marsultor/email-tracker
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

should be it

Browse Source
This commit is contained in:
Mars Ultor
2025-10-24 19:21:19 -05:00
parent a4b23fc57c
commit f09560c7b1
14047 changed files with 3161551 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

54
external/duckdb/third_party/snappy/LICENSE vendored Normal file
View File

@@ -0,0 +1,54 @@
Copyright 2011, Google Inc.
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above
copyright notice, this list of conditions and the following disclaimer
in the documentation and/or other materials provided with the
distribution.
* Neither the name of Google Inc. nor the names of its
contributors may be used to endorse or promote products derived from
this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
===
Some of the benchmark data in testdata/ is licensed differently:
- fireworks.jpeg is Copyright 2013 Steinar H. Gunderson, and
is licensed under the Creative Commons Attribution 3.0 license
(CC-BY-3.0). See https://creativecommons.org/licenses/by/3.0/
for more information.
- kppkn.gtb is taken from the Gaviota chess tablebase set, and
is licensed under the MIT License. See
https://sites.google.com/site/gaviotachessengine/Home/endgame-tablebases-1
for more information.
- paper-100k.pdf is an excerpt (bytes 92160 to 194560) from the paper
“Combinatorial Modeling of Chromatin Features Quantitatively Predicts DNA
Replication Timing in _Drosophila_” by Federico Comoglio and Renato Paro,
which is licensed under the CC-BY license. See
http://www.ploscompbiol.org/static/license for more ifnormation.
- alice29.txt, asyoulik.txt, plrabn12.txt and lcet10.txt are from Project
Gutenberg. The first three have expired copyrights and are in the public
domain; the latter does not have expired copyright, but is still in the
public domain according to the license information
(http://www.gutenberg.org/ebooks/53).

647
external/duckdb/third_party/snappy/snappy-internal.h vendored Executable file
View File

