences_rejects_invalid_match_lengths_and_block_mismatches() {
|
|
let invalid_match = [
|
|
ZSTD_Sequence {
|
|
offset: 1,
|
|
litLength: 0,
|
|
matchLength: 2,
|
|
rep: 0,
|
|
},
|
|
ZSTD_Sequence {
|
|
offset: 0,
|
|
litLength: 0,
|
|
matchLength: 0,
|
|
rep: 0,
|
|
},
|
|
];
|
|
let source = [1u8, 2];
|
|
let mut output = [SeqDef::default(); 2];
|
|
let mut literals = [0u8; 1];
|
|
let mut seq_store = SeqStore_t {
|
|
sequencesStart: output.as_mut_ptr(),
|
|
sequences: output.as_mut_ptr(),
|
|
litStart: literals.as_mut_ptr(),
|
|
lit: literals.as_mut_ptr(),
|
|
llCode: std::ptr::null_mut(),
|
|
mlCode: std::ptr::null_mut(),
|
|
ofCode: std::ptr::null_mut(),
|
|
maxNbSeq: output.len(),
|
|
maxNbLit: literals.len(),
|
|
longLengthType: 0,
|
|
longLengthPos: 0,
|
|
};
|
|
let mut position = ZSTD_SequencePosition::default();
|
|
let previous_repcodes = [5u32, 6, 7];
|
|
let mut next_repcodes = [0u32; ZSTD_REP_NUM];
|
|
let result = unsafe {
|
|
ZSTD_rust_transferSequencesWBlockDelim(
|
|
&mut seq_store,
|
|
&mut position,
|
|
invalid_match.as_ptr(),
|
|
invalid_match.len(),
|
|
source.as_ptr(),
|
|
source.len(),
|
|
ZSTD_PS_DISABLE,
|
|
previous_repcodes.as_ptr(),
|
|
next_repcodes.as_mut_ptr(),
|
|
0,
|
|
1,
|
|
4,
|
|
10,
|
|
0,
|
|
)
|
|
};
|
|
assert_eq!(result, ERROR(ZstdErrorCode::ExternalSequencesInvalid));
|
|
assert_eq!(position.posInSrc, source.len());
|
|
assert_eq!(seq_store.sequences, seq_store.sequencesStart);
|
|
|
|
let mismatch = [
|
|
ZSTD_Sequence {
|
|
offset: 1,
|
|
litLength: 1,
|
|
matchLength: 3,
|
|
rep: 0,
|
|
},
|
|
ZSTD_Sequence {
|
|
offset: 0,
|
|
litLength: 1,
|
|
matchLength: 0,
|
|
rep: 0,
|
|
},
|
|
];
|
|
let mut mismatch_store = SeqStore_t {
|
|
sequencesStart: output.as_mut_ptr(),
|
|
sequences: output.as_mut_ptr(),
|
|
litStart: literals.as_mut_ptr(),
|
|
lit: literals.as_mut_ptr(),
|
|
llCode: std::ptr::null_mut(),
|
|
mlCode: std::ptr::null_mut(),
|
|
ofCode: std::ptr::null_mut(),
|
|
maxNbSeq: output.len(),
|
|
maxNbLit: literals.len(),
|
|
longLengthType: 0,
|
|
longLengthPos: 0,
|
|
};
|
|
let mut mismatch_position = ZSTD_SequencePosition::default();
|
|
let result = unsafe {
|
|
ZSTD_rust_transferSequencesWBlockDelim(
|
|
&mut mismatch_store,
|
|
&mut mismatch_position,
|
|
mismatch.as_ptr(),
|
|
mismatch.len(),
|
|
source.as_ptr(),
|
|
source.len(),
|
|
ZSTD_PS_DISABLE,
|
|
previous_repcodes.as_ptr(),
|
|
next_repcodes.as_mut_ptr(),
|
|
0,
|
|
0,
|
|
4,
|
|
10,
|
|
0,
|
|
)
|
|
};
|
|
assert_eq!(result, ERROR(ZstdErrorCode::ExternalSequencesInvalid));
|
|
}
|
|
|
|
#[test]
|
|
fn transfer_sequences_preserves_a_long_literal_side_band() {
|
|
let literal_length = 65536usize;
|
|
let input = [
|
|
ZSTD_Sequence {
|
|
offset: 1,
|
|
litLength: literal_length as u32,
|
|
matchLength: 3,
|
|
rep: 0,
|
|
},
|
|
ZSTD_Sequence {
|
|
offset: 0,
|
|
litLength: 0,
|
|
matchLength: 0,
|
|
rep: 0,
|
|
},
|
|
];
|
|
let source = vec![0xA5u8; literal_length + 3];
|
|
let mut output = [SeqDef::default(); 2];
|
|
let mut literals = vec![0u8; literal_length];
|
|
let mut seq_store = SeqStore_t {
|
|
sequencesStart: output.as_mut_ptr(),
|
|
sequences: output.as_mut_ptr(),
|
|
litStart: literals.as_mut_ptr(),
|
|
lit: literals.as_mut_ptr(),
|
|
llCode: std::ptr::null_mut(),
|
|
mlCode: std::ptr::null_mut(),
|
|
ofCode: std::ptr::null_mut(),
|
|
maxNbSeq: output.len(),
|
|
maxNbLit: literals.len(),
|
|
longLengthType: 0,
|
|
longLengthPos: 0,
|
|
};
|
|
let mut position = ZSTD_SequencePosition::default();
|
|
let repcodes = [5u32, 6, 7];
|
|
let mut next_repcodes = [0u32; ZSTD_REP_NUM];
|
|
|
|
let result = unsafe {
|
|
ZSTD_rust_transferSequencesWBlockDelim(
|
|
&mut seq_store,
|
|
&mut position,
|
|
input.as_ptr(),
|
|
input.len(),
|
|
source.as_ptr(),
|
|
source.len(),
|
|
ZSTD_PS_DISABLE,
|
|
repcodes.as_ptr(),
|
|
next_repcodes.as_mut_ptr(),
|
|
0,
|
|
0,
|
|
4,
|
|
10,
|
|
0,
|
|
)
|
|
};
|
|
|
|
assert_eq!(result, source.len());
|
|
assert_eq!(output[0].litLength, 0);
|
|
assert_eq!(output[0].mlBase, 0);
|
|
assert_eq!(seq_store.longLengthType, ZSTD_LLT_LITERAL_LENGTH);
|
|
assert_eq!(seq_store.longLengthPos, 0);
|
|
assert_eq!(literals, source[..literal_length]);
|
|
}
|
|
|
|
#[test]
|
|
fn estimate_block_size_counts_basic_literals_and_rle_sequences() {
|
|
let entropy = ZSTD_entropyCTables_t {
|
|
huf: empty_huf_tables(HUF_REPEAT_NONE),
|
|
fse: empty_fse_tables(FSE_REPEAT_NONE),
|
|
};
|
|
let mut metadata = unsafe {
|
|
