diff options
author | dwight <dwight@10gen.com> | 2011-07-18 09:28:26 -0400 |
---|---|---|
committer | dwight <dwight@10gen.com> | 2011-08-02 02:38:00 -0400 |
commit | 57b8eaa186ed07844f5eb563850e7a12704e3007 (patch) | |
tree | 3021be849245fe91d644f1c971786e3c2d89c6d5 /third_party | |
parent | 63c87fe20c42a21eae42052531589f7547748e69 (diff) | |
download | mongo-57b8eaa186ed07844f5eb563850e7a12704e3007.tar.gz |
compression lib
Diffstat (limited to 'third_party')
-rwxr-xr-x | third_party/snappy/COPYING | 28 | ||||
-rwxr-xr-x | third_party/snappy/README | 135 | ||||
-rwxr-xr-x | third_party/snappy/config.h | 124 | ||||
-rwxr-xr-x | third_party/snappy/snappy-internal.h | 150 | ||||
-rwxr-xr-x | third_party/snappy/snappy-sinksource.cc | 72 | ||||
-rwxr-xr-x | third_party/snappy/snappy-sinksource.h | 136 | ||||
-rwxr-xr-x | third_party/snappy/snappy-stubs-internal.cc | 42 | ||||
-rwxr-xr-x | third_party/snappy/snappy-stubs-internal.h | 478 | ||||
-rwxr-xr-x | third_party/snappy/snappy-stubs-public.h | 85 | ||||
-rwxr-xr-x | third_party/snappy/snappy.cc | 1021 | ||||
-rwxr-xr-x | third_party/snappy/snappy.h | 155 |
11 files changed, 2426 insertions, 0 deletions
diff --git a/third_party/snappy/COPYING b/third_party/snappy/COPYING new file mode 100755 index 00000000000..8d6bd9fed4e --- /dev/null +++ b/third_party/snappy/COPYING @@ -0,0 +1,28 @@ +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. diff --git a/third_party/snappy/README b/third_party/snappy/README new file mode 100755 index 00000000000..df8f0e178e2 --- /dev/null +++ b/third_party/snappy/README @@ -0,0 +1,135 @@ +Snappy, a fast compressor/decompressor. + + +Introduction +============ + +Snappy is a compression/decompression library. It does not aim for maximum +compression, or compatibility with any other compression library; instead, +it aims for very high speeds and reasonable compression. For instance, +compared to the fastest mode of zlib, Snappy is an order of magnitude faster +for most inputs, but the resulting compressed files are anywhere from 20% to +100% bigger. (For more information, see "Performance", below.) + +Snappy has the following properties: + + * Fast: Compression speeds at 250 MB/sec and beyond, with no assembler code. + See "Performance" below. + * Stable: Over the last few years, Snappy has compressed and decompressed + petabytes of data in Google's production environment. The Snappy bitstream + format is stable and will not change between versions. + * Robust: The Snappy decompressor is designed not to crash in the face of + corrupted or malicious input. + * Free and open source software: Snappy is licensed under a BSD-type license. + For more information, see the included COPYING file. + +Snappy has previously been called "Zippy" in some Google presentations +and the like. + + +Performance +=========== + +Snappy is intended to be fast. On a single core of a Core i7 processor +in 64-bit mode, it compresses at about 250 MB/sec or more and decompresses at +about 500 MB/sec or more. (These numbers are for the slowest inputs in our +benchmark suite; others are much faster.) In our tests, Snappy usually +is faster than algorithms in the same class (e.g. LZO, LZF, FastLZ, QuickLZ, +etc.) while achieving comparable compression ratios. + +Typical compression ratios (based on the benchmark suite) are about 1.5-1.7x +for plain text, about 2-4x for HTML, and of course 1.0x for JPEGs, PNGs and +other already-compressed data. Similar numbers for zlib in its fastest mode +are 2.6-2.8x, 3-7x and 1.0x, respectively. More sophisticated algorithms are +capable of achieving yet higher compression rates, although usually at the +expense of speed. Of course, compression ratio will vary significantly with +the input. + +Although Snappy should be fairly portable, it is primarily optimized +for 64-bit x86-compatible processors, and may run slower in other environments. +In particular: + + - Snappy uses 64-bit operations in several places to process more data at + once than would otherwise be possible. + - Snappy assumes unaligned 32- and 64-bit loads and stores are cheap. + On some platforms, these must be emulated with single-byte loads + and stores, which is much slower. + - Snappy assumes little-endian throughout, and needs to byte-swap data in + several places if running on a big-endian platform. + +Experience has shown that even heavily tuned code can be improved. +Performance optimizations, whether for 64-bit x86 or other platforms, +are of course most welcome; see "Contact", below. + + +Usage +===== + +Note that Snappy, both the implementation and the main interface, +is written in C++. However, several third-party bindings to other languages +are available; see the Google Code page at http://code.google.com/p/snappy/ +for more information. Also, if you want to use