@@ -0,0 +1,647 @@
// Copyright 2008 Google Inc. All Rights Reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following disclaimer
// in the documentation and/or other materials provided with the
// distribution.
// * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// Internals shared between the Snappy implementation and its unittest.
#ifndef THIRD_PARTY_SNAPPY_SNAPPY_INTERNAL_H_
#define THIRD_PARTY_SNAPPY_SNAPPY_INTERNAL_H_
#include "snappy_version.hpp"
#if SNAPPY_NEW_VERSION
#include <utility>
#include "snappy-stubs-internal.h"
#if SNAPPY_HAVE_SSSE3
// Please do not replace with <x86intrin.h> or with headers that assume more
// advanced SSE versions without checking with all the OWNERS.
#include <emmintrin.h>
#include <tmmintrin.h>
#endif
#if SNAPPY_HAVE_NEON
#include <arm_neon.h>
#endif
#if SNAPPY_HAVE_SSSE3 || SNAPPY_HAVE_NEON
#define SNAPPY_HAVE_VECTOR_BYTE_SHUFFLE 1
#else
#define SNAPPY_HAVE_VECTOR_BYTE_SHUFFLE 0
#endif
namespace duckdb_snappy {
namespace internal {
#if SNAPPY_HAVE_VECTOR_BYTE_SHUFFLE
#if SNAPPY_HAVE_SSSE3
using V128 = __m128i;
#elif SNAPPY_HAVE_NEON
using V128 = uint8x16_t;
#endif
// Load 128 bits of integer data. `src` must be 16-byte aligned.
inline V128 V128_Load(const V128* src);
// Load 128 bits of integer data. `src` does not need to be aligned.
inline V128 V128_LoadU(const V128* src);
// Store 128 bits of integer data. `dst` does not need to be aligned.
inline void V128_StoreU(V128* dst, V128 val);
// Shuffle packed 8-bit integers using a shuffle mask.
// Each packed integer in the shuffle mask must be in [0,16).
inline V128 V128_Shuffle(V128 input, V128 shuffle_mask);
// Constructs V128 with 16 chars |c|.
inline V128 V128_DupChar(char c);
#if SNAPPY_HAVE_SSSE3
inline V128 V128_Load(const V128* src) { return _mm_load_si128(src); }
inline V128 V128_LoadU(const V128* src) { return _mm_loadu_si128(src); }
inline void V128_StoreU(V128* dst, V128 val) { _mm_storeu_si128(dst, val); }
inline V128 V128_Shuffle(V128 input, V128 shuffle_mask) {
return _mm_shuffle_epi8(input, shuffle_mask);
}
inline V128 V128_DupChar(char c) { return _mm_set1_epi8(c); }
#elif SNAPPY_HAVE_NEON
inline V128 V128_Load(const V128* src) {
return vld1q_u8(reinterpret_cast<const uint8_t*>(src));
}
inline V128 V128_LoadU(const V128* src) {
return vld1q_u8(reinterpret_cast<const uint8_t*>(src));
}
inline void V128_StoreU(V128* dst, V128 val) {
vst1q_u8(reinterpret_cast<uint8_t*>(dst), val);
}
inline V128 V128_Shuffle(V128 input, V128 shuffle_mask) {
assert(vminvq_u8(shuffle_mask) >= 0 && vmaxvq_u8(shuffle_mask) <= 15);
return vqtbl1q_u8(input, shuffle_mask);
}
inline V128 V128_DupChar(char c) { return vdupq_n_u8(c); }
#endif
#endif // SNAPPY_HAVE_VECTOR_BYTE_SHUFFLE
// Working memory performs a single allocation to hold all scratch space
// required for compression.
class WorkingMemory {
public:
explicit WorkingMemory(size_t input_size);
~WorkingMemory();
// Allocates and clears a hash table using memory in "*this",
// stores the number of buckets in "*table_size" and returns a pointer to
// the base of the hash table.
uint16_t* GetHashTable(size_t fragment_size, int* table_size) const;
char* GetScratchInput() const { return input_; }
char* GetScratchOutput() const { return output_; }
private:
char* mem_; // the allocated memory, never nullptr
size_t size_; // the size of the allocated memory, never 0
uint16_t* table_; // the pointer to the hashtable
char* input_; // the pointer to the input scratch buffer
char* output_; // the pointer to the output scratch buffer
// No copying
WorkingMemory(const WorkingMemory&);
void operator=(const WorkingMemory&);
};
// Flat array compression that does not emit the "uncompressed length"
// prefix. Compresses "input" string to the "*op" buffer.
//
// REQUIRES: "input_length <= kBlockSize"
// REQUIRES: "op" points to an array of memory that is at least
// "MaxCompressedLength(input_length)" in size.
// REQUIRES: All elements in "table[0..table_size-1]" are initialized to zero.
// REQUIRES: "table_size" is a power of two
//
// Returns an "end" pointer into "op" buffer.
// "end - op" is the compressed size of "input".
char* CompressFragment(const char* input,
size_t input_length,
char* op,
uint16_t* table,
const int table_size);
// Find the largest n such that
//
// s1[0,n-1] == s2[0,n-1]
// and n <= (s2_limit - s2).
//
// Return make_pair(n, n < 8).
// Does not read *s2_limit or beyond.
// Does not read *(s1 + (s2_limit - s2)) or beyond.
// Requires that s2_limit >= s2.
//
// In addition populate *data with the next 5 bytes from the end of the match.
// This is only done if 8 bytes are available (s2_limit - s2 >= 8). The point is
// that on some arch's this can be done faster in this routine than subsequent
// loading from s2 + n.
//
// Separate implementation for 64-bit, little-endian cpus.
#if !SNAPPY_IS_BIG_ENDIAN && \
(defined(__x86_64__) || defined(_M_X64) || defined(ARCH_PPC) || \
defined(ARCH_ARM))
static inline std::pair<size_t, bool> FindMatchLength(const char* s1,
const char* s2,
const char* s2_limit,
uint64_t* data) {
assert(s2_limit >= s2);
size_t matched = 0;
// This block isn't necessary for correctness; we could just start looping
// immediately. As an optimization though, it is useful. It creates some not
// uncommon code paths that determine, without extra effort, whether the match
// length is less than 8. In short, we are hoping to avoid a conditional
// branch, and perhaps get better code layout from the C++ compiler.
if (SNAPPY_PREDICT_TRUE(s2 <= s2_limit - 16)) {
uint64_t a1 = UNALIGNED_LOAD64(s1);
uint64_t a2 = UNALIGNED_LOAD64(s2);
if (SNAPPY_PREDICT_TRUE(a1 != a2)) {
// This code is critical for performance. The reason is that it determines
// how much to advance `ip` (s2). This obviously depends on both the loads
// from the `candidate` (s1) and `ip`. Furthermore the next `candidate`
// depends on the advanced `ip` calculated here through a load, hash and
// new candidate hash lookup (a lot of cycles). This makes s1 (ie.
// `candidate`) the variable that limits throughput. This is the reason we
// go through hoops to have this function update `data` for the next iter.
// The straightforward code would use *data, given by
//
// *data = UNALIGNED_LOAD64(s2 + matched_bytes) (Latency of 5 cycles),
//
// as input for the hash table lookup to find next candidate. However
// this forces the load on the data dependency chain of s1, because
// matched_bytes directly depends on s1. However matched_bytes is 0..7, so
// we can also calculate *data by
//
// *data = AlignRight(UNALIGNED_LOAD64(s2), UNALIGNED_LOAD64(s2 + 8),
// matched_bytes);
//
// The loads do not depend on s1 anymore and are thus off the bottleneck.
// The straightforward implementation on x86_64 would be to use
//
// shrd rax, rdx, cl (cl being matched_bytes * 8)
//
// unfortunately shrd with a variable shift has a 4 cycle latency. So this
// only wins 1 cycle. The BMI2 shrx instruction is a 1 cycle variable
// shift instruction but can only shift 64 bits. If we focus on just
// obtaining the least significant 4 bytes, we can obtain this by
//
// *data = ConditionalMove(matched_bytes < 4, UNALIGNED_LOAD64(s2),
// UNALIGNED_LOAD64(s2 + 4) >> ((matched_bytes & 3) * 8);
//
// Writen like above this is not a big win, the conditional move would be
// a cmp followed by a cmov (2 cycles) followed by a shift (1 cycle).
// However matched_bytes < 4 is equal to
// static_cast<uint32_t>(xorval) != 0. Writen that way, the conditional
// move (2 cycles) can execute in parallel with FindLSBSetNonZero64
// (tzcnt), which takes 3 cycles.
uint64_t xorval = a1 ^ a2;
int shift = Bits::FindLSBSetNonZero64(xorval);
size_t matched_bytes = shift >> 3;
uint64_t a3 = UNALIGNED_LOAD64(s2 + 4);
#ifndef __x86_64__
a2 = static_cast<uint32_t>(xorval) == 0 ? a3 : a2;
#else
// Ideally this would just be
//
// a2 = static_cast<uint32_t>(xorval) == 0 ? a3 : a2;
//
// However clang correctly infers that the above statement participates on
// a critical data dependency chain and thus, unfortunately, refuses to
// use a conditional move (it's tuned to cut data dependencies). In this
// case there is a longer parallel chain anyway AND this will be fairly
// unpredictable.
asm("testl %k2, %k2\n\t"
"cmovzq %1, %0\n\t"
: "+r"(a2)
: "r"(a3), "r"(xorval)
: "cc");
#endif
*data = a2 >> (shift & (3 * 8));
return std::pair<size_t, bool>(matched_bytes, true);
} else {
matched = 8;
s2 += 8;
}
}
SNAPPY_PREFETCH(s1 + 64);
SNAPPY_PREFETCH(s2 + 64);
// Find out how long the match is. We loop over the data 64 bits at a
// time until we find a 64-bit block that doesn't match; then we find
// the first non-matching bit and use that to calculate the total
// length of the match.
while (SNAPPY_PREDICT_TRUE(s2 <= s2_limit - 16)) {
uint64_t a1 = UNALIGNED_LOAD64(s1 + matched);
uint64_t a2 = UNALIGNED_LOAD64(s2);
if (a1 == a2) {
s2 += 8;
matched += 8;
} else {
uint64_t xorval = a1 ^ a2;
int shift = Bits::FindLSBSetNonZero64(xorval);
size_t matched_bytes = shift >> 3;
uint64_t a3 = UNALIGNED_LOAD64(s2 + 4);
#ifndef __x86_64__
a2 = static_cast<uint32_t>(xorval) == 0 ? a3 : a2;
#else
asm("testl %k2, %k2\n\t"
"cmovzq %1, %0\n\t"
: "+r"(a2)
: "r"(a3), "r"(xorval)
: "cc");
#endif
*data = a2 >> (shift & (3 * 8));
matched += matched_bytes;
assert(matched >= 8);
return std::pair<size_t, bool>(matched, false);
}
}
while (SNAPPY_PREDICT_TRUE(s2 < s2_limit)) {
if (s1[matched] == *s2) {
++s2;
++matched;
} else {
if (s2 <= s2_limit - 8) {
*data = UNALIGNED_LOAD64(s2);
}
return std::pair<size_t, bool>(matched, matched < 8);
}
}
return std::pair<size_t, bool>(matched, matched < 8);
}
#else
static inline std::pair<size_t, bool> FindMatchLength(const char* s1,
const char* s2,
const char* s2_limit,
uint64_t* data) {
// Implementation based on the x86-64 version, above.
assert(s2_limit >= s2);
int matched = 0;
while (s2 <= s2_limit - 4 &&
UNALIGNED_LOAD32(s2) == UNALIGNED_LOAD32(s1 + matched)) {
s2 += 4;
matched += 4;
}
if (LittleEndian::IsLittleEndian() && s2 <= s2_limit - 4) {
uint32_t x = UNALIGNED_LOAD32(s2) ^ UNALIGNED_LOAD32(s1 + matched);
int matching_bits = Bits::FindLSBSetNonZero(x);
matched += matching_bits >> 3;
s2 += matching_bits >> 3;
} else {
while ((s2 < s2_limit) && (s1[matched] == *s2)) {
++s2;
++matched;
}
}
if (s2 <= s2_limit - 8) *data = LittleEndian::Load64(s2);
return std::pair<size_t, bool>(matched, matched < 8);
}
#endif
static inline size_t FindMatchLengthPlain(const char* s1, const char* s2,
const char* s2_limit) {
// Implementation based on the x86-64 version, above.
assert(s2_limit >= s2);
int matched = 0;
while (s2 <= s2_limit - 8 &&
UNALIGNED_LOAD64(s2) == UNALIGNED_LOAD64(s1 + matched)) {
s2 += 8;
matched += 8;
}
if (LittleEndian::IsLittleEndian() && s2 <= s2_limit - 8) {
uint64_t x = UNALIGNED_LOAD64(s2) ^ UNALIGNED_LOAD64(s1 + matched);
int matching_bits = Bits::FindLSBSetNonZero64(x);
matched += matching_bits >> 3;
s2 += matching_bits >> 3;
} else {
while ((s2 < s2_limit) && (s1[matched] == *s2)) {
++s2;
++matched;
}
}
return matched;
}
// Lookup tables for decompression code. Give --snappy_dump_decompression_table