std::mem::MaybeUninit::<ZSTD_entropyCTablesMetadata_t>::zeroed().assume_init()
|
|
};
|
|
metadata.hufMetadata.hType = SET_BASIC;
|
|
metadata.fseMetadata.ofType = SET_RLE;
|
|
metadata.fseMetadata.llType = SET_RLE;
|
|
metadata.fseMetadata.mlType = SET_RLE;
|
|
metadata.fseMetadata.fseTablesSize = 7;
|
|
let literals = *b"abcd";
|
|
let codes = [0u8];
|
|
let mut workspace = vec![0u64; ENTROPY_WORKSPACE_SIZE / size_of::<u64>()];
|
|
|
|
let without_entropy = unsafe {
|
|
ZSTD_rust_estimateBlockSize(
|
|
literals.as_ptr(),
|
|
literals.len(),
|
|
codes.as_ptr(),
|
|
codes.as_ptr(),
|
|
codes.as_ptr(),
|
|
1,
|
|
&entropy,
|
|
&metadata,
|
|
workspace.as_mut_ptr().cast(),
|
|
ENTROPY_WORKSPACE_SIZE,
|
|
0,
|
|
0,
|
|
)
|
|
};
|
|
assert_eq!(without_entropy, 9);
|
|
|
|
let with_entropy = unsafe {
|
|
ZSTD_rust_estimateBlockSize(
|
|
literals.as_ptr(),
|
|
literals.len(),
|
|
codes.as_ptr(),
|
|
codes.as_ptr(),
|
|
codes.as_ptr(),
|
|
1,
|
|
&entropy,
|
|
&metadata,
|
|
workspace.as_mut_ptr().cast(),
|
|
ENTROPY_WORKSPACE_SIZE,
|
|
0,
|
|
1,
|
|
)
|
|
};
|
|
assert_eq!(with_entropy, 16);
|
|
}
|
|
|
|
#[test]
|
|
fn estimate_block_size_counts_rle_literals() {
|
|
let entropy = ZSTD_entropyCTables_t {
|
|
huf: empty_huf_tables(HUF_REPEAT_NONE),
|
|
fse: empty_fse_tables(FSE_REPEAT_NONE),
|
|
};
|
|
let mut metadata = unsafe {
|
|
std::mem::MaybeUninit::<ZSTD_entropyCTablesMetadata_t>::zeroed().assume_init()
|
|
};
|
|
metadata.hufMetadata.hType = SET_RLE;
|
|
metadata.fseMetadata.ofType = SET_RLE;
|
|
metadata.fseMetadata.llType = SET_RLE;
|
|
metadata.fseMetadata.mlType = SET_RLE;
|
|
let literals = [b'x'; 1024];
|
|
let codes = [0u8];
|
|
let mut workspace = vec![0u64; ENTROPY_WORKSPACE_SIZE / size_of::<u64>()];
|
|
|
|
let estimate = unsafe {
|
|
ZSTD_rust_estimateBlockSize(
|
|
literals.as_ptr(),
|
|
literals.len(),
|
|
codes.as_ptr(),
|
|
codes.as_ptr(),
|
|
codes.as_ptr(),
|
|
1,
|
|
&entropy,
|
|
&metadata,
|
|
workspace.as_mut_ptr().cast(),
|
|
ENTROPY_WORKSPACE_SIZE,
|
|
0,
|
|
0,
|
|
)
|
|
};
|
|
assert_eq!(estimate, 6);
|
|
}
|
|
|
|
#[test]
|
|
fn compressed_block_state_reset_restores_repcodes_and_repeat_modes() {
|
|
let mut block_state =
|
|
unsafe { std::mem::MaybeUninit::<ZSTD_compressedBlockState_t>::zeroed().assume_init() };
|
|
block_state.rep = [99; ZSTD_REP_NUM];
|
|
block_state.entropy.huf.repeatMode = HUF_REPEAT_VALID;
|
|
block_state.entropy.fse.offcode_repeatMode = 2;
|
|
block_state.entropy.fse.matchlength_repeatMode = 1;
|
|
block_state.entropy.fse.litlength_repeatMode = 2;
|
|
|
|
unsafe { ZSTD_rust_resetCompressedBlockState(&mut block_state) };
|
|
|
|
assert_eq!(block_state.rep, [1, 4, 8]);
|
|
assert_eq!(block_state.entropy.huf.repeatMode, HUF_REPEAT_NONE);
|
|
assert_eq!(block_state.entropy.fse.offcode_repeatMode, FSE_REPEAT_NONE);
|
|
assert_eq!(
|
|
block_state.entropy.fse.matchlength_repeatMode,
|
|
FSE_REPEAT_NONE
|
|
);
|
|
assert_eq!(
|
|
block_state.entropy.fse.litlength_repeatMode,
|
|
FSE_REPEAT_NONE
|
|
);
|
|
}
|
|
|
|
#[test]
|
|
fn store_last_literals_copies_bytes_and_advances_only_the_literal_cursor() {
|
|
let mut sequences = [SeqDef::default(); 2];
|
|
let mut literals = [0xA5u8; 8];
|
|
let sequences_start = sequences.as_mut_ptr();
|
|
let literals_start = literals.as_mut_ptr();
|
|
let mut seq_store = SeqStore_t {
|
|
sequencesStart: sequences_start,
|
|
sequences: unsafe { sequences_start.add(1) },
|
|
litStart: literals_start,
|
|
lit: unsafe { literals_start.add(2) },
|
|
llCode: std::ptr::null_mut(),
|
|
mlCode: std::ptr::null_mut(),
|
|
ofCode: std::ptr::null_mut(),
|
|
maxNbSeq: sequences.len(),
|
|
maxNbLit: literals.len(),
|
|
longLengthType: ZSTD_LLT_MATCH_LENGTH,
|
|
longLengthPos: 7,
|
|
};
|
|
let source = [1u8, 2, 3, 4];
|
|
|
|
unsafe {
|
|
ZSTD_rust_storeLastLiterals(&mut seq_store, source.as_ptr(), 3);
|
|
}
|
|
assert_eq!(literals, [0xA5, 0xA5, 1, 2, 3, 0xA5, 0xA5, 0xA5]);
|
|
assert_eq!(seq_store.lit, unsafe { literals_start.add(5) });
|
|
assert_eq!(seq_store.sequences, unsafe { sequences_start.add(1) });
|
|
|
|
let cursor = seq_store.lit;
|
|
unsafe {
|
|
ZSTD_rust_storeLastLiterals(&mut seq_store, source.as_ptr(), 0);
|
|
}
|
|
assert_eq!(seq_store.lit, cursor);
|
|
assert_eq!(literals, [0xA5, 0xA5, 1, 2, 3, 0xA5, 0xA5, 0xA5]);
|
|
}
|
|
|
|
#[test]
|
|
fn reset_seq_store_resets_cursors_and_long_length_type_but_not_position() {
|
|
let mut sequences = [SeqDef::default(); 2];
|
|