Snappy from C code, you can +use the included C bindings in snappy-c.h. + +To use Snappy from your own C++ program, include the file "snappy.h" from +your calling file, and link against the compiled library. + +There are many ways to call Snappy, but the simplest possible is + + snappy::Compress(input, &output); + +and similarly + + snappy::Uncompress(input, &output); + +where "input" and "output" are both instances of std::string. + +There are other interfaces that are more flexible in various ways, including +support for custom (non-array) input sources. See the header file for more +information. + + +Tests and benchmarks +==================== + +When you compile Snappy, snappy_unittest is compiled in addition to the +library itself. You do not need it to use the compressor from your own library, +but it contains several useful components for Snappy development. + +First of all, it contains unit tests, verifying correctness on your machine in +various scenarios. If you want to change or optimize Snappy, please run the +tests to verify you have not broken anything. Note that if you have the +Google Test library installed, unit test behavior (especially failures) will be +significantly more user-friendly. You can find Google Test at + + http://code.google.com/p/googletest/ + +You probably also want the gflags library for handling of command-line flags; +you can find it at + + http://code.google.com/p/google-gflags/ + +In addition to the unit tests, snappy contains microbenchmarks used to +tune compression and decompression performance. These are automatically run +before the unit tests, but you can disable them using the flag +--run_microbenchmarks=false if you have gflags installed (otherwise you will +need to edit the source). + +Finally, snappy can benchmark Snappy against a few other compression libraries +(zlib, LZO, LZF, FastLZ and QuickLZ), if they were detected at configure time. +To benchmark using a given file, give the compression algorithm you want to test +Snappy against (e.g. --zlib) and then a list of one or more file names on the +command line. The testdata/ directory contains the files used by the +microbenchmark, which should provide a reasonably balanced starting point for +benchmarking. (Note that baddata[1-3].snappy are not intended as benchmarks; they +are used to verify correctness in the presence of corrupted data in the unit +test.) + + +Contact +======= + +Snappy is distributed through Google Code. For the latest version, a bug tracker, +and other information, see + + http://code.google.com/p/snappy/ diff --git a/third_party/snappy/config.h b/third_party/snappy/config.h new file mode 100755 index 00000000000..bfc3b30087f --- /dev/null +++ b/third_party/snappy/config.h @@ -0,0 +1,124 @@ +/* config.h.in. Generated from configure.ac by autoheader. */ + +/* Define if building universal (internal helper macro) */ +//#undef AC_APPLE_UNIVERSAL_BUILD + +#if defined(_WIN32) +// signed/unsigned mismatch +#pragma warning( disable : 4018 ) +#endif + +/* Define to 1 if the compiler supports __builtin_ctz and friends. */ +#if defined(__GNUC__) +#definfe HAVE_BUILTIN_CTZ 1 +#endif + +/* Define to 1 if the compiler supports __builtin_expect. */ +#if defined(__GNUC__) +#definfe HAVE_BUILTIN_EXPECT 1 +#endif + +/* Define to 1 if you have the <dlfcn.h> header file. */ +#if !defined(_WIN32) +#define HAVE_DLFCN_H 1 +#endif + +/* Use the gflags package for command-line parsing. */ +#undef HAVE_GFLAGS + +/* Defined when Google Test is available. */ +#undef HAVE_GTEST + +/* Define to 1 if you have the <inttypes.h> header file. */ +#define HAVE_INTTYPES_H 1 + +/* Define to 1 if you have the `fastlz' library (-lfastlz). */ +#undef HAVE_LIBFASTLZ + +/* Define to 1 if you have the `lzf' library (-llzf). */ +#undef HAVE_LIBLZF + +/* Define to 1 if you have the `lzo2' library (-llzo2). */ +#undef HAVE_LIBLZO2 + +/* Define to 1 if you have the `quicklz' library (-lquicklz). */ +#undef HAVE_LIBQUICKLZ + +/* Define to 1 if you have the `z' library (-lz). */ +#undef HAVE_LIBZ + +/* Define to 1 if you have the <memory.h> header file. */ +#define HAVE_MEMORY_H 1 + +/* Define to 1 if you have the <stddef.h> header file. */ +#define HAVE_STDDEF_H 1 + +/* Define to 1 if you have the <stdint.h> header file. */ +#define HAVE_STDINT_H 1 + +/* Define to 1 if you have the <stdlib.h> header file. */ +#define HAVE_STDLIB_H 1 + +/* Define to 1 if you have the <strings.h> header file. */ +#define HAVE_STRINGS_H 1 + +/* Define to 1 if you have the <string.h> header file. */ +#define HAVE_STRING_H 1 + +/* Define to 1 if you have the <sys/mman.h> header file. */ +#if !defined(_WIN32) +#define HAVE_SYS_MMAN_H 1 +#endif + +/* Define to 1 if you have the <sys/resource.h> header file. */ +#define HAVE_SYS_RESOURCE_H 1 + +/* Define to 1 if you have the <sys/stat.h> header