// to the unit test to recompute char_table.
enum {
LITERAL = 0,
COPY_1_BYTE_OFFSET = 1, // 3 bit length + 3 bits of offset in opcode
COPY_2_BYTE_OFFSET = 2,
COPY_4_BYTE_OFFSET = 3
};
static const int kMaximumTagLength = 5; // COPY_4_BYTE_OFFSET plus the actual offset.
// Data stored per entry in lookup table:
// Range Bits-used Description
// ------------------------------------
// 1..64 0..7 Literal/copy length encoded in opcode byte
// 0..7 8..10 Copy offset encoded in opcode byte / 256
// 0..4 11..13 Extra bytes after opcode
//
// We use eight bits for the length even though 7 would have sufficed
// because of efficiency reasons:
// (1) Extracting a byte is faster than a bit-field
// (2) It properly aligns copy offset so we do not need a <<8
static constexpr uint16_t char_table[256] = {
// clang-format off
0x0001, 0x0804, 0x1001, 0x2001, 0x0002, 0x0805, 0x1002, 0x2002,
0x0003, 0x0806, 0x1003, 0x2003, 0x0004, 0x0807, 0x1004, 0x2004,
0x0005, 0x0808, 0x1005, 0x2005, 0x0006, 0x0809, 0x1006, 0x2006,
0x0007, 0x080a, 0x1007, 0x2007, 0x0008, 0x080b, 0x1008, 0x2008,
0x0009, 0x0904, 0x1009, 0x2009, 0x000a, 0x0905, 0x100a, 0x200a,
0x000b, 0x0906, 0x100b, 0x200b, 0x000c, 0x0907, 0x100c, 0x200c,
0x000d, 0x0908, 0x100d, 0x200d, 0x000e, 0x0909, 0x100e, 0x200e,
0x000f, 0x090a, 0x100f, 0x200f, 0x0010, 0x090b, 0x1010, 0x2010,
0x0011, 0x0a04, 0x1011, 0x2011, 0x0012, 0x0a05, 0x1012, 0x2012,
0x0013, 0x0a06, 0x1013, 0x2013, 0x0014, 0x0a07, 0x1014, 0x2014,
0x0015, 0x0a08, 0x1015, 0x2015, 0x0016, 0x0a09, 0x1016, 0x2016,
0x0017, 0x0a0a, 0x1017, 0x2017, 0x0018, 0x0a0b, 0x1018, 0x2018,
0x0019, 0x0b04, 0x1019, 0x2019, 0x001a, 0x0b05, 0x101a, 0x201a,
0x001b, 0x0b06, 0x101b, 0x201b, 0x001c, 0x0b07, 0x101c, 0x201c,
0x001d, 0x0b08, 0x101d, 0x201d, 0x001e, 0x0b09, 0x101e, 0x201e,
0x001f, 0x0b0a, 0x101f, 0x201f, 0x0020, 0x0b0b, 0x1020, 0x2020,
0x0021, 0x0c04, 0x1021, 0x2021, 0x0022, 0x0c05, 0x1022, 0x2022,
0x0023, 0x0c06, 0x1023, 0x2023, 0x0024, 0x0c07, 0x1024, 0x2024,
0x0025, 0x0c08, 0x1025, 0x2025, 0x0026, 0x0c09, 0x1026, 0x2026,
0x0027, 0x0c0a, 0x1027, 0x2027, 0x0028, 0x0c0b, 0x1028, 0x2028,
0x0029, 0x0d04, 0x1029, 0x2029, 0x002a, 0x0d05, 0x102a, 0x202a,
0x002b, 0x0d06, 0x102b, 0x202b, 0x002c, 0x0d07, 0x102c, 0x202c,
0x002d, 0x0d08, 0x102d, 0x202d, 0x002e, 0x0d09, 0x102e, 0x202e,
0x002f, 0x0d0a, 0x102f, 0x202f, 0x0030, 0x0d0b, 0x1030, 0x2030,
0x0031, 0x0e04, 0x1031, 0x2031, 0x0032, 0x0e05, 0x1032, 0x2032,
0x0033, 0x0e06, 0x1033, 0x2033, 0x0034, 0x0e07, 0x1034, 0x2034,
0x0035, 0x0e08, 0x1035, 0x2035, 0x0036, 0x0e09, 0x1036, 0x2036,
0x0037, 0x0e0a, 0x1037, 0x2037, 0x0038, 0x0e0b, 0x1038, 0x2038,
0x0039, 0x0f04, 0x1039, 0x2039, 0x003a, 0x0f05, 0x103a, 0x203a,
0x003b, 0x0f06, 0x103b, 0x203b, 0x003c, 0x0f07, 0x103c, 0x203c,
0x0801, 0x0f08, 0x103d, 0x203d, 0x1001, 0x0f09, 0x103e, 0x203e,
0x1801, 0x0f0a, 0x103f, 0x203f, 0x2001, 0x0f0b, 0x1040, 0x2040,
// clang-format on
};
} // end namespace internal
} // end namespace duckdb_snappy
#else // #if SNAPPY_NEW_VERSION
#include "snappy-stubs-internal.h"
namespace duckdb_snappy {
namespace internal {
// Working memory performs a single allocation to hold all scratch space
// required for compression.
class WorkingMemory {
public:
explicit WorkingMemory(size_t input_size);
~WorkingMemory();
// Allocates and clears a hash table using memory in "*this",
// stores the number of buckets in "*table_size" and returns a pointer to
// the base of the hash table.
uint16* GetHashTable(size_t fragment_size, int* table_size) const;
char* GetScratchInput() const { return input_; }
char* GetScratchOutput() const { return output_; }
private:
char* mem_; // the allocated memory, never nullptr
size_t size_; // the size of the allocated memory, never 0
uint16* table_; // the pointer to the hashtable
char* input_; // the pointer to the input scratch buffer
char* output_; // the pointer to the output scratch buffer
// No copying
WorkingMemory(const WorkingMemory&);
void operator=(const WorkingMemory&);
};
// Flat array compression that does not emit the "uncompressed length"
// prefix. Compresses "input" string to the "*op" buffer.
//
// REQUIRES: "input_length <= kBlockSize"
// REQUIRES: "op" points to an array of memory that is at least
// "MaxCompressedLength(input_length)" in size.
// REQUIRES: All elements in "table[0..table_size-1]" are initialized to zero.
// REQUIRES: "table_size" is a power of two
//
// Returns an "end" pointer into "op" buffer.
// "end - op" is the compressed size of "input".
char* CompressFragment(const char* input,
size_t input_length,
char* op,
uint16* table,
const int table_size);
// Find the largest n such that
//
// s1[0,n-1] == s2[0,n-1]
// and n <= (s2_limit - s2).
//
// Return make_pair(n, n < 8).
// Does not read *s2_limit or beyond.
// Does not read *(s1 + (s2_limit - s2)) or beyond.
// Requires that s2_limit >= s2.
//
// Separate implementation for 64-bit, little-endian cpus.
#if !defined(SNAPPY_IS_BIG_ENDIAN) && \
(defined(ARCH_K8) || defined(ARCH_PPC) || defined(ARCH_ARM))
static inline std::pair<size_t, bool> FindMatchLength(const char* s1,
const char* s2,
const char* s2_limit) {
assert(s2_limit >= s2);
size_t matched = 0;
// This block isn't necessary for correctness; we could just start looping
// immediately. As an optimization though, it is useful. It creates some not
// uncommon code paths that determine, without extra effort, whether the match
// length is less than 8. In short, we are hoping to avoid a conditional
// branch, and perhaps get better code layout from the C++ compiler.
if (SNAPPY_PREDICT_TRUE(s2 <= s2_limit - 8)) {
uint64 a1 = UNALIGNED_LOAD64(s1);
uint64 a2 = UNALIGNED_LOAD64(s2);
if (a1 != a2) {
return std::pair<size_t, bool>(Bits::FindLSBSetNonZero64(a1 ^ a2) >> 3,
true);
} else {
matched = 8;
s2 += 8;
}
}
// Find out how long the match is. We loop over the data 64 bits at a
// time until we find a 64-bit block that doesn't match; then we find
// the first non-matching bit and use that to calculate the total
// length of the match.
while (SNAPPY_PREDICT_TRUE(s2 <= s2_limit - 8)) {
if (UNALIGNED_LOAD64(s2) == UNALIGNED_LOAD64(s1 + matched)) {
s2 += 8;
matched += 8;
} else {
uint64 x = UNALIGNED_LOAD64(s2) ^ UNALIGNED_LOAD64(s1 + matched);
int matching_bits = Bits::FindLSBSetNonZero64(x);
matched += matching_bits >> 3;
assert(matched >= 8);
return std::pair<size_t, bool>(matched, false);
}
}
while (SNAPPY_PREDICT_TRUE(s2 < s2_limit)) {
if (s1[matched] == *s2) {
++s2;
++matched;
} else {
return std::pair<size_t, bool>(matched, matched < 8);
}
}
return std::pair<size_t, bool>(matched, matched < 8);
}
#else
static inline std::pair<size_t, bool> FindMatchLength(const char* s1,
const char* s2,
const char* s2_limit) {
// Implementation based on the x86-64 version, above.
assert(s2_limit >= s2);
int matched = 0;
while (s2 <= s2_limit - 4 &&
UNALIGNED_LOAD32(s2) == UNALIGNED_LOAD32(s1 + matched)) {
s2 += 4;
matched += 4;
}
if (LittleEndian::IsLittleEndian() && s2 <= s2_limit - 4) {
uint32 x = UNALIGNED_LOAD32(s2) ^ UNALIGNED_LOAD32(s1 + matched);
int matching_bits = Bits::FindLSBSetNonZero(x);
matched += matching_bits >> 3;
} else {
while ((s2 < s2_limit) && (s1[matched] == *s2)) {
++s2;
++matched;
}
}
return std::pair<size_t, bool>(matched, matched < 8);
}
#endif
// Lookup tables for decompression code. Give --snappy_dump_decompression_table
// to the unit test to recompute char_table.
enum {
LITERAL = 0,
COPY_1_BYTE_OFFSET = 1, // 3 bit length + 3 bits of offset in opcode
COPY_2_BYTE_OFFSET = 2,
COPY_4_BYTE_OFFSET = 3
};
static const int kMaximumTagLength = 5; // COPY_4_BYTE_OFFSET plus the actual offset.
// Data stored per entry in lookup table:
// Range Bits-used Description
// ------------------------------------
// 1..64 0..7 Literal/copy length encoded in opcode byte
// 0..7 8..10 Copy offset encoded in opcode byte / 256
// 0..4 11..13 Extra bytes after opcode
//
// We use eight bits for the length even though 7 would have sufficed
// because of efficiency reasons:
// (1) Extracting a byte is faster than a bit-field
// (2) It properly aligns copy offset so we do not need a <<8
static const uint16 char_table[256] = {
0x0001, 0x0804, 0x1001, 0x2001, 0x0002, 0x0805, 0x1002, 0x2002,
0x0003, 0x0806, 0x1003, 0x2003, 0x0004, 0x0807, 0x1004, 0x2004,
0x0005, 0x0808, 0x1005, 0x2005, 0x0006, 0x0809, 0x1006, 0x2006,
0x0007, 0x080a, 0x1007, 0x2007, 0x0008, 0x080b, 0x1008, 0x2008,
0x0009, 0x0904, 0x1009, 0x2009, 0x000a, 0x0905, 0x100a, 0x200a,
0x000b, 0x0906, 0x100b, 0x200b, 0x000c, 0x0907, 0x100c, 0x200c,
0x000d, 0x0908, 0x100d, 0x200d, 0x000e, 0x0909, 0x100e, 0x200e,
0x000f, 0x090a, 0x100f, 0x200f, 0x0010, 0x090b, 0x1010, 0x2010,
0x0011, 0x0a04, 0x1011, 0x2011, 0x0012, 0x0a05, 0x1012, 0x2012,
0x0013, 0x0a06, 0x1013, 0x2013, 0x0014, 0x0a07, 0x1014, 0x2014,
0x0015, 0x0a08, 0x1015, 0x2015, 0x0016, 0x0a09, 0x1016, 0x2016,
0x0017, 0x0a0a, 0x1017, 0x2017, 0x0018, 0x0a0b, 0x1018, 0x2018,
0x0019, 0x0b04, 0x1019, 0x2019, 0x001a, 0x0b05, 0x101a, 0x201a,
0x001b, 0x0b06, 0x101b, 0x201b, 0x001c, 0x0b07, 0x101c, 0x201c,
0x001d, 0x0b08, 0x101d, 0x201d, 0x001e, 0x0b09, 0x101e, 0x201e,
0x001f, 0x0b0a, 0x101f, 0x201f, 0x0020, 0x0b0b, 0x1020, 0x2020,
0x0021, 0x0c04, 0x1021, 0x2021, 0x0022, 0x0c05, 0x1022, 0x2022,
0x0023, 0x0c06, 0x1023, 0x2023, 0x0024, 0x0c07, 0x1024, 0x2024,
0x0025, 0x0c08, 0x1025, 0x2025, 0x0026, 0x0c09, 0x1026, 0x2026,
0x0027, 0x0c0a, 0x1027, 0x2027, 0x0028, 0x0c0b, 0x1028, 0x2028,
0x0029, 0x0d04, 0x1029, 0x2029, 0x002a, 0x0d05, 0x102a, 0x202a,
0x002b, 0x0d06, 0x102b, 0x202b, 0x002c, 0x0d07, 0x102c, 0x202c,
0x002d, 0x0d08, 0x102d, 0x202d, 0x002e, 0x0d09, 0x102e, 0x202e,
0x002f, 0x0d0a, 0x102f, 0x202f, 0x0030, 0x0d0b, 0x1030, 0x2030,
0x0031, 0x0e04, 0x1031, 0x2031, 0x0032, 0x0e05, 0x1032, 0x2032,
0x0033, 0x0e06, 0x1033, 0x2033, 0x0034, 0x0e07, 0x1034, 0x2034,
0x0035, 0x0e08, 0x1035, 0x2035, 0x0036, 0x0e09, 0x1036, 0x2036,
0x0037, 0x0e0a, 0x1037, 0x2037, 0x0038, 0x0e0b, 0x1038, 0x2038,
0x0039, 0x0f04, 0x1039, 0x2039, 0x003a, 0x0f05, 0x103a, 0x203a,
0x003b, 0x0f06, 0x103b, 0x203b, 0x003c, 0x0f07, 0x103c, 0x203c,
0x0801, 0x0f08, 0x103d, 0x203d, 0x1001, 0x0f09, 0x103e, 0x203e,
0x1801, 0x0f0a, 0x103f, 0x203f, 0x2001, 0x0f0b, 0x1040, 0x2040
};
} // end namespace internal
// The size of a compression block. Note that many parts of the compression
// code assumes that kBlockSize <= 65536; in particular, the hash table
// can only store 16-bit offsets, and EmitCopy() also assumes the offset
// is 65535 bytes or less. Note also that if you change this, it will
// affect the framing format (see framing_format.txt).
//
// Note that there might be older data around that is compressed with larger
// block sizes, so the decompression code should not rely on the
// non-existence of long backreferences.
static const int kBlockLog = 16;
static const size_t kBlockSize = 1 << kBlockLog;
static const int kMaxHashTableBits = 14;
static const size_t kMaxHashTableSize = 1 << kMaxHashTableBits;
} // end namespace duckdb_snappy
#endif // #if SNAPPY_NEW_VERSION # else
#endif // THIRD_PARTY_SNAPPY_SNAPPY_INTERNAL_H_