let mut literals = [0u8; 5];
|
|
let sequences_start = sequences.as_mut_ptr();
|
|
let literals_start = literals.as_mut_ptr();
|
|
let mut seq_store = SeqStore_t {
|
|
sequencesStart: sequences_start,
|
|
sequences: unsafe { sequences_start.add(2) },
|
|
litStart: literals_start,
|
|
lit: unsafe { literals_start.add(4) },
|
|
llCode: std::ptr::null_mut(),
|
|
mlCode: std::ptr::null_mut(),
|
|
ofCode: std::ptr::null_mut(),
|
|
maxNbSeq: sequences.len(),
|
|
maxNbLit: literals.len(),
|
|
longLengthType: ZSTD_LLT_LITERAL_LENGTH,
|
|
longLengthPos: 11,
|
|
};
|
|
|
|
unsafe { ZSTD_rust_resetSeqStore(&mut seq_store) };
|
|
|
|
assert_eq!(seq_store.sequences, sequences_start);
|
|
assert_eq!(seq_store.lit, literals_start);
|
|
assert_eq!(seq_store.longLengthType, 0);
|
|
assert_eq!(seq_store.longLengthPos, 11);
|
|
}
|
|
|
|
#[test]
|
|
fn fast_sequence_length_sum_handles_empty_and_populated_buffers() {
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_fastSequenceLengthSum(std::ptr::null(), 0) },
|
|
0
|
|
);
|
|
|
|
let sequences = [
|
|
ZSTD_Sequence {
|
|
offset: 0,
|
|
litLength: 3,
|
|
matchLength: 7,
|
|
rep: 0,
|
|
},
|
|
ZSTD_Sequence {
|
|
offset: 10,
|
|
litLength: 11,
|
|
matchLength: 13,
|
|
rep: 1,
|
|
},
|
|
];
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_fastSequenceLengthSum(sequences.as_ptr(), sequences.len()) },
|
|
34
|
|
);
|
|
}
|
|
|
|
#[test]
|
|
fn post_process_rejects_external_count_overflow() {
|
|
let mut out = [ZSTD_Sequence {
|
|
offset: 0xA5A5_A5A5,
|
|
litLength: 0x5A5A_5A5A,
|
|
matchLength: 0xA5A5_A5A5,
|
|
rep: 0x5A5A_5A5A,
|
|
}];
|
|
let result = unsafe {
|
|
ZSTD_rust_postProcessSequenceProducerResult(out.as_mut_ptr(), usize::MAX, out.len(), 1)
|
|
};
|
|
assert_eq!(result, ERROR(ZstdErrorCode::SequenceProducerFailed));
|
|
assert_eq!(out[0].offset, 0xA5A5_A5A5);
|
|
assert_eq!(out[0].litLength, 0x5A5A_5A5A);
|
|
assert_eq!(out[0].matchLength, 0xA5A5_A5A5);
|
|
assert_eq!(out[0].rep, 0x5A5A_5A5A);
|
|
}
|
|
|
|
#[test]
|
|
fn post_process_empty_source_zeroes_the_output_sequence() {
|
|
let mut out = [ZSTD_Sequence {
|
|
offset: 1,
|
|
litLength: 2,
|
|
matchLength: 3,
|
|
rep: 4,
|
|
}];
|
|
let result =
|
|
unsafe { ZSTD_rust_postProcessSequenceProducerResult(out.as_mut_ptr(), 0, 1, 0) };
|
|
assert_eq!(result, 1);
|
|
assert_eq!(out[0].offset, 0);
|
|
assert_eq!(out[0].litLength, 0);
|
|
assert_eq!(out[0].matchLength, 0);
|
|
assert_eq!(out[0].rep, 0);
|
|
}
|
|
|
|
#[test]
|
|
fn post_process_preserves_an_existing_delimiter() {
|
|
let mut out = [ZSTD_Sequence {
|
|
offset: 0,
|
|
litLength: 0xA5A5_A5A5,
|
|
matchLength: 0,
|
|
rep: 0x5A5A_5A5A,
|
|
}];
|
|
let result =
|
|
unsafe { ZSTD_rust_postProcessSequenceProducerResult(out.as_mut_ptr(), 1, 1, 7) };
|
|
assert_eq!(result, 1);
|
|
assert_eq!(out[0].offset, 0);
|
|
assert_eq!(out[0].litLength, 0xA5A5_A5A5);
|
|
assert_eq!(out[0].matchLength, 0);
|
|
assert_eq!(out[0].rep, 0x5A5A_5A5A);
|
|
}
|
|
|
|
#[test]
|
|
fn post_process_appends_a_zeroed_delimiter() {
|
|
let mut out = [
|
|
ZSTD_Sequence {
|
|
offset: 3,
|
|
litLength: 4,
|
|
matchLength: 5,
|
|
rep: 6,
|
|
},
|
|
ZSTD_Sequence {
|
|
offset: 0xA5A5_A5A5,
|
|
litLength: 0x5A5A_5A5A,
|
|
matchLength: 0xA5A5_A5A5,
|
|
rep: 0x5A5A_5A5A,
|
|
},
|
|
];
|
|
let result = unsafe {
|
|
ZSTD_rust_postProcessSequenceProducerResult(out.as_mut_ptr(), 1, out.len(), 7)
|
|
};
|
|
assert_eq!(result, 2);
|
|
assert_eq!(out[0].offset, 3);
|
|
assert_eq!(out[0].litLength, 4);
|
|
assert_eq!(out[0].matchLength, 5);
|
|
assert_eq!(out[0].rep, 6);
|
|
assert_eq!(out[1].offset, 0);
|
|
assert_eq!(out[1].litLength, 0);
|
|
assert_eq!(out[1].matchLength, 0);
|
|
assert_eq!(out[1].rep, 0);
|
|
}
|
|
|
|
#[test]
|
|
fn post_process_reports_empty_parse_for_non_empty_source() {
|
|
let mut out = [ZSTD_Sequence {
|
|
offset: 0xA5A5_A5A5,
|
|
litLength: 0x5A5A_5A5A,
|
|
matchLength: 0xA5A5_A5A5,
|
|
rep: 0x5A5A_5A5A,
|
|
}];
|
|
let result = unsafe {
|
|
ZSTD_rust_postProcessSequenceProducerResult(out.as_mut_ptr(), 0, out.len(), 1)
|
|
};
|
|
assert_eq!(result, ERROR(ZstdErrorCode::SequenceProducerFailed));
|
|
}
|
|
|
|
#[test]
|
|
fn post_process_reports_full_capacity_without_a_delimiter() {
|
|
let mut out = [ZSTD_Sequence {
|
|
offset: 1,
|
|
litLength: 2,
|
|
matchLength: 3,
|
|
rep: 4,
|
|
}];
|
|
let result = unsafe {
|
|
ZSTD_rust_postProcessSequenceProducerResult(out.as_mut_ptr(), 1, out.len(), 7)
|
|
};
|
|