file. */ +#define HAVE_SYS_STAT_H 1 + +/* Define to 1 if you have the <sys/types.h> header file. */ +#define HAVE_SYS_TYPES_H 1 + +/* Define to 1 if you have the <unistd.h> header file. */ +#define HAVE_UNISTD_H 1 + +/* Define to 1 if you have the <windows.h> header file. */ +#if defined(_WIN32) +#define HAVE_WINDOWS_H 1 +#endif + +/* Define to the sub-directory in which libtool stores uninstalled libraries. + */ +#define LT_OBJDIR "libs/" + +/* Name of package */ +#define PACKAGE "snappy" + +#define PACKAGE_BUGREPORT "" + +/* Define to the full name of this package. */ +#define PACKAGE_NAME "snappy" + +/* Define to the full name and version of this package. */ +#define PACKAGE_STRING "snappy 1.0.3" + +/* Define to the one symbol short name of this package. */ +#define PACKAGE_TARNAME "snappy" + +/* Define to the home page for this package. */ +#define PACKAGE_URL "" + +/* Define to the version of this package. */ +#define PACKAGE_VERSION "1.0.3" + +/* Define to 1 if you have the ANSI C header files. */ +#define STDC_HEADERS 1 + +/* Version number of package */ +#define VERSION "1.0.3" + +/* Define WORDS_BIGENDIAN to 1 if your processor stores words with the most + significant byte first (like Motorola and SPARC, unlike Intel). */ +#if defined(__BIG_ENDIAN__) +#define WORDS_BIGENDIAN 1 +#endif diff --git a/third_party/snappy/snappy-internal.h b/third_party/snappy/snappy-internal.h new file mode 100755 index 00000000000..a32eda59fb2 --- /dev/null +++ b/third_party/snappy/snappy-internal.h @@ -0,0 +1,150 @@ +// 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 UTIL_SNAPPY_SNAPPY_INTERNAL_H_ +#define UTIL_SNAPPY_SNAPPY_INTERNAL_H_ + +#include "snappy-stubs-internal.h" + +namespace snappy { +namespace internal { + +class WorkingMemory { + public: + WorkingMemory() : large_table_(NULL) { } + ~WorkingMemory() { delete[] large_table_; } + + // 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 input_size, int* table_size); + + private: + uint16 small_table_[1<<10]; // 2KB + uint16* large_table_; // Allocated only when needed + + DISALLOW_COPY_AND_ASSIGN(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); + +// Return the largest n such that +// +// s1[0,n-1] == s2[0,n-1] +// and n <= (s2_limit - s2). +// +// Does not read *s2_limit or beyond. +// Does not read *(s1 + (s2_limit - s2)) or beyond. +// Requires that s2_limit >= s2. +// +// Separate implementation for x86_64, for speed. Uses the fact that +// x86_64 is little endian. +#if defined(ARCH_K8) +static inline int FindMatchLength(const char* s1, + const char* s2, + const char* s2_limit) { + DCHECK_GE(s2_limit, s2); + int matched = 0; + + // 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 (PREDICT_TRUE(s2 <= s2_limit - 8)) { + if (PREDICT_FALSE(UNALIGNED_LOAD64(s2) == UNALIGNED_LOAD64(s1 + matched))) { + s2 += 8; + matched += 8; + } else { + // On current (mid-2008) Opteron models there is a 3% more + // efficient code sequence to find the first non-matching byte. + // However, what follows is ~10% better on Intel Core 2 and newer, + // and we expect AMD's bsf instruction to improve. + uint64 x = UNALIGNED_LOAD64(s2) ^ UNALIGNED_LOAD64(s1 + matched); + int matching_bits = Bits::FindLSBSetNonZero64(x); + matched += matching_bits >> 3; + return matched; + } + } + while (PREDICT_TRUE(s2 < s2_limit)) { + if (PREDICT_TRUE(s1[matched] == *s2)) { + ++s2; + ++matched; + } else { + return matched; + } + } + return matched; +} +#else +static inline int FindMatchLength(const char* s1, + const char* s2, + const char* s2_limit) { + // Implementation based on the x86-64 version, above. + DCHECK_GE(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 matched; +} +#endif + +} // end namespace internal +} // end namespace snappy + +#endif // UTIL_SNAPPY_SNAPPY_INTERNAL_H_ diff --git a/third_party/snappy/snappy-sinksource.cc b/third_party/snappy/snappy-sinksource.cc new file mode 100755 index 00000000000..1017895f962 --- /dev/null +++ b/third_party/snappy/snappy-sinksource.cc @@ -0,0 +1,72 @@ +// 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 <string.h> + +#include "snappy-sinksource.h" + +namespace snappy { + +Source::~Source() { } + +Sink::~Sink() { } + +char* Sink::GetAppendBuffer(size_t length, char* scratch) { + return scratch; +} + +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_; +} + + +} diff --git a/third_party/snappy/snappy-sinksource.h b/third_party/snappy/snappy-sinksource.h new file mode 100755 index 00000000000..430baeabb0e --- /dev/null +++ b/third_party/snappy/snappy-sinksource.h @@ -0,0 +1,136 @@ +// 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 UTIL_SNAPPY_SNAPPY_SINKSOURCE_H_ +#define UTIL_SNAPPY_SNAPPY_SINKSOURCE_H_ + +#include <stddef.h> + + +namespace 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); + + 