206
external/duckdb/third_party/snappy/snappy-sinksource.cc vendored Normal file
View File

@@ -0,0 +1,206 @@
// Copyright 2011 Google Inc. All Rights Reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following disclaimer
// in the documentation and/or other materials provided with the
// distribution.
// * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#include "snappy_version.hpp"
#if SNAPPY_NEW_VERSION
#include <stddef.h>
#include <cstring>
#include "snappy-sinksource.h"
namespace duckdb_snappy {
Source::~Source() = default;
Sink::~Sink() = default;
char* Sink::GetAppendBuffer(size_t length, char* scratch) {
// TODO: Switch to [[maybe_unused]] when we can assume C++17.
(void)length;
return scratch;
}
char* Sink::GetAppendBufferVariable(
size_t min_size, size_t desired_size_hint, char* scratch,
size_t scratch_size, size_t* allocated_size) {
// TODO: Switch to [[maybe_unused]] when we can assume C++17.
(void)min_size;
(void)desired_size_hint;
*allocated_size = scratch_size;
return scratch;
}
void Sink::AppendAndTakeOwnership(
char* bytes, size_t n,
void (*deleter)(void*, const char*, size_t),
void *deleter_arg) {
Append(bytes, n);
(*deleter)(deleter_arg, bytes, n);
}
ByteArraySource::~ByteArraySource() = default;
size_t ByteArraySource::Available() const { return left_; }
const char* ByteArraySource::Peek(size_t* len) {
*len = left_;
return ptr_;
}
void ByteArraySource::Skip(size_t n) {
left_ -= n;
ptr_ += n;
}
UncheckedByteArraySink::~UncheckedByteArraySink() { }
void UncheckedByteArraySink::Append(const char* data, size_t n) {
// Do no copying if the caller filled in the result of GetAppendBuffer()
if (data != dest_) {
std::memcpy(dest_, data, n);
}
dest_ += n;
}
char* UncheckedByteArraySink::GetAppendBuffer(size_t len, char* scratch) {
// TODO: Switch to [[maybe_unused]] when we can assume C++17.
(void)len;
(void)scratch;
return dest_;
}
void UncheckedByteArraySink::AppendAndTakeOwnership(
char* bytes, size_t n,
void (*deleter)(void*, const char*, size_t),
void *deleter_arg) {
if (bytes != dest_) {
std::memcpy(dest_, bytes, n);
(*deleter)(deleter_arg, bytes, n);
}
dest_ += n;
}
char* UncheckedByteArraySink::GetAppendBufferVariable(
size_t min_size, size_t desired_size_hint, char* scratch,
size_t scratch_size, size_t* allocated_size) {
// TODO: Switch to [[maybe_unused]] when we can assume C++17.
(void)min_size;
(void)scratch;
(void)scratch_size;
*allocated_size = desired_size_hint;
return dest_;
}
} // namespace duckdb_snappy
#else // #if SNAPPY_NEW_VERSION
#include <string.h>
#include "snappy-sinksource.h"
namespace duckdb_snappy {
Source::~Source() { }
Sink::~Sink() { }
char* Sink::GetAppendBuffer(size_t length, char* scratch) {
return scratch;
}
char* Sink::GetAppendBufferVariable(
size_t min_size, size_t desired_size_hint, char* scratch,
size_t scratch_size, size_t* allocated_size) {
*allocated_size = scratch_size;
return scratch;
}
void Sink::AppendAndTakeOwnership(
char* bytes, size_t n,
void (*deleter)(void*, const char*, size_t),
void *deleter_arg) {
Append(bytes, n);
(*deleter)(deleter_arg, bytes, n);
}
ByteArraySource::~ByteArraySource() { }
size_t ByteArraySource::Available() const { return left_; }
const char* ByteArraySource::Peek(size_t* len) {
*len = left_;
return ptr_;
}
void ByteArraySource::Skip(size_t n) {
left_ -= n;
ptr_ += n;
}
UncheckedByteArraySink::~UncheckedByteArraySink() { }
void UncheckedByteArraySink::Append(const char* data, size_t n) {
// Do no copying if the caller filled in the result of GetAppendBuffer()
if (data != dest_) {
memcpy(dest_, data, n);
}
dest_ += n;
}
char* UncheckedByteArraySink::GetAppendBuffer(size_t len, char* scratch) {
return dest_;
}
void UncheckedByteArraySink::AppendAndTakeOwnership(
char* data, size_t n,
void (*deleter)(void*, const char*, size_t),
void *deleter_arg) {
if (data != dest_) {
memcpy(dest_, data, n);
(*deleter)(deleter_arg, data, n);
}
dest_ += n;
}
char* UncheckedByteArraySink::GetAppendBufferVariable(
size_t min_size, size_t desired_size_hint, char* scratch,
size_t scratch_size, size_t* allocated_size) {
*allocated_size = desired_size_hint;
return dest_;
}
} // namespace duckdb_snappy
#endif // #if SNAPPY_NEW_VERSION # else