assert_eq!(result, ERROR(ZstdErrorCode::SequenceProducerFailed));
|
|
}
|
|
|
|
#[test]
|
|
fn is_rle_matches_empty_one_byte_and_unrolled_inputs() {
|
|
let one_byte = [0x5Au8];
|
|
assert_eq!(unsafe { ZSTD_rust_isRLE(one_byte.as_ptr(), 0) }, 1);
|
|
assert_eq!(unsafe { ZSTD_rust_isRLE(one_byte.as_ptr(), 1) }, 1);
|
|
|
|
let unrolled_size = size_of::<usize>() * 4;
|
|
let mut repeated = vec![0xA7u8; unrolled_size + 3];
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_isRLE(repeated.as_ptr(), repeated.len()) },
|
|
1
|
|
);
|
|
|
|
repeated[1] ^= 1;
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_isRLE(repeated.as_ptr(), repeated.len()) },
|
|
0
|
|
);
|
|
repeated[1] ^= 1;
|
|
repeated[3] ^= 1;
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_isRLE(repeated.as_ptr(), repeated.len()) },
|
|
0
|
|
);
|
|
}
|
|
|
|
#[test]
|
|
fn maybe_rle_uses_the_original_sequence_and_literal_thresholds() {
|
|
let mut sequences = [SeqDef::default(); 4];
|
|
let mut literals = [0u8; 10];
|
|
let sequences_start = sequences.as_mut_ptr();
|
|
let literals_start = literals.as_mut_ptr();
|
|
let mut seq_store = SeqStore_t {
|
|
sequencesStart: sequences_start,
|
|
sequences: sequences_start,
|
|
litStart: literals_start,
|
|
lit: literals_start,
|
|
llCode: std::ptr::null_mut(),
|
|
mlCode: std::ptr::null_mut(),
|
|
ofCode: std::ptr::null_mut(),
|
|
maxNbSeq: sequences.len(),
|
|
maxNbLit: literals.len(),
|
|
longLengthType: 0,
|
|
longLengthPos: 0,
|
|
};
|
|
|
|
assert_eq!(unsafe { ZSTD_rust_maybeRLE(&seq_store) }, 1);
|
|
seq_store.sequences = unsafe { sequences_start.add(3) };
|
|
seq_store.lit = unsafe { literals_start.add(9) };
|
|
assert_eq!(unsafe { ZSTD_rust_maybeRLE(&seq_store) }, 1);
|
|
seq_store.sequences = unsafe { sequences_start.add(4) };
|
|
assert_eq!(unsafe { ZSTD_rust_maybeRLE(&seq_store) }, 0);
|
|
seq_store.sequences = unsafe { sequences_start.add(3) };
|
|
seq_store.lit = unsafe { literals_start.add(10) };
|
|
assert_eq!(unsafe { ZSTD_rust_maybeRLE(&seq_store) }, 0);
|
|
}
|
|
|
|
#[test]
|
|
fn sequence_store_byte_counts_include_the_side_band_long_length() {
|
|
let mut sequences = [
|
|
SeqDef {
|
|
offBase: 4,
|
|
litLength: 2,
|
|
mlBase: 4,
|
|
},
|
|
SeqDef {
|
|
offBase: 5,
|
|
litLength: 5,
|
|
mlBase: 5,
|
|
},
|
|
SeqDef {
|
|
offBase: 6,
|
|
litLength: 7,
|
|
mlBase: 6,
|
|
},
|
|
];
|
|
let mut literals = [0u8; 32];
|
|
let sequences_start = sequences.as_mut_ptr();
|
|
let literals_start = literals.as_mut_ptr();
|
|
let mut seq_store = SeqStore_t {
|
|
sequencesStart: sequences_start,
|
|
sequences: unsafe { sequences_start.add(sequences.len()) },
|
|
litStart: literals_start,
|
|
lit: unsafe { literals_start.add(14) },
|
|
llCode: std::ptr::null_mut(),
|
|
mlCode: std::ptr::null_mut(),
|
|
ofCode: std::ptr::null_mut(),
|
|
maxNbSeq: sequences.len(),
|
|
maxNbLit: literals.len(),
|
|
longLengthType: ZSTD_LLT_LITERAL_LENGTH,
|
|
longLengthPos: 1,
|
|
};
|
|
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_countSeqStoreLiteralsBytes(&seq_store) },
|
|
2 + 5 + 7 + 0x10000
|
|
);
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_countSeqStoreMatchBytes(&seq_store) },
|
|
(4 + MINMATCH) + (5 + MINMATCH) + (6 + MINMATCH)
|
|
);
|
|
|
|
seq_store.longLengthType = ZSTD_LLT_MATCH_LENGTH;
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_countSeqStoreLiteralsBytes(&seq_store) },
|
|
2 + 5 + 7
|
|
);
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_countSeqStoreMatchBytes(&seq_store) },
|
|
(4 + MINMATCH) + (5 + MINMATCH) + (6 + MINMATCH) + 0x10000
|
|
);
|
|
}
|
|
|
|
#[test]
|
|
fn derive_seq_store_chunk_shifts_views_and_long_length_position() {
|
|
let mut sequences = [
|
|
SeqDef {
|
|
offBase: 4,
|
|
litLength: 2,
|
|
mlBase: 0,
|
|
},
|
|
SeqDef {
|
|
offBase: 5,
|
|
litLength: 5,
|
|
mlBase: 1,
|
|
},
|
|
SeqDef {
|
|
offBase: 6,
|
|
litLength: 7,
|
|
mlBase: 2,
|
|
},
|
|
SeqDef {
|
|
offBase: 7,
|
|
litLength: 3,
|
|
mlBase: 3,
|
|
},
|
|
];
|
|
let mut literals = [0u8; 20];
|
|
let mut ll_codes = [10u8; 4];
|
|
let mut ml_codes = [20u8; 4];
|
|
let mut of_codes = [30u8; 4];
|
|
let sequences_start = sequences.as_mut_ptr();
|
|
let literals_start = literals.as_mut_ptr();
|
|
let original = SeqStore_t {
|
|
sequencesStart: sequences_start,
|
|
sequences: unsafe { sequences_start.add(sequences.len()) },
|
|
litStart: literals_start,
|
|
lit: unsafe { literals_start.add(17) },
|
|
llCode: ll_codes.as_mut_ptr(),
|
|