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); + + // 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_; +}; + + +} + +#endif // UTIL_SNAPPY_SNAPPY_SINKSOURCE_H_ diff --git a/third_party/snappy/snappy-stubs-internal.cc b/third_party/snappy/snappy-stubs-internal.cc new file mode 100755 index 00000000000..6ed334371f1 --- /dev/null +++ b/third_party/snappy/snappy-stubs-internal.cc @@ -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 snappy { + +void Varint::Append32(string* s, uint32 value) { + char buf[Varint::kMax32]; + const char* p = Varint::Encode32(buf, value); + s->append(buf, p - buf); +} + +} // namespace snappy diff --git a/third_party/snappy/snappy-stubs-internal.h b/third_party/snappy/snappy-stubs-internal.h new file mode 100755 index 00000000000..355a06bc568 --- /dev/null +++ b/third_party/snappy/snappy-stubs-internal.h @@ -0,0 +1,478 @@ +// 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 stubs for the open-source version of Snappy. + +#ifndef UTIL_SNAPPY_OPENSOURCE_SNAPPY_STUBS_INTERNAL_H_ +#define UTIL_SNAPPY_OPENSOURCE_SNAPPY_STUBS_INTERNAL_H_ + +#ifdef HAVE_CONFIG_H +#include "config.h" +#endif + +#include <iostream> +#include <string> + +#include <assert.h> +#include <stdlib.h> +#include <string.h> + +#ifdef HAVE_SYS_MMAN +#include <sys/mman.h> +#endif + +#include "snappy-stubs-public.h" + +#if defined(__x86_64__) + +// Enable 64-bit optimized versions of some routines. +#define ARCH_K8 1 + +#endif + +// Needed by OS X, among others. +#ifndef MAP_ANONYMOUS +#define MAP_ANONYMOUS MAP_ANON +#endif + +// Pull in std::min, std::ostream, and the likes. This is safe because this +// header file is never used from any public header files. +using namespace std; + +// The size of an array, if known at compile-time. +// Will give unexpected results if used on a pointer. +// We undefine it first, since some compilers already have a definition. +#ifdef ARRAYSIZE +#undef ARRAYSIZE +#endif +#define ARRAYSIZE(a) (sizeof(a) / sizeof(*(a))) + +// Static prediction hints. +#ifdef HAVE_BUILTIN_EXPECT +#define PREDICT_FALSE(x) (__builtin_expect(x, 0)) +#define PREDICT_TRUE(x) (__builtin_expect(!!(x), 1)) +#else +#define PREDICT_FALSE(x) x +#define PREDICT_TRUE(x) x +#endif + +// This is only used for recomputing the tag byte table used during +// decompression; for simplicity we just remove it from the open-source +// version (anyone who wants to regenerate it can just do the call +// themselves within main()). +#define DEFINE_bool(flag_name, default_value, description) \ + bool FLAGS_ ## flag_name = default_value; +#define DECLARE_bool(flag_name) \ + extern bool FLAGS_ ## flag_name; +#define REGISTER_MODULE_INITIALIZER(name, code) + +namespace snappy { + +static const uint32 kuint32max = static_cast<uint32>(0xFFFFFFFF); +static const int64 kint64max = static_cast<int64>(0x7FFFFFFFFFFFFFFFLL); + +// Logging. + +#define LOG(level) LogMessage() +#define VLOG(level) true ? (void)0 : \ + snappy::LogMessageVoidify() & snappy::LogMessage() + +class LogMessage { + public: + LogMessage() { } + ~LogMessage() { + cerr << endl; + } + + LogMessage& operator<<(const std::string& msg) { + cerr << msg; + return *this; + } + LogMessage& operator<<(int x) { + cerr << x; + return *this; + } +}; + +// Asserts, both versions activated in debug mode only, +// and ones that are always active. + +#define CRASH_UNLESS(condition) \ + PREDICT_TRUE(condition) ? (void)0 : \ + snappy::LogMessageVoidify() & snappy::LogMessageCrash() + +class LogMessageCrash : public LogMessage { + public: + LogMessageCrash() { } + ~LogMessageCrash() { + cerr << endl; + abort(); + } +}; + +// This class is used to explicitly ignore values in the conditional +// logging macros. This avoids compiler warnings like "value computed +// is not used" and "statement has no effect". + +class LogMessageVoidify { + public: + LogMessageVoidify() { } + // This has to be an operator with a precedence lower than << but + // higher than ?: + void operator&(const LogMessage&) { } +}; + +#define CHECK(cond) CRASH_UNLESS(cond) +#define CHECK_LE(a, b) CRASH_UNLESS((a) <= (b)) +#define CHECK_GE(a, b) CRASH_UNLESS((a) >= (b)) +#define CHECK_EQ(a, b) CRASH_UNLESS((a) == (b)) +#define CHECK_NE(a, b) CRASH_UNLESS((a) != (b)) +#define CHECK_LT(a, b) CRASH_UNLESS((a) < (b)) +#define CHECK_GT(a, b) CRASH_UNLESS((a) > (b)) + +#ifdef NDEBUG + +#define DCHECK(cond) CRASH_UNLESS(true) +#define DCHECK_LE(a, b) CRASH_UNLESS(true) +#define DCHECK_GE(a, b) CRASH_UNLESS(true) +#define DCHECK_EQ(a, b) CRASH_UNLESS(true) +#define DCHECK_NE(a, b) CRASH_UNLESS(true) +#define DCHECK_LT(a, b) CRASH_UNLESS(true) +#define DCHECK_GT(a, b) CRASH_UNLESS(true) + +#else + +#define DCHECK(cond) CHECK(cond) +#define DCHECK_LE(a, b) CHECK_LE(a, b) +#define DCHECK_GE(a, b) CHECK_GE(a, b) +#define DCHECK_EQ(a, b) CHECK_EQ(a, b) +#define