340
external/duckdb/third_party/snappy/snappy-sinksource.h vendored Normal file
View File

@@ -0,0 +1,340 @@
// Copyright 2011 Google Inc. All Rights Reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following disclaimer
// in the documentation and/or other materials provided with the
// distribution.
// * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#ifndef THIRD_PARTY_SNAPPY_SNAPPY_SINKSOURCE_H_
#define THIRD_PARTY_SNAPPY_SNAPPY_SINKSOURCE_H_
#include "snappy_version.hpp"
#if SNAPPY_NEW_VERSION
#include <stddef.h>
namespace duckdb_snappy {
// A Sink is an interface that consumes a sequence of bytes.
class Sink {
public:
Sink() { }
virtual ~Sink();
// Append "bytes[0,n-1]" to this.
virtual void Append(const char* bytes, size_t n) = 0;
// Returns a writable buffer of the specified length for appending.
// May return a pointer to the caller-owned scratch buffer which
// must have at least the indicated length. The returned buffer is
// only valid until the next operation on this Sink.
//
// After writing at most "length" bytes, call Append() with the
// pointer returned from this function and the number of bytes
// written. Many Append() implementations will avoid copying
// bytes if this function returned an internal buffer.
//
// If a non-scratch buffer is returned, the caller may only pass a
// prefix of it to Append(). That is, it is not correct to pass an
// interior pointer of the returned array to Append().
//
// The default implementation always returns the scratch buffer.
virtual char* GetAppendBuffer(size_t length, char* scratch);
// For higher performance, Sink implementations can provide custom
// AppendAndTakeOwnership() and GetAppendBufferVariable() methods.
// These methods can reduce the number of copies done during
// compression/decompression.
// Append "bytes[0,n-1] to the sink. Takes ownership of "bytes"
// and calls the deleter function as (*deleter)(deleter_arg, bytes, n)
// to free the buffer. deleter function must be non NULL.
//
// The default implementation just calls Append and frees "bytes".
// Other implementations may avoid a copy while appending the buffer.
virtual void AppendAndTakeOwnership(
char* bytes, size_t n, void (*deleter)(void*, const char*, size_t),
void *deleter_arg);
// Returns a writable buffer for appending and writes the buffer's capacity to
// *allocated_size. Guarantees *allocated_size >= min_size.
// May return a pointer to the caller-owned scratch buffer which must have
// scratch_size >= min_size.
//
// The returned buffer is only valid until the next operation
// on this ByteSink.
//
// After writing at most *allocated_size bytes, call Append() with the
// pointer returned from this function and the number of bytes written.
// Many Append() implementations will avoid copying bytes if this function
// returned an internal buffer.
//
// If the sink implementation allocates or reallocates an internal buffer,
// it should use the desired_size_hint if appropriate. If a caller cannot
// provide a reasonable guess at the desired capacity, it should set
// desired_size_hint = 0.
//
// If a non-scratch buffer is returned, the caller may only pass
// a prefix to it to Append(). That is, it is not correct to pass an
// interior pointer to Append().
//
// The default implementation always returns the scratch buffer.
virtual char* GetAppendBufferVariable(
size_t min_size, size_t desired_size_hint, char* scratch,
size_t scratch_size, size_t* allocated_size);
private:
// No copying
Sink(const Sink&);
void operator=(const Sink&);
};
// A Source is an interface that yields a sequence of bytes
class Source {
public:
Source() { }
virtual ~Source();
// Return the number of bytes left to read from the source
virtual size_t Available() const = 0;
// Peek at the next flat region of the source. Does not reposition
// the source. The returned region is empty iff Available()==0.
//
// Returns a pointer to the beginning of the region and store its
// length in *len.
//
// The returned region is valid until the next call to Skip() or
// until this object is destroyed, whichever occurs first.
//
// The returned region may be larger than Available() (for example
// if this ByteSource is a view on a substring of a larger source).
// The caller is responsible for ensuring that it only reads the
// Available() bytes.
virtual const char* Peek(size_t* len) = 0;
// Skip the next n bytes. Invalidates any buffer returned by
// a previous call to Peek().
// REQUIRES: Available() >= n
virtual void Skip(size_t n) = 0;
private:
// No copying
Source(const Source&);
void operator=(const Source&);
};
// A Source implementation that yields the contents of a flat array
class ByteArraySource : public Source {
public:
ByteArraySource(const char* p, size_t n) : ptr_(p), left_(n) { }
~ByteArraySource() override;
size_t Available() const override;
const char* Peek(size_t* len) override;
void Skip(size_t n) override;
private:
const char* ptr_;
size_t left_;
};
// A Sink implementation that writes to a flat array without any bound checks.
class UncheckedByteArraySink : public Sink {
public:
explicit UncheckedByteArraySink(char* dest) : dest_(dest) { }
~UncheckedByteArraySink() override;
void Append(const char* data, size_t n) override;
char* GetAppendBuffer(size_t len, char* scratch) override;
char* GetAppendBufferVariable(
size_t min_size, size_t desired_size_hint, char* scratch,
size_t scratch_size, size_t* allocated_size) override;
void AppendAndTakeOwnership(
char* bytes, size_t n, void (*deleter)(void*, const char*, size_t),
void *deleter_arg) override;
// Return the current output pointer so that a caller can see how
// many bytes were produced.
// Note: this is not a Sink method.
char* CurrentDestination() const { return dest_; }
private:
char* dest_;
};
} // namespace duckdb_snappy
#else // #if SNAPPY_NEW_VERSION
#include <stddef.h>
namespace duckdb_snappy {
// A Sink is an interface that consumes a sequence of bytes.
class Sink {
public:
Sink() { }
virtual ~Sink();
// Append "bytes[0,n-1]" to this.
virtual void Append(const char* bytes, size_t n) = 0;
// Returns a writable buffer of the specified length for appending.
// May return a pointer to the caller-owned scratch buffer which
// must have at least the indicated length. The returned buffer is
// only valid until the next operation on this Sink.
//
// After writing at most "length" bytes, call Append() with the
// pointer returned from this function and the number of bytes
// written. Many Append() implementations will avoid copying
// bytes if this function returned an internal buffer.
//
// If a non-scratch buffer is returned, the caller may only pass a
// prefix of it to Append(). That is, it is not correct to pass an
// interior pointer of the returned array to Append().
//
// The default implementation always returns the scratch buffer.
virtual char* GetAppendBuffer(size_t length, char* scratch);
// For higher performance, Sink implementations can provide custom
// AppendAndTakeOwnership() and GetAppendBufferVariable() methods.
// These methods can reduce the number of copies done during
// compression/decompression.
// Append "bytes[0,n-1] to the sink. Takes ownership of "bytes"
// and calls the deleter function as (*deleter)(deleter_arg, bytes, n)
// to free the buffer. deleter function must be non NULL.
//
// The default implementation just calls Append and frees "bytes".
// Other implementations may avoid a copy while appending the buffer.
virtual void AppendAndTakeOwnership(
char* bytes, size_t n, void (*deleter)(void*, const char*, size_t),
void *deleter_arg);
// Returns a writable buffer for appending and writes the buffer's capacity to
// *allocated_size. Guarantees *allocated_size >= min_size.
// May return a pointer to the caller-owned scratch buffer which must have
// scratch_size >= min_size.
//
// The returned buffer is only valid until the next operation
// on this ByteSink.
//
// After writing at most *allocated_size bytes, call Append() with the
// pointer returned from this function and the number of bytes written.
// Many Append() implementations will avoid copying bytes if this function
// returned an internal buffer.
//
// If the sink implementation allocates or reallocates an internal buffer,
// it should use the desired_size_hint if appropriate. If a caller cannot
// provide a reasonable guess at the desired capacity, it should set
// desired_size_hint = 0.
//
// If a non-scratch buffer is returned, the caller may only pass
// a prefix to it to Append(). That is, it is not correct to pass an
// interior pointer to Append().
//
// The default implementation always returns the scratch buffer.
virtual char* GetAppendBufferVariable(
size_t min_size, size_t desired_size_hint, char* scratch,
size_t scratch_size, size_t* allocated_size);
private:
// No copying
Sink(const Sink&);
void operator=(const Sink&);
};
// A Source is an interface that yields a sequence of bytes
class Source {
public:
Source() { }
virtual ~Source();
// Return the number of bytes left to read from the source
virtual size_t Available() const = 0;
// Peek at the next flat region of the source. Does not reposition
// the source. The returned region is empty iff Available()==0.
//
// Returns a pointer to the beginning of the region and store its
// length in *len.
//
// The returned region is valid until the next call to Skip() or
// until this object is destroyed, whichever occurs first.
//
// The returned region may be larger than Available() (for example
// if this ByteSource is a view on a substring of a larger source).
// The caller is responsible for ensuring that it only reads the
// Available() bytes.
virtual const char* Peek(size_t* len) = 0;
// Skip the next n bytes. Invalidates any buffer returned by
// a previous call to Peek().
// REQUIRES: Available() >= n
virtual void Skip(size_t n) = 0;
private:
// No copying
Source(const Source&);
void operator=(const Source&);
};
// A Source implementation that yields the contents of a flat array
class ByteArraySource : public Source {
public:
ByteArraySource(const char* p, size_t n) : ptr_(p), left_(n) { }
virtual ~ByteArraySource();
virtual size_t Available() const;
virtual const char* Peek(size_t* len);
virtual void Skip(size_t n);
private:
const char* ptr_;
size_t left_;
};
// A Sink implementation that writes to a flat array without any bound checks.
class UncheckedByteArraySink : public Sink {
public:
explicit UncheckedByteArraySink(char* dest) : dest_(dest) { }
virtual ~UncheckedByteArraySink();
virtual void Append(const char* data, size_t n);
virtual char* GetAppendBuffer(size_t len, char* scratch);
virtual char* GetAppendBufferVariable(
size_t min_size, size_t desired_size_hint, char* scratch,
size_t scratch_size, size_t* allocated_size);
virtual void AppendAndTakeOwnership(
char* bytes, size_t n, void (*deleter)(void*, const char*, size_t),
void *deleter_arg);
// Return the current output pointer so that a caller can see how
// many bytes were produced.
// Note: this is not a Sink method.
char* CurrentDestination() const { return dest_; }
private:
char* dest_;
};
} // namespace duckdb_snappy
#endif // #if SNAPPY_NEW_VERSION # else
#endif // THIRD_PARTY_SNAPPY_SNAPPY_SINKSOURCE_H_