mlCode: ml_codes.as_mut_ptr(),
|
|
ofCode: of_codes.as_mut_ptr(),
|
|
maxNbSeq: sequences.len(),
|
|
maxNbLit: literals.len(),
|
|
longLengthType: ZSTD_LLT_MATCH_LENGTH,
|
|
longLengthPos: 2,
|
|
};
|
|
let mut result = SeqStore_t {
|
|
sequencesStart: std::ptr::null_mut(),
|
|
sequences: std::ptr::null_mut(),
|
|
litStart: std::ptr::null_mut(),
|
|
lit: std::ptr::null_mut(),
|
|
llCode: std::ptr::null_mut(),
|
|
mlCode: std::ptr::null_mut(),
|
|
ofCode: std::ptr::null_mut(),
|
|
maxNbSeq: 0,
|
|
maxNbLit: 0,
|
|
longLengthType: 0,
|
|
longLengthPos: 0,
|
|
};
|
|
|
|
unsafe {
|
|
ZSTD_rust_deriveSeqStoreChunk(&mut result, &original, 1, 3);
|
|
}
|
|
assert_eq!(result.sequencesStart, unsafe { sequences_start.add(1) });
|
|
assert_eq!(result.sequences, unsafe { sequences_start.add(3) });
|
|
assert_eq!(result.litStart, unsafe { literals_start.add(2) });
|
|
assert_eq!(result.lit, unsafe { literals_start.add(14) });
|
|
assert_eq!(result.llCode, unsafe { ll_codes.as_mut_ptr().add(1) });
|
|
assert_eq!(result.mlCode, unsafe { ml_codes.as_mut_ptr().add(1) });
|
|
assert_eq!(result.ofCode, unsafe { of_codes.as_mut_ptr().add(1) });
|
|
assert_eq!(result.longLengthType, ZSTD_LLT_MATCH_LENGTH);
|
|
assert_eq!(result.longLengthPos, 1);
|
|
assert_eq!(result.maxNbSeq, original.maxNbSeq);
|
|
assert_eq!(result.maxNbLit, original.maxNbLit);
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_countSeqStoreMatchBytes(&result) },
|
|
(1 + MINMATCH) + (2 + MINMATCH) + 0x10000
|
|
);
|
|
|
|
unsafe {
|
|
ZSTD_rust_deriveSeqStoreChunk(&mut result, &original, 3, 4);
|
|
}
|
|
assert_eq!(result.sequencesStart, unsafe { sequences_start.add(3) });
|
|
assert_eq!(result.sequences, unsafe { sequences_start.add(4) });
|
|
assert_eq!(result.litStart, unsafe { literals_start.add(14) });
|
|
assert_eq!(result.lit, original.lit);
|
|
assert_eq!(result.longLengthType, 0);
|
|
assert_eq!(result.longLengthPos, original.longLengthPos);
|
|
}
|
|
|
|
#[test]
|
|
fn resolve_repcode_to_raw_offset_matches_c_repcode_numbering() {
|
|
let reps = [11u32, 22, 33];
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_resolveRepcodeToRawOffset(reps.as_ptr(), 1, 0) },
|
|
11
|
|
);
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_resolveRepcodeToRawOffset(reps.as_ptr(), 1, 1) },
|
|
22
|
|
);
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_resolveRepcodeToRawOffset(reps.as_ptr(), 2, 0) },
|
|
22
|
|
);
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_resolveRepcodeToRawOffset(reps.as_ptr(), 2, 1) },
|
|
33
|
|
);
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_resolveRepcodeToRawOffset(reps.as_ptr(), 3, 0) },
|
|
33
|
|
);
|
|
assert_eq!(
|
|
unsafe { ZSTD_rust_resolveRepcodeToRawOffset(reps.as_ptr(), 3, 1) },
|
|
10
|
|
);
|
|
}
|
|
|
|
#[test]
|
|
fn repcode_updates_match_the_c_sum_type_rules() {
|
|
let mut reps = [1, 4, 8];
|
|
update_rep(&mut reps, 10, false);
|
|
assert_eq!(reps, [7, 1, 4]);
|
|
update_rep(&mut reps, 1, true);
|
|
assert_eq!(reps, [1, 7, 4]);
|
|
update_rep(&mut reps, 2, false);
|
|
assert_eq!(reps, [7, 1, 4]);
|
|
update_rep(&mut reps, 3, false);
|
|
assert_eq!(reps, [4, 7, 1]);
|
|
}
|
|
|
|
#[test]
|
|
fn resolve_off_codes_keeps_matching_histories_aligned() {
|
|
let mut fixture = TestSeqStore::new(
|
|
vec![
|
|
SeqDef {
|
|
offBase: 1,
|
|
litLength: 4,
|
|
mlBase: 0,
|
|
},
|
|
SeqDef {
|
|
offBase: 2,
|
|
litLength: 0,
|
|
mlBase: 0,
|
|
},
|
|
],
|
|
vec![0; 4],
|
|
);
|
|
let seq_store = fixture.seq_store();
|
|
let mut d_rep = [11, 22, 33];
|
|
let mut c_rep = d_rep;
|
|
|
|
unsafe {
|
|
ZSTD_rust_seqStore_resolveOffCodes(
|
|
d_rep.as_mut_ptr(),
|
|
c_rep.as_mut_ptr(),
|
|
&seq_store,
|
|
2,
|
|
);
|
|
}
|
|
|
|
assert_eq!(fixture.sequences[0].offBase, 1);
|
|
assert_eq!(fixture.sequences[1].offBase, 2);
|
|
assert_eq!(d_rep, [33, 11, 22]);
|
|
assert_eq!(c_rep, d_rep);
|
|
}
|
|
|
|
#[test]
|
|
fn resolve_off_codes_materializes_mismatched_repcode() {
|
|
let mut fixture = TestSeqStore::new(
|
|
vec![SeqDef {
|
|
offBase: 2,
|
|
litLength: 1,
|
|
mlBase: 0,
|
|
}],
|
|
vec![0; 1],
|
|
);
|
|
let seq_store = fixture.seq_store();
|
|
let mut d_rep = [11, 22, 33];
|
|
let mut c_rep = [11, 44, 33];
|
|
|
|
unsafe {
|
|
ZSTD_rust_seqStore_resolveOffCodes(
|
|
d_rep.as_mut_ptr(),
|
|
c_rep.as_mut_ptr(),
|
|
&seq_store,
|
|
1,
|
|
);
|
|
}
|
|
|
|
assert_eq!(fixture.sequences[0].offBase, 44 + ZSTD_REP_NUM as u32);
|
|
assert_eq!(d_rep, [44, 11, 22]);
|
|