DCHECK_NE(a, b) CHECK_NE(a, b) +#define DCHECK_LT(a, b) CHECK_LT(a, b) +#define DCHECK_GT(a, b) CHECK_GT(a, b) + +#endif + +// Potentially unaligned loads and stores. + +#if 1 +//#if defined(__i386__) || defined(__x86_64__) || defined(__powerpc__) || defined(_WIN32) + +#define UNALIGNED_LOAD16(_p) (*reinterpret_cast<const uint16 *>(_p)) +#define UNALIGNED_LOAD32(_p) (*reinterpret_cast<const uint32 *>(_p)) +#define UNALIGNED_LOAD64(_p) (*reinterpret_cast<const uint64 *>(_p)) + +#define UNALIGNED_STORE16(_p, _val) (*reinterpret_cast<uint16 *>(_p) = (_val)) +#define UNALIGNED_STORE32(_p, _val) (*reinterpret_cast<uint32 *>(_p) = (_val)) +#define UNALIGNED_STORE64(_p, _val) (*reinterpret_cast<uint64 *>(_p) = (_val)) + +#else + +// These functions are provided for architectures that don't support +// unaligned loads and stores. + +inline uint16 UNALIGNED_LOAD16(const void *p) { + uint16 t; + memcpy(&t, p, sizeof t); + return t; +} + +inline uint32 UNALIGNED_LOAD32(const void *p) { + uint32 t; + memcpy(&t, p, sizeof t); + return t; +} + +inline uint64 UNALIGNED_LOAD64(const void *p) { + uint64 t; + memcpy(&t, p, sizeof t); + return t; +} + +inline void UNALIGNED_STORE16(void *p, uint16 v) { + memcpy(p, &v, sizeof v); +} + +inline void UNALIGNED_STORE32(void *p, uint32 v) { + memcpy(p, &v, sizeof v); +} + +inline void UNALIGNED_STORE64(void *p, uint64 v) { + memcpy(p, &v, sizeof v); +} + +#endif + +// The following guarantees declaration of the byte swap functions. +#ifdef WORDS_BIGENDIAN + +#ifdef _MSC_VER +#include <stdlib.h> +#define bswap_16(x) _byteswap_ushort(x) +#define bswap_32(x) _byteswap_ulong(x) +#define bswap_64(x) _byteswap_uint64(x) + +#elif defined(__APPLE__) +// Mac OS X / Darwin features +#include <libkern/OSByteOrder.h> +#define bswap_16(x) OSSwapInt16(x) +#define bswap_32(x) OSSwapInt32(x) +#define bswap_64(x) OSSwapInt64(x) + +#else +#include <byteswap.h> +#endif + +#endif // WORDS_BIGENDIAN + +// Convert to little-endian storage, opposite of network format. +// Convert x from host to little endian: x = LittleEndian.FromHost(x); +// convert x from little endian to host: x = LittleEndian.ToHost(x); +// +// Store values into unaligned memory converting to little endian order: +// LittleEndian.Store16(p, x); +// +// Load unaligned values stored in little endian converting to host order: +// x = LittleEndian.Load16(p); +class LittleEndian { + public: + // Conversion functions. +#ifdef WORDS_BIGENDIAN + + static uint16 FromHost16(uint16 x) { return bswap_16(x); } + static uint16 ToHost16(uint16 x) { return bswap_16(x); } + + static uint32 FromHost32(uint32 x) { return bswap_32(x); } + static uint32 ToHost32(uint32 x) { return bswap_32(x); } + + static bool IsLittleEndian() { return false; } + +#else // !defined(WORDS_BIGENDIAN) + + static uint16 FromHost16(uint16 x) { return x; } + static uint16 ToHost16(uint16 x) { return x; } + + static uint32 FromHost32(uint32 x) { return x; } + static uint32 ToHost32(uint32 x) { return x; } + + static bool IsLittleEndian() { return true; } + +#endif // !defined(WORDS_BIGENDIAN) + + // Functions to do unaligned loads and stores in little-endian order. + static uint16 Load16(const void *p) { + return ToHost16(UNALIGNED_LOAD16(p)); + } + + static void Store16(void *p, uint16 v) { + UNALIGNED_STORE16(p, FromHost16(v)); + } + + static uint32 Load32(const void *p) { + return ToHost32(UNALIGNED_LOAD32(p)); + } + + static void Store32(void *p, uint32 v) { + UNALIGNED_STORE32(p, FromHost32(v)); + } +}; + +// Some bit-manipulation functions. +class Bits { + public: + // Return floor(log2(n)) for positive integer n. Returns -1 iff n == 0. + static int Log2Floor(uint32 n); + + // Return the first set least / most significant bit, 0-indexed. Returns an + // undefined value if n == 0. FindLSBSetNonZero() is similar to ffs() except + // that it's 0-indexed. + static int FindLSBSetNonZero(uint32 n); + static int FindLSBSetNonZero64(uint64 n); + + private: + DISALLOW_COPY_AND_ASSIGN(Bits); +}; + +#ifdef HAVE_BUILTIN_CTZ + +inline int Bits::Log2Floor(uint32 n) { + return n == 0 ? -1 : 31 ^ __builtin_clz(n); +} + +inline int Bits::FindLSBSetNonZero(uint32 n) { + return __builtin_ctz(n); +} + +inline int Bits::FindLSBSetNonZero64(uint64 n) { + return __builtin_ctzll(n); +} + +#else // Portable versions. + +inline int Bits::Log2Floor(uint32 n) { + if (n == 0) + return -1; + int log = 0; + uint32 value = n; + for (int i = 4; i >= 0; --i) { + int shift = (1 << i); + uint32 x = value >> shift; + if (x != 0) { + value = x; + log += shift; + } + } + assert(value == 1); + return log; +} + +inline int Bits::FindLSBSetNonZero(uint32 n) { + int rc = 31; + for (int i = 4, shift = 1 << 4; i >= 0; --i) { + const uint32 x = n << shift; + if (x != 0) { + n = x; + rc -= shift; + } + shift >>= 1; + } + return