42
external/duckdb/third_party/snappy/snappy-stubs-internal.cc vendored Normal file
View File

@@ -0,0 +1,42 @@
// Copyright 2011 Google Inc. All Rights Reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following disclaimer
// in the documentation and/or other materials provided with the
// distribution.
// * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#include <algorithm>
#include <string>
#include "snappy-stubs-internal.h"
namespace duckdb_snappy {
void Bignum::Append32(std::string* s, uint32_t value) {
char buf[Bignum::kMax32];
const char* p = Bignum::Encode32(buf, value);
s->append(buf, p - buf);
}
} // namespace duckdb_snappy

1026
external/duckdb/third_party/snappy/snappy-stubs-internal.h vendored Normal file
View File

File diff suppressed because it is too large Load Diff

109
external/duckdb/third_party/snappy/snappy-stubs-public.h vendored Executable file
View File

@@ -0,0 +1,109 @@
// Copyright 2011 Google Inc. All Rights Reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following disclaimer
// in the documentation and/or other materials provided with the
// distribution.
// * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// Various type stubs for the open-source version of Snappy.
//
// This file cannot include config.h, as it is included from snappy.h,
// which is a public header. Instead, snappy-stubs-public.h is generated by
// from snappy-stubs-public.h.in at configure time.
#ifndef THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_PUBLIC_H_
#define THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_PUBLIC_H_
#include "snappy_version.hpp"
#if SNAPPY_NEW_VERSION
#include <cstddef>
#ifndef _WIN32 // HAVE_SYS_UIO_H
#include <sys/uio.h>
#endif // HAVE_SYS_UIO_H
#define SNAPPY_MAJOR 1
#define SNAPPY_MINOR 2
#define SNAPPY_PATCHLEVEL 1
#define SNAPPY_VERSION \
((SNAPPY_MAJOR << 16) | (SNAPPY_MINOR << 8) | SNAPPY_PATCHLEVEL)
namespace duckdb_snappy {
#ifdef _WIN32 // !HAVE_SYS_UIO_H
// Windows does not have an iovec type, yet the concept is universally useful.
// It is simple to define it ourselves, so we put it inside our own namespace.
struct iovec {
void* iov_base;
size_t iov_len;
};
#endif // !HAVE_SYS_UIO_H
} // namespace duckdb_snappy
#else // #if SNAPPY_NEW_VERSION
#include <cstddef>
#include <cstdint>
#include <string>
#ifndef _WIN32 // HAVE_SYS_UIO_H
#include <sys/uio.h>
#endif // HAVE_SYS_UIO_H
#define SNAPPY_MAJOR 1
#define SNAPPY_MINOR 1
#define SNAPPY_PATCHLEVEL 7
#define SNAPPY_VERSION \
((SNAPPY_MAJOR << 16) | (SNAPPY_MINOR << 8) | SNAPPY_PATCHLEVEL)
namespace duckdb_snappy {
using int8 = std::int8_t;
using uint8 = std::uint8_t;
using int16 = std::int16_t;
using uint16 = std::uint16_t;
using int32 = std::int32_t;
using uint32 = std::uint32_t;
using int64 = std::int64_t;
using uint64 = std::uint64_t;
using string = std::string;
#ifdef _WIN32 // !HAVE_SYS_UIO_H
// Windows does not have an iovec type, yet the concept is universally useful.
// It is simple to define it ourselves, so we put it inside our own namespace.
struct iovec {
void* iov_base;
size_t iov_len;
};
#endif // !HAVE_SYS_UIO_H
} // namespace duckdb_snappy
#endif // #if SNAPPY_NEW_VERSION # else
#endif // THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_PUBLIC_H_

63
external/duckdb/third_party/snappy/snappy-stubs-public.h.in vendored Normal file
View File