assert_eq!(c_rep, [44, 11, 33]);
|
|
}
|
|
|
|
#[test]
|
|
fn resolve_off_codes_updates_both_histories_for_full_offsets() {
|
|
let mut fixture = TestSeqStore::new(
|
|
vec![SeqDef {
|
|
offBase: 103,
|
|
litLength: 1,
|
|
mlBase: 0,
|
|
}],
|
|
vec![0; 1],
|
|
);
|
|
let seq_store = fixture.seq_store();
|
|
let mut d_rep = [10, 20, 30];
|
|
let mut c_rep = [90, 80, 70];
|
|
|
|
unsafe {
|
|
ZSTD_rust_seqStore_resolveOffCodes(
|
|
d_rep.as_mut_ptr(),
|
|
c_rep.as_mut_ptr(),
|
|
&seq_store,
|
|
1,
|
|
);
|
|
}
|
|
|
|
assert_eq!(fixture.sequences[0].offBase, 103);
|
|
assert_eq!(d_rep, [100, 10, 20]);
|
|
assert_eq!(c_rep, [100, 90, 80]);
|
|
}
|
|
|
|
#[test]
|
|
fn resolve_off_codes_excludes_long_literal_from_ll0() {
|
|
let mut fixture = TestSeqStore::new(
|
|
vec![
|
|
SeqDef {
|
|
offBase: 1,
|
|
litLength: 0,
|
|
mlBase: 0,
|
|
},
|
|
SeqDef {
|
|
offBase: 1,
|
|
litLength: 0,
|
|
mlBase: 0,
|
|
},
|
|
],
|
|
vec![0; 2],
|
|
);
|
|
let mut seq_store = fixture.seq_store();
|
|
seq_store.longLengthType = ZSTD_LLT_LITERAL_LENGTH;
|
|
seq_store.longLengthPos = 0;
|
|
let mut d_rep = [11, 22, 33];
|
|
let mut c_rep = d_rep;
|
|
|
|
unsafe {
|
|
ZSTD_rust_seqStore_resolveOffCodes(
|
|
d_rep.as_mut_ptr(),
|
|
c_rep.as_mut_ptr(),
|
|
&seq_store,
|
|
2,
|
|
);
|
|
}
|
|
|
|
assert_eq!(fixture.sequences[0].offBase, 1);
|
|
assert_eq!(fixture.sequences[1].offBase, 1);
|
|
assert_eq!(d_rep, [22, 11, 33]);
|
|
assert_eq!(c_rep, d_rep);
|
|
}
|
|
|
|
#[test]
|
|
fn seq_to_codes_matches_the_reference_tables() {
|
|
let mut store = TestSeqStore::new(
|
|
vec![
|
|
SeqDef {
|
|
offBase: 4,
|
|
litLength: 0,
|
|
mlBase: 0,
|
|
},
|
|
SeqDef {
|
|
offBase: 1,
|
|
litLength: 63,
|
|
mlBase: 127,
|
|
},
|
|
SeqDef {
|
|
offBase: 1 << 20,
|
|
litLength: 64,
|
|
mlBase: 128,
|
|
},
|
|
],
|
|
vec![0; 127],
|
|
);
|
|
let seq_store = store.seq_store();
|
|
let long_offsets = unsafe { ZSTD_seqToCodes(&seq_store) };
|
|
assert_eq!(long_offsets, 0);
|
|
assert_eq!(store.ll_codes, [0, 24, 25]);
|
|
assert_eq!(store.ml_codes, [0, 42, 43]);
|
|
assert_eq!(store.of_codes, [2, 0, 20]);
|
|
}
|
|
|
|
#[test]
|
|
fn rle_sequences_produce_one_byte_table_headers() {
|
|
// Five identical sequences select set_rle for all three symbol kinds;
|
|
// each table header is then exactly the one RLE symbol byte.
|
|
let mut store = TestSeqStore::new(
|
|
vec![
|
|
SeqDef {
|
|
offBase: 8,
|
|
litLength: 2,
|
|
mlBase: 3,
|
|
};
|
|
5
|
|
],
|
|
vec![b'x'; 10],
|
|
);
|
|
let seq_store = store.seq_store();
|
|
let prev = empty_fse_tables(FSE_REPEAT_NONE);
|
|
let mut next = empty_fse_tables(FSE_REPEAT_NONE);
|
|
let mut metadata =
|
|
unsafe { std::mem::MaybeUninit::<ZSTD_fseCTablesMetadata_t>::zeroed().assume_init() };
|
|
let mut workspace = vec![0u64; ENTROPY_WORKSPACE_SIZE / size_of::<u64>()];
|
|
|
|
let size = unsafe {
|
|
build_block_entropy_stats_sequences(
|
|
&seq_store,
|
|
&prev,
|
|
&mut next,
|
|
1,
|
|
&mut metadata,
|
|
workspace.as_mut_ptr().cast::<c_void>(),
|
|
ENTROPY_WORKSPACE_SIZE,
|
|
)
|
|
};
|
|
assert_eq!(size, 3);
|
|
assert_eq!(metadata.llType, SET_RLE);
|
|
assert_eq!(metadata.ofType, SET_RLE);
|
|
assert_eq!(metadata.mlType, SET_RLE);
|
|
assert_eq!(metadata.lastCountSize, 0);
|
|
assert_eq!(&metadata.fseTablesBuffer[..3], &[2, 3, 3]);
|
|
assert_eq!(next.litlength_repeatMode, FSE_REPEAT_NONE);
|
|
assert_eq!(next.offcode_repeatMode, FSE_REPEAT_NONE);
|
|
assert_eq!(next.matchlength_repeatMode, FSE_REPEAT_NONE);
|
|
}
|
|
|
|
#[test]
|
|
fn empty_seq_store_emits_raw_literals_and_repeats_tables() {
|
|
let mut store = TestSeqStore::new(Vec::new(), b"hello".to_vec());
|
|
let seq_store = store.seq_store();
|
|
let prev = ZSTD_entropyCTables_t {
|
|
huf: empty_huf_tables(HUF_REPEAT_NONE),
|
|
fse: empty_fse_tables(2),
|
|
};
|
|
let mut next = ZSTD_entropyCTables_t {
|
|
huf: empty_huf_tables(HUF_REPEAT_NONE),
|
|
fse: empty_fse_tables(FSE_REPEAT_NONE),
|
|
};
|
|
let mut workspace = vec![0u64; ENTROPY_WORKSPACE_SIZE / size_of::<u64>()];
|
|
let mut output = [0xA5u8; 32];
|
|
|
|
let size = unsafe {
|
|
ZSTD_rust_entropyCompressSeqStore_internal(
|
|
output.as_mut_ptr().cast::<c_void>(),
|
|
output.len(),
|
|
seq_store.litStart.cast::<c_void>(),
|
|
5,
|
|
&seq_store,
|
|
&prev,
|
|
&mut next,
|
|
1,
|
|
0,
|
|
workspace.as_mut_ptr().cast::<c_void>(),
|
|
ENTROPY_WORKSPACE_SIZE,
|
|
0,
|
|
)
|
|
};
|
|
// Raw literals header (5 << 3), five literal bytes, nbSeq == 0.