rc; +} + +// FindLSBSetNonZero64() is defined in terms of FindLSBSetNonZero(). +inline int Bits::FindLSBSetNonZero64(uint64 n) { + const uint32 bottombits = static_cast<uint32>(n); + if (bottombits == 0) { + // Bottom bits are zero, so scan in top bits + return 32 + FindLSBSetNonZero(static_cast<uint32>(n >> 32)); + } else { + return FindLSBSetNonZero(bottombits); + } +} + +#endif // End portable versions. + +// Variable-length integer encoding. +class Varint { + public: + // Maximum lengths of varint encoding of uint32. + static const int kMax32 = 5; + + // Attempts to parse a varint32 from a prefix of the bytes in [ptr,limit-1]. + // Never reads a character at or beyond limit. If a valid/terminated varint32 + // was found in the range, stores it in *OUTPUT and returns a pointer just + // past the last byte of the varint32. Else returns NULL. On success, + // "result <= limit". + static const char* Parse32WithLimit(const char* ptr, const char* limit, + uint32* OUTPUT); + + // REQUIRES "ptr" points to a buffer of length sufficient to hold "v". + // EFFECTS Encodes "v" into "ptr" and returns a pointer to the + // byte just past the last encoded byte. + static char* Encode32(char* ptr, uint32 v); + + // EFFECTS Appends the varint representation of "value" to "*s". + static void Append32(string* s, uint32 value); +}; + +inline const char* Varint::Parse32WithLimit(const char* p, + const char* l, + uint32* OUTPUT) { + const unsigned char* ptr = reinterpret_cast<const unsigned char*>(p); + const unsigned char* limit = reinterpret_cast<const unsigned char*>(l); + uint32 b, result; + if (ptr >= limit) return NULL; + b = *(ptr++); result = b & 127; if (b < 128) goto done; + if (ptr >= limit) return NULL; + b = *(ptr++); result |= (b & 127) << 7; if (b < 128) goto done; + if (ptr >= limit) return NULL; + b = *(ptr++); result |= (b & 127) << 14; if (b < 128) goto done; + if (ptr >= limit) return NULL; + b = *(ptr++); result |= (b & 127) << 21; if (b < 128) goto done; + if (ptr >= limit) return NULL; + b = *(ptr++); result |= (b & 127) << 28; if (b < 16) goto done; + return NULL; // Value is too long to be a varint32 + done: + *OUTPUT = result; + return reinterpret_cast<const char*>(ptr); +} + +inline char* Varint::Encode32(char* sptr, uint32 v) { + // Operate on characters as unsigneds + unsigned char* ptr = reinterpret_cast<unsigned char*>(sptr); + static const int B = 128; + if (v < (1<<7)) { + *(ptr++) = v; + } else if (v < (1<<14)) { + *(ptr++) = v | B; + *(ptr++) = v>>7; + } else if (v < (1<<21)) { + *(ptr++) = v | B; + *(ptr++) = (v>>7) | B; + *(ptr++) = v>>14; + } else if (v < (1<<28)) { + *(ptr++) = v | B; + *(ptr++) = (v>>7) | B; + *(ptr++) = (v>>14) | B; + *(ptr++) = v>>21; + } else { + *(ptr++) = v | B; + *(ptr++) = (v>>7) | B; + *(ptr++) = (v>>14) | B; + *(ptr++) = (v>>21) | B; + *(ptr++) = v>>28; + } + return reinterpret_cast<char*>(ptr); +} + +// If you know the internal layout of the std::string in use, you can +// replace this function with one that resizes the string without +// filling the new space with zeros (if applicable) -- +// it will be non-portable but faster. +inline void STLStringResizeUninitialized(string* s, size_t new_size) { + s->resize(new_size); +} + +// Return a mutable char* pointing to a string's internal buffer, +// which may not be null-terminated. Writing through this pointer will +// modify the string. +// +// string_as_array(&str)[i] is valid for 0 <= i < str.size() until the +// next call to a string method that invalidates iterators. +// +// As of 2006-04, there is no standard-blessed way of getting a +// mutable reference to a string's internal buffer. However, issue 530 +// (http://www.open-std.org/JTC1/SC22/WG21/docs/lwg-defects.html#530) +// proposes this as the method. It will officially be part of the standard +// for C++0x. This should already work on all current implementations. +inline char* string_as_array(string* str) { + return str->empty() ? NULL : &*str->begin(); +} + +} // namespace snappy + +#endif // UTIL_SNAPPY_OPENSOURCE_SNAPPY_STUBS_INTERNAL_H_ diff --git a/third_party/snappy/snappy-stubs-public.h b/third_party/snappy/snappy-stubs-public.h new file mode 100755 index 00000000000..b089baca9b5 --- /dev/null +++ b/third_party/snappy/snappy-stubs-public.h @@ -0,0 +1,85 @@ +// Copyright 2011 Google Inc. All Rights Reserved. +// Author: sesse@google.com (Steinar H. Gunderson) +// +// 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 UTIL_SNAPPY_OPENSOURCE_SNAPPY_STUBS_PUBLIC_H_ +#define UTIL_SNAPPY_OPENSOURCE_SNAPPY_STUBS_PUBLIC_H_ + +#if 1 +#include <stdint.h> +#endif + +#if 1 +#include <stddef.h> +#endif + +#define SNAPPY_MAJOR 1 +#define SNAPPY_MINOR 0 +#define SNAPPY_PATCHLEVEL 3 +#define SNAPPY_VERSION \ + ((SNAPPY_MAJOR << 16) | (SNAPPY_MINOR << 8) | SNAPPY_PATCHLEVEL) + +#include <string> + +namespace snappy { + +#if 1 +typedef int8_t int8; +typedef uint8_t uint8; +typedef int16_t int16; +typedef uint16_t uint16; +typedef int32_t int32; +typedef uint32_t uint32; +typedef int64_t int64; +typedef uint64_t uint64; +#else +typedef signed char int8; +typedef unsigned char uint8; +typedef short int16; +typedef unsigned short uint16; +typedef int int32; +typedef unsigned int uint32; +typedef long long int64; +typedef unsigned long long uint64; +#endif + +typedef std::string string; + +#define DISALLOW_COPY_AND_ASSIGN(TypeName) \ + TypeName(const TypeName&); \ + void operator=(const TypeName&) + +} // namespace snappy + +#endif // UTIL_SNAPPY_OPENSOURCE_SNAPPY_STUBS_PUBLIC_H_ diff --git a/third_party/snappy/snappy.cc b/third_party/snappy/snappy.cc new file mode 100755 index 00000000000..a591aba5945 --- /dev/null +++ b/third_party/snappy/snappy.cc @@ -0,0 +1,1021 @@ +// Copyright 2005 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.h" +#include "snappy-internal.h" +#include "snappy-sinksource.h" + +#include <stdio.h> + +#include <algorithm> +#include <string> +#include <vector> + + +namespace snappy { + +// Any hash function will produce a valid compressed bitstream, but a good +// hash function reduces the number of collisions and thus yields better +// compression for compressible input, and more speed for incompressible +// input. Of course, it doesn't hurt if the hash function is reasonably fast +// either, as it gets called a lot. +static inline uint32 HashBytes(uint32 bytes, int shift) { + uint32 kMul = 0x1e35a7bd; + return (bytes * kMul) >> shift; +} +static inline uint32 Hash(const char* p, int shift) { + return HashBytes(UNALIGNED_LOAD32(p), shift); +} + +size_t MaxCompressedLength(size_t source_len) { + // Compressed data can be defined as: + // compressed := item* literal* + // item := literal* copy + // + // The trailing literal sequence has a space blowup of at most 62/60 + // since a literal of length 60 needs one tag byte + one extra byte + // for length information. + // + // Item blowup is trickier to measure. Suppose the "copy" op copies + // 4 bytes of data. Because of a special check in the encoding code, + // we produce a 4-byte copy only if the offset is < 65536. Therefore + // the copy op takes 3 bytes to encode, and this type of item leads + // to at most the 62/60 blowup for representing literals. + // + // Suppose the "copy" op copies 5 bytes of data. If the offset is big + // enough, it will take 5 bytes to encode the copy op. Therefore the + // worst case here is a one-byte literal followed by a five-byte copy. + // I.e., 6 bytes of input turn into 7 bytes of "compressed" data. + // + // This last factor dominates the blowup, so the final estimate is: + return 32 + source_len + source_len/6; +} + +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 +}; + +// Copy "len" bytes from "src" to "op", one byte at a time. Used for +// handling COPY operations where the input and output regions may +// overlap. For example, suppose: +// src == "ab" +// op == src + 2 +// len == 20 +// After IncrementalCopy(src, op, len), the result will have +// eleven copies of "ab" +// ababababababababababab +// Note that this does not match the semantics of either memcpy() +// or memmove(). +static inline void IncrementalCopy(const char* src, char* op, int len) { + DCHECK_GT(len, 0); + do { + *op++ = *src++; + } while (--len > 0); +} + +// Equivalent to IncrementalCopy except that it can write up to ten extra +// bytes after the end of the copy, and that it is faster. +// +// The main part of this loop is a simple copy of eight bytes at a time until +// we've copied (at least) the requested amount of bytes. However, if op and +// src are less than eight bytes apart (indicating a repeating pattern of +// length < 8), we first need to expand the pattern in order to get the correct +// results. For instance, if the buffer looks like this, with the eight-byte +// <src> and <op> patterns marked as intervals: +// +// abxxxxxxxxxxxx +// [------] src +// [------] op +// +// a single eight-byte copy from <src> to <op> will repeat the pattern once, +// after which we can move <op> two bytes without moving <src>: +// +// ababxxxxxxxxxx +// [------] src +// [------] op +// +// and repeat the exercise until the two no longer overlap. +// +// This allows us to do very well in the special case of one single byte +// repeated many times, without taking a big hit for more general cases. +// +// The worst case of extra writing past the end of the match occurs when +// op - src == 1 and len == 1; the last copy will read from byte positions +// [0..7] and write to [4..11], whereas it was only supposed to write to +// position 1. Thus, ten excess bytes. + +namespace { + +const int kMaxIncrementCopyOverflow = 10; + +} // namespace + +static