@@ -0,0 +1,63 @@
// Copyright 2011 Google Inc. All Rights Reserved.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following disclaimer
// in the documentation and/or other materials provided with the
// distribution.
// * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// Various type stubs for the open-source version of Snappy.
//
// This file cannot include config.h, as it is included from snappy.h,
// which is a public header. Instead, snappy-stubs-public.h is generated by
// from snappy-stubs-public.h.in at configure time.
#ifndef THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_PUBLIC_H_
#define THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_PUBLIC_H_
#include <cstddef>
#if ${HAVE_SYS_UIO_H_01} // HAVE_SYS_UIO_H
#include <sys/uio.h>
#endif // HAVE_SYS_UIO_H
#define SNAPPY_MAJOR ${PROJECT_VERSION_MAJOR}
#define SNAPPY_MINOR ${PROJECT_VERSION_MINOR}
#define SNAPPY_PATCHLEVEL ${PROJECT_VERSION_PATCH}
#define SNAPPY_VERSION \
((SNAPPY_MAJOR << 16) | (SNAPPY_MINOR << 8) | SNAPPY_PATCHLEVEL)
namespace snappy {
#if !${HAVE_SYS_UIO_H_01} // !HAVE_SYS_UIO_H
// Windows does not have an iovec type, yet the concept is universally useful.
// It is simple to define it ourselves, so we put it inside our own namespace.
struct iovec {
void* iov_base;
size_t iov_len;
};
#endif // !HAVE_SYS_UIO_H
} // namespace snappy
#endif // THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_PUBLIC_H_

4288
external/duckdb/third_party/snappy/snappy.cc vendored Normal file
View File

File diff suppressed because it is too large Load Diff

412
external/duckdb/third_party/snappy/snappy.h vendored Executable file
View File