|
|
assert_eq!(size, 7);
|
|
assert_eq!(&output[..7], &[0x28, b'h', b'e', b'l', b'l', b'o', 0x00]);
|
|
// With no sequences, the previous FSE tables are copied unchanged.
|
|
assert_eq!(next.fse.offcode_repeatMode, 2);
|
|
assert_eq!(next.fse.matchlength_repeatMode, 2);
|
|
assert_eq!(next.fse.litlength_repeatMode, 2);
|
|
}
|
|
|
|
#[test]
|
|
fn literal_stats_pick_rle_and_compressed_types() {
|
|
let mut workspace = vec![0u64; ENTROPY_WORKSPACE_SIZE / size_of::<u64>()];
|
|
let prev = empty_huf_tables(HUF_REPEAT_NONE);
|
|
let mut next = empty_huf_tables(HUF_REPEAT_NONE);
|
|
let mut metadata =
|
|
unsafe { std::mem::MaybeUninit::<ZSTD_hufCTablesMetadata_t>::zeroed().assume_init() };
|
|
|
|
let rle = [7u8; 128];
|
|
let size = unsafe {
|
|
build_block_entropy_stats_literals(
|
|
rle.as_ptr().cast::<c_void>(),
|
|
rle.len(),
|
|
&prev,
|
|
&mut next,
|
|
&mut metadata,
|
|
0,
|
|
workspace.as_mut_ptr().cast::<c_void>(),
|
|
ENTROPY_WORKSPACE_SIZE,
|
|
0,
|
|
)
|
|
};
|
|
assert_eq!(size, 0);
|
|
assert_eq!(metadata.hType, SET_RLE);
|
|
|
|
let mut compressible = [0u8; 1024];
|
|
for (index, byte) in compressible.iter_mut().enumerate() {
|
|
*byte = (index % 4) as u8;
|
|
}
|
|
let size = unsafe {
|
|
build_block_entropy_stats_literals(
|
|
compressible.as_ptr().cast::<c_void>(),
|
|
compressible.len(),
|
|
&prev,
|
|
&mut next,
|
|
&mut metadata,
|
|
0,
|
|
workspace.as_mut_ptr().cast::<c_void>(),
|
|
ENTROPY_WORKSPACE_SIZE,
|
|
0,
|
|
)
|
|
};
|
|
assert!(!ERR_isError(size));
|
|
assert!(size > 0);
|
|
assert_eq!(metadata.hType, SET_COMPRESSED);
|
|
assert_eq!(metadata.hufDesSize, 0); // caller stores the returned size
|
|
assert_eq!(next.repeatMode, HUF_REPEAT_CHECK);
|
|
}
|
|
|
|
#[test]
|
|
fn copied_sequences_resolve_repcodes_and_append_the_delimiter() {
|
|
let mut store = TestSeqStore::new(
|
|
vec![
|
|
SeqDef {
|
|
offBase: 3 + 100, // raw offset 100
|
|
litLength: 4,
|
|
mlBase: 5,
|
|
},
|
|
SeqDef {
|
|
offBase: 1, // repcode 1 with literals: rep[0] == 100
|
|
litLength: 2,
|
|
mlBase: 7,
|
|
},
|
|
SeqDef {
|
|
offBase: 3, // repcode 3 without literals: rep[0] - 1 == 99
|
|
litLength: 0,
|
|
mlBase: 9,
|
|
},
|
|
],
|
|
vec![0; 10],
|
|
);
|
|
let seq_store = store.seq_store();
|
|
let mut out = [ZSTD_Sequence {
|
|
offset: 0xAA,
|
|
litLength: 0xAA,
|
|
matchLength: 0xAA,
|
|
rep: 0xAA,
|
|
}; 8];
|
|
let mut collector = SeqCollector {
|
|
collectSequences: 1,
|
|
seqStart: out.as_mut_ptr(),
|
|
seqIndex: 1,
|
|
maxSequences: 8,
|
|
};
|
|
let reps = [1u32, 4, 8];
|
|
|
|
let result =
|
|
unsafe { ZSTD_rust_copyBlockSequences(&mut collector, &seq_store, reps.as_ptr()) };
|
|
assert_eq!(result, 0);
|
|
assert_eq!(collector.seqIndex, 5);
|
|
let expect = [
|
|
(100, 4, 5 + 3, 0),
|
|
(100, 2, 7 + 3, 1),
|
|
(99, 0, 9 + 3, 3),
|
|
(0, 10 - 6, 0, 0xAA), // delimiter: last literals; rep untouched
|
|
];
|
|
for (index, &(offset, lit, matched, rep)) in expect.iter().enumerate() {
|
|
assert_eq!(out[1 + index].offset, offset, "offset {index}");
|
|
assert_eq!(out[1 + index].litLength, lit, "litLength {index}");
|
|
assert_eq!(out[1 + index].matchLength, matched, "matchLength {index}");
|
|
assert_eq!(out[1 + index].rep, rep, "rep {index}");
|
|
}
|
|
|
|
collector.maxSequences = 5;
|
|
let result =
|
|
unsafe { ZSTD_rust_copyBlockSequences(&mut collector, &seq_store, reps.as_ptr()) };
|
|
assert_eq!(result, ERROR(ZstdErrorCode::DstSizeTooSmall));
|
|
}
|
|
|
|
#[test]
|
|
fn finalize_off_base_preserves_repcode_boundaries() {
|
|
let reps = [10u32, 20, 30];
|
|
let finalize =
|
|
|raw_offset, ll0| unsafe { ZSTD_rust_finalizeOffBase(raw_offset, reps.as_ptr(), ll0) };
|
|
|
|
assert_eq!(finalize(10, 0), 1); // repcode 1 with literals
|
|
assert_eq!(finalize(10, 1), 13); // raw offset, not repcode 1
|
|
assert_eq!(finalize(20, 0), 2); // repcode 2 with literals
|
|
assert_eq!(finalize(20, 1), 1); // repcode 1 without literals