inline void IncrementalCopyFastPath(const char* src, char* op, int len) { + while (op - src < 8) { + UNALIGNED_STORE64(op, UNALIGNED_LOAD64(src)); + len -= op - src; + op += op - src; + } + while (len > 0) { + UNALIGNED_STORE64(op, UNALIGNED_LOAD64(src)); + src += 8; + op += 8; + len -= 8; + } +} + +static inline char* EmitLiteral(char* op, + const char* literal, + int len, + bool allow_fast_path) { + int n = len - 1; // Zero-length literals are disallowed + if (n < 60) { + // Fits in tag byte + *op++ = LITERAL | (n << 2); + + // The vast majority of copies are below 16 bytes, for which a + // call to memcpy is overkill. This fast path can sometimes + // copy up to 15 bytes too much, but that is okay in the + // main loop, since we have a bit to go on for both sides: + // + // - The input will always have kInputMarginBytes = 15 extra + // available bytes, as long as we're in the main loop, and + // if not, allow_fast_path = false. + // - The output will always have 32 spare bytes (see + // MaxCompressedLength). + if (allow_fast_path && len <= 16) { + UNALIGNED_STORE64(op, UNALIGNED_LOAD64(literal)); + UNALIGNED_STORE64(op + 8, UNALIGNED_LOAD64(literal + 8)); + return op + len; + } + } else { + // Encode in upcoming bytes + char* base = op; + int count = 0; + op++; + while (n > 0) { + *op++ = n & 0xff; + n >>= 8; + count++; + } + assert(count >= 1); + assert(count <= 4); + *base = LITERAL | ((59+count) << 2); + } + memcpy(op, literal, len); + return op + len; +} + +static inline char* EmitCopyLessThan64(char* op, int offset, int len) { + DCHECK_LE(len, 64); + DCHECK_GE(len, 4); + DCHECK_LT(offset, 65536); + + if ((len < 12) && (offset < 2048)) { + int len_minus_4 = len - 4; + assert(len_minus_4 < 8); // Must fit in 3 bits + *op++ = COPY_1_BYTE_OFFSET | ((len_minus_4) << 2) | ((offset >> 8) << 5); + *op++ = offset & 0xff; + } else { + *op++ = COPY_2_BYTE_OFFSET | ((len-1) << 2); + LittleEndian::Store16(op, offset); + op += 2; + } + return op; +} + +static inline char* EmitCopy(char* op, int offset, int len) { + // Emit 64 byte copies but make sure to keep at least four bytes reserved + while (len >= 68) { + op = EmitCopyLessThan64(op, offset, 64); + len -= 64; + } + + // Emit an extra 60 byte copy if have too much data to fit in one copy + if (len > 64) { + op = EmitCopyLessThan64(op, offset, 60); + len -= 60; + } + + // Emit remainder + op = EmitCopyLessThan64(op, offset, len); + return op; +} + + +bool GetUncompressedLength(const char* start, size_t n, size_t* result) { + uint32 v = 0; + const char* limit = start + n; + if (Varint::Parse32WithLimit(start, limit, &v) != NULL) { + *result = v; + return true; + } else { + return false; + } +} + +namespace internal { +uint16* WorkingMemory::GetHashTable(size_t input_size, int* table_size) { + // Use smaller hash table when input.size() is smaller, since we + // fill the table, incurring O(hash table size) overhead for + // compression, and if the input is short, we won't need that + // many hash table entries anyway. + assert(kMaxHashTableSize >= 256); + int htsize = 256; + while (htsize < kMaxHashTableSize && htsize < input_size) { + htsize <<= 1; + } + CHECK_EQ(0, htsize & (htsize - 1)) << ": must be power of two"; + CHECK_LE(htsize, kMaxHashTableSize) << ": hash table too large"; + + uint16* table; + if (htsize <= ARRAYSIZE(small_table_)) { + table = small_table_; + } else { + if (large_table_ == NULL) { + large_table_ = new uint16[kMaxHashTableSize]; + } + table = large_table_; + } + + *table_size = htsize; + memset(table, 0, htsize * sizeof(*table)); + return table; +} +} // end namespace internal + +// For 0 <= offset <= 4, GetUint32AtOffset(UNALIGNED_LOAD64(p), offset) will +// equal UNALIGNED_LOAD32(p + offset). Motivation: On x86-64 hardware we have +// empirically found that overlapping loads such as +// UNALIGNED_LOAD32(p) ... UNALIGNED_LOAD32(p+1) ... UNALIGNED_LOAD32(p+2) +// are slower than UNALIGNED_LOAD64(p) followed by shifts and casts to uint32. +static inline uint32 GetUint32AtOffset(uint64 v, int offset) { + DCHECK(0 <= offset && offset <= 4) << offset; + return v >> (LittleEndian::IsLittleEndian() ? 8 * offset : 32 - 8 * offset); +} + +// Flat array compression that does not emit the "uncompressed length" +// prefix. Compresses "input" string to the "*op" buffer. +// +// REQUIRES: "input" is at most "kBlockSize" bytes long. +// REQUIRES: "op" points to an array of memory that is at least +// "MaxCompressedLength(input.size())" 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". +namespace internal { +char* CompressFragment(const char* const input, + const size_t input_size, + char* op, + uint16* table, + const int table_size) { + // "ip" is the input pointer, and "op" is the output pointer. + const char* ip = input; + CHECK_LE(input_size, kBlockSize); + CHECK