@@ -0,0 +1,412 @@
// Copyright 2005 and onwards Google Inc.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following disclaimer
// in the documentation and/or other materials provided with the
// distribution.
// * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// A light-weight compression algorithm. It is designed for speed of
// compression and decompression, rather than for the utmost in space
// savings.
//
// For getting better compression ratios when you are compressing data
// with long repeated sequences or compressing data that is similar to
// other data, while still compressing fast, you might look at first
// using BMDiff and then compressing the output of BMDiff with
// Snappy.
#ifndef THIRD_PARTY_SNAPPY_SNAPPY_H__
#define THIRD_PARTY_SNAPPY_SNAPPY_H__
#include "snappy_version.hpp"
#if SNAPPY_NEW_VERSION
#include <stddef.h>
#include <stdint.h>
#include <string>
#include "snappy-stubs-public.h"
namespace duckdb_snappy {
class Source;
class Sink;
struct CompressionOptions {
// Compression level.
// Level 1 is the fastest
// Level 2 is a little slower but provides better compression. Level 2 is
// **EXPERIMENTAL** for the time being. It might happen that we decide to
// fall back to level 1 in the future.
// Levels 3+ are currently not supported. We plan to support levels up to
// 9 in the future.
// If you played with other compression algorithms, level 1 is equivalent to
// fast mode (level 1) of LZ4, level 2 is equivalent to LZ4's level 2 mode
// and compresses somewhere around zstd:-3 and zstd:-2 but generally with
// faster decompression speeds than snappy:1 and zstd:-3.
int level = DefaultCompressionLevel();
constexpr CompressionOptions() = default;
constexpr CompressionOptions(int compression_level)
: level(compression_level) {}
static constexpr int MinCompressionLevel() { return 1; }
static constexpr int MaxCompressionLevel() { return 2; }
static constexpr int DefaultCompressionLevel() { return 1; }
};
// ------------------------------------------------------------------------
// Generic compression/decompression routines.
// ------------------------------------------------------------------------
// Compress the bytes read from "*reader" and append to "*writer". Return the
// number of bytes written.
// First version is to preserve ABI.
size_t Compress(Source* reader, Sink* writer);
size_t Compress(Source* reader, Sink* writer,
CompressionOptions options);
// Find the uncompressed length of the given stream, as given by the header.
// Note that the true length could deviate from this; the stream could e.g.
// be truncated.
//
// Also note that this leaves "*source" in a state that is unsuitable for
// further operations, such as RawUncompress(). You will need to rewind
// or recreate the source yourself before attempting any further calls.
bool GetUncompressedLength(Source* source, uint32_t* result);
// ------------------------------------------------------------------------
// Higher-level string based routines (should be sufficient for most users)
// ------------------------------------------------------------------------
// Sets "*compressed" to the compressed version of "input[0..input_length-1]".
// Original contents of *compressed are lost.
//
// REQUIRES: "input[]" is not an alias of "*compressed".
// First version is to preserve ABI.
size_t Compress(const char* input, size_t input_length,
std::string* compressed);
size_t Compress(const char* input, size_t input_length,
std::string* compressed, CompressionOptions options);
// Same as `Compress` above but taking an `iovec` array as input. Note that
// this function preprocesses the inputs to compute the sum of
// `iov[0..iov_cnt-1].iov_len` before reading. To avoid this, use
// `RawCompressFromIOVec` below.
// First version is to preserve ABI.
size_t CompressFromIOVec(const struct iovec* iov, size_t iov_cnt,
std::string* compressed);
size_t CompressFromIOVec(const struct iovec* iov, size_t iov_cnt,
std::string* compressed,
CompressionOptions options);
// Decompresses "compressed[0..compressed_length-1]" to "*uncompressed".
// Original contents of "*uncompressed" are lost.
//
// REQUIRES: "compressed[]" is not an alias of "*uncompressed".
//
// returns false if the message is corrupted and could not be decompressed
bool Uncompress(const char* compressed, size_t compressed_length,
std::string* uncompressed);
// Decompresses "compressed" to "*uncompressed".
//
// returns false if the message is corrupted and could not be decompressed
bool Uncompress(Source* compressed, Sink* uncompressed);
// This routine uncompresses as much of the "compressed" as possible
// into sink. It returns the number of valid bytes added to sink
// (extra invalid bytes may have been added due to errors; the caller
// should ignore those). The emitted data typically has length
// GetUncompressedLength(), but may be shorter if an error is
// encountered.
size_t UncompressAsMuchAsPossible(Source* compressed, Sink* uncompressed);
// ------------------------------------------------------------------------
// Lower-level character array based routines. May be useful for
// efficiency reasons in certain circumstances.
// ------------------------------------------------------------------------
// REQUIRES: "compressed" must point to an area of memory that is at
// least "MaxCompressedLength(input_length)" bytes in length.
//
// Takes the data stored in "input[0..input_length]" and stores
// it in the array pointed to by "compressed".
//
// "*compressed_length" is set to the length of the compressed output.
//
// Example:
// char* output = new char[snappy::MaxCompressedLength(input_length)];
// size_t output_length;
// RawCompress(input, input_length, output, &output_length);
// ... Process(output, output_length) ...
// delete [] output;
void RawCompress(const char* input, size_t input_length, char* compressed,
size_t* compressed_length);
void RawCompress(const char* input, size_t input_length, char* compressed,
size_t* compressed_length, CompressionOptions options);
// Same as `RawCompress` above but taking an `iovec` array as input. Note that
// `uncompressed_length` is the total number of bytes to be read from the
// elements of `iov` (_not_ the number of elements in `iov`).
void RawCompressFromIOVec(const struct iovec* iov, size_t uncompressed_length,
char* compressed, size_t* compressed_length);
void RawCompressFromIOVec(const struct iovec* iov, size_t uncompressed_length,
char* compressed, size_t* compressed_length,
CompressionOptions options);
// Given data in "compressed[0..compressed_length-1]" generated by
// calling the Snappy::Compress routine, this routine
// stores the uncompressed data to
// uncompressed[0..GetUncompressedLength(compressed)-1]
// returns false if the message is corrupted and could not be decrypted
bool RawUncompress(const char* compressed, size_t compressed_length,
char* uncompressed);
// Given data from the byte source 'compressed' generated by calling
// the Snappy::Compress routine, this routine stores the uncompressed
// data to
// uncompressed[0..GetUncompressedLength(compressed,compressed_length)-1]
// returns false if the message is corrupted and could not be decrypted
bool RawUncompress(Source* compressed, char* uncompressed);
// Given data in "compressed[0..compressed_length-1]" generated by
// calling the Snappy::Compress routine, this routine
// stores the uncompressed data to the iovec "iov". The number of physical
// buffers in "iov" is given by iov_cnt and their cumulative size
// must be at least GetUncompressedLength(compressed). The individual buffers
// in "iov" must not overlap with each other.
//
// returns false if the message is corrupted and could not be decrypted
bool RawUncompressToIOVec(const char* compressed, size_t compressed_length,
const struct iovec* iov, size_t iov_cnt);
// Given data from the byte source 'compressed' generated by calling
// the Snappy::Compress routine, this routine stores the uncompressed
// data to the iovec "iov". The number of physical
// buffers in "iov" is given by iov_cnt and their cumulative size
// must be at least GetUncompressedLength(compressed). The individual buffers
// in "iov" must not overlap with each other.
//
// returns false if the message is corrupted and could not be decrypted
bool RawUncompressToIOVec(Source* compressed, const struct iovec* iov,
size_t iov_cnt);
// Returns the maximal size of the compressed representation of
// input data that is "source_bytes" bytes in length;
size_t MaxCompressedLength(size_t source_bytes);
// REQUIRES: "compressed[]" was produced by RawCompress() or Compress()
// Returns true and stores the length of the uncompressed data in
// *result normally. Returns false on parsing error.
// This operation takes O(1) time.
bool GetUncompressedLength(const char* compressed, size_t compressed_length,
size_t* result);
// Returns true iff the contents of "compressed[]" can be uncompressed
// successfully. Does not return the uncompressed data. Takes
// time proportional to compressed_length, but is usually at least
// a factor of four faster than actual decompression.
bool IsValidCompressedBuffer(const char* compressed,
size_t compressed_length);
// Returns true iff the contents of "compressed" can be uncompressed
// successfully. Does not return the uncompressed data. Takes
// time proportional to *compressed length, but is usually at least
// a factor of four faster than actual decompression.
// On success, consumes all of *compressed. On failure, consumes an
// unspecified prefix of *compressed.
bool IsValidCompressed(Source* compressed);
// The size of a compression block. Note that many parts of the compression
// code assumes that kBlockSize <= 65536; in particular, the hash table
// can only store 16-bit offsets, and EmitCopy() also assumes the offset
// is 65535 bytes or less. Note also that if you change this, it will
// affect the framing format (see framing_format.txt).
//
// Note that there might be older data around that is compressed with larger
// block sizes, so the decompression code should not rely on the
// non-existence of long backreferences.
static constexpr int kBlockLog = 16;
static constexpr size_t kBlockSize = 1 << kBlockLog;
static constexpr int kMinHashTableBits = 8;
static constexpr size_t kMinHashTableSize = 1 << kMinHashTableBits;
static constexpr int kMaxHashTableBits = 15;
static constexpr size_t kMaxHashTableSize = 1 << kMaxHashTableBits;
} // end namespace duckdb_snappy
#else // #if SNAPPY_NEW_VERSION
#include <cstddef>
#include <string>
#include "snappy-stubs-public.h"
namespace duckdb_snappy {
class Source;
class Sink;
// ------------------------------------------------------------------------
// Generic compression/decompression routines.
// ------------------------------------------------------------------------
// Compress the bytes read from "*source" and append to "*sink". Return the
// number of bytes written.
size_t Compress(Source* source, Sink* sink);
// Find the uncompressed length of the given stream, as given by the header.
// Note that the true length could deviate from this; the stream could e.g.
// be truncated.
//
// Also note that this leaves "*source" in a state that is unsuitable for
// further operations, such as RawUncompress(). You will need to rewind
// or recreate the source yourself before attempting any further calls.
bool GetUncompressedLength(Source* source, uint32* result);
// ------------------------------------------------------------------------
// Higher-level string based routines (should be sufficient for most users)
// ------------------------------------------------------------------------
// Sets "*output" to the compressed version of "input[0,input_length-1]".
// Original contents of *output are lost.
//
// REQUIRES: "input[]" is not an alias of "*output".
size_t Compress(const char* input, size_t input_length, string* output);
// Decompresses "compressed[0,compressed_length-1]" to "*uncompressed".
// Original contents of "*uncompressed" are lost.
//
// REQUIRES: "compressed[]" is not an alias of "*uncompressed".
//
// returns false if the message is corrupted and could not be decompressed
bool Uncompress(const char* compressed, size_t compressed_length,
string* uncompressed);
// Decompresses "compressed" to "*uncompressed".
//
// returns false if the message is corrupted and could not be decompressed
bool Uncompress(Source* compressed, Sink* uncompressed);
// This routine uncompresses as much of the "compressed" as possible
// into sink. It returns the number of valid bytes added to sink
// (extra invalid bytes may have been added due to errors; the caller
// should ignore those). The emitted data typically has length
// GetUncompressedLength(), but may be shorter if an error is
// encountered.
size_t UncompressAsMuchAsPossible(Source* compressed, Sink* uncompressed);
// ------------------------------------------------------------------------
// Lower-level character array based routines. May be useful for
// efficiency reasons in certain circumstances.
// ------------------------------------------------------------------------
// REQUIRES: "compressed" must point to an area of memory that is at
// least "MaxCompressedLength(input_length)" bytes in length.
//
// Takes the data stored in "input[0..input_length]" and stores
// it in the array pointed to by "compressed".
//
// "*compressed_length" is set to the length of the compressed output.
//
// Example:
// char* output = new char[snappy::MaxCompressedLength(input_length)];
// size_t output_length;
// RawCompress(input, input_length, output, &output_length);
// ... Process(output, output_length) ...
// delete [] output;
void RawCompress(const char* input,
size_t input_length,
char* compressed,
size_t* compressed_length);
// Given data in "compressed[0..compressed_length-1]" generated by
// calling the Snappy::Compress routine, this routine
// stores the uncompressed data to
// uncompressed[0..GetUncompressedLength(compressed)-1]
// returns false if the message is corrupted and could not be decrypted
bool RawUncompress(const char* compressed, size_t compressed_length,
char* uncompressed);
// Given data from the byte source 'compressed' generated by calling
// the Snappy::Compress routine, this routine stores the uncompressed
// data to
// uncompressed[0..GetUncompressedLength(compressed,compressed_length)-1]
// returns false if the message is corrupted and could not be decrypted
bool RawUncompress(Source* compressed, char* uncompressed);
// Given data in "compressed[0..compressed_length-1]" generated by
// calling the Snappy::Compress routine, this routine
// stores the uncompressed data to the iovec "iov". The number of physical
// buffers in "iov" is given by iov_cnt and their cumulative size
// must be at least GetUncompressedLength(compressed). The individual buffers
// in "iov" must not overlap with each other.
//
// returns false if the message is corrupted and could not be decrypted
bool RawUncompressToIOVec(const char* compressed, size_t compressed_length,
const struct iovec* iov, size_t iov_cnt);
// Given data from the byte source 'compressed' generated by calling
// the Snappy::Compress routine, this routine stores the uncompressed
// data to the iovec "iov". The number of physical
// buffers in "iov" is given by iov_cnt and their cumulative size
// must be at least GetUncompressedLength(compressed). The individual buffers
// in "iov" must not overlap with each other.
//
// returns false if the message is corrupted and could not be decrypted
bool RawUncompressToIOVec(Source* compressed, const struct iovec* iov,
size_t iov_cnt);
// Returns the maximal size of the compressed representation of
// input data that is "source_bytes" bytes in length;
size_t MaxCompressedLength(size_t source_bytes);
// REQUIRES: "compressed[]" was produced by RawCompress() or Compress()
// Returns true and stores the length of the uncompressed data in
// *result normally. Returns false on parsing error.
// This operation takes O(1) time.
bool GetUncompressedLength(const char* compressed, size_t compressed_length,
size_t* result);
// Returns true iff the contents of "compressed[]" can be uncompressed
// successfully. Does not return the uncompressed data. Takes
// time proportional to compressed_length, but is usually at least
// a factor of four faster than actual decompression.
bool IsValidCompressedBuffer(const char* compressed,
size_t compressed_length);
// Returns true iff the contents of "compressed" can be uncompressed
// successfully. Does not return the uncompressed data. Takes
// time proportional to *compressed length, but is usually at least
// a factor of four faster than actual decompression.
// On success, consumes all of *compressed. On failure, consumes an
// unspecified prefix of *compressed.
bool IsValidCompressed(Source* compressed);
} // end namespace duckdb_snappy
#endif // #if SNAPPY_NEW_VERSION # else
#endif // THIRD_PARTY_SNAPPY_SNAPPY_H__

11
external/duckdb/third_party/snappy/snappy_version.hpp vendored Normal file
View File

@@ -0,0 +1,11 @@
#pragma once
//! The new version of snappy is much faster when compiled with clang, but slower when compiled with GCC
//! For DuckDB, we default to the old version if the compiler is not clang
#ifndef SNAPPY_NEW_VERSION
#ifdef __clang__
#define SNAPPY_NEW_VERSION true
#else
#define SNAPPY_NEW_VERSION false
#endif
#endif