|
|
assert_eq!(finalize(30, 0), 3); // repcode 3 with literals
|
|
assert_eq!(finalize(30, 1), 2); // repcode 2 without literals
|
|
assert_eq!(finalize(9, 1), 3); // repcode 3's rep[0] - 1 form
|
|
assert_eq!(finalize(31, 0), 34); // ordinary raw offset
|
|
}
|
|
|
|
#[test]
|
|
fn validate_sequence_accepts_valid_input_and_caps_at_window_size() {
|
|
let result = ZSTD_rust_validateSequence(1027, 4, 4, 2048, 10, 99, 0);
|
|
assert_eq!(result, 0);
|
|
}
|
|
|
|
#[test]
|
|
fn validate_sequence_rejects_oversized_offset() {
|
|
let result = ZSTD_rust_validateSequence(1028, 4, 4, 2048, 10, 99, 0);
|
|
assert_eq!(result, ERROR(ZstdErrorCode::ExternalSequencesInvalid));
|
|
}
|
|
|
|
#[test]
|
|
fn validate_sequence_rejects_too_short_match() {
|
|
let result = ZSTD_rust_validateSequence(103, 3, 4, 100, 10, 0, 0);
|
|
assert_eq!(result, ERROR(ZstdErrorCode::ExternalSequencesInvalid));
|
|
}
|
|
|
|
#[test]
|
|
fn validate_sequence_accepts_three_byte_match_for_min_match_three() {
|
|
let result = ZSTD_rust_validateSequence(103, 3, 3, 100, 10, 0, 0);
|
|
assert_eq!(result, 0);
|
|
}
|
|
|
|
#[test]
|
|
fn validate_sequence_accepts_three_byte_match_from_sequence_producer() {
|
|
let result = ZSTD_rust_validateSequence(103, 3, 4, 100, 10, 0, 1);
|
|
assert_eq!(result, 0);
|
|
}
|
|
|
|
#[test]
|
|
fn determine_block_size_without_delimiters_returns_minimum() {
|
|
let result = unsafe { ZSTD_rust_determineBlockSize(0, 128, 50, ptr::null(), 0, 0) };
|
|
assert_eq!(result, 50);
|
|
|
|
let result = unsafe { ZSTD_rust_determineBlockSize(0, 32, 50, ptr::null(), 0, 0) };
|
|
assert_eq!(result, 32);
|
|
}
|
|
|
|
#[test]
|
|
fn explicit_delimiter_size_includes_delimiter_literals() {
|
|
let sequences = [
|
|
ZSTD_Sequence {
|
|
offset: 7,
|
|
litLength: 3,
|
|
matchLength: 4,
|
|
rep: 0,
|
|
},
|
|
ZSTD_Sequence {
|
|
offset: 0,
|
|
litLength: 5,
|
|
matchLength: 0,
|
|
rep: 0,
|
|
},
|
|
];
|
|
let result =
|
|
unsafe { ZSTD_rust_blockSizeExplicitDelimiter(sequences.as_ptr(), sequences.len(), 0) };
|
|
assert_eq!(result, 12);
|
|
|
|
let result = unsafe {
|
|
ZSTD_rust_determineBlockSize(1, 12, 12, sequences.as_ptr(), sequences.len(), 0)
|
|
};
|
|
assert_eq!(result, 12);
|
|
}
|
|
|
|
#[test]
|
|
fn explicit_delimiter_size_rejects_missing_delimiter() {
|
|
let sequences = [ZSTD_Sequence {
|
|
offset: 7,
|
|
litLength: 3,
|
|
matchLength: 4,
|
|
rep: 0,
|
|
}];
|
|
let result =
|
|
unsafe { ZSTD_rust_blockSizeExplicitDelimiter(sequences.as_ptr(), sequences.len(), 0) };
|
|
assert_eq!(result, ERROR(ZstdErrorCode::ExternalSequencesInvalid));
|
|
}
|
|
|
|
#[test]
|
|
fn explicit_delimiter_size_rejects_match_length() {
|
|
let sequences = [ZSTD_Sequence {
|
|
offset: 0,
|
|
litLength: 3,
|
|
matchLength: 4,
|
|
rep: 0,
|
|
}];
|
|
let result =
|
|
unsafe { ZSTD_rust_blockSizeExplicitDelimiter(sequences.as_ptr(), sequences.len(), 0) };
|
|
assert_eq!(result, ERROR(ZstdErrorCode::ExternalSequencesInvalid));
|
|
}
|
|
|
|
#[test]
|
|
fn determine_block_size_rejects_too_large_block() {
|
|
let sequences = [
|
|
ZSTD_Sequence {
|
|
offset: 7,
|
|
litLength: 3,
|
|
matchLength: 4,
|
|
rep: 0,
|
|
},
|
|
ZSTD_Sequence {
|
|
offset: 0,
|
|
litLength: 1,
|
|
matchLength: 0,
|
|
rep: 0,
|
|
},
|
|
];
|
|
let result = unsafe {
|
|
ZSTD_rust_determineBlockSize(1, 7, 8, sequences.as_ptr(), sequences.len(), 0)
|
|
};
|
|
assert_eq!(result, ERROR(ZstdErrorCode::ExternalSequencesInvalid));
|
|
}
|
|
|
|
#[test]
|
|
fn determine_block_size_rejects_too_long_frame() {
|
|
let sequences = [
|
|
ZSTD_Sequence {
|
|
offset: 7,
|
|
litLength: 3,
|
|
matchLength: 4,
|
|
rep: 0,
|
|
},
|
|
ZSTD_Sequence {
|
|
offset: 0,
|
|
litLength: 1,
|
|
matchLength: 0,
|
|
rep: 0,
|
|
},
|
|
];
|
|
let result = unsafe {
|
|
ZSTD_rust_determineBlockSize(1, 8, 7, sequences.as_ptr(), sequences.len(), 0)
|
|
};
|
|
assert_eq!(result, ERROR(ZstdErrorCode::ExternalSequencesInvalid));
|
|
}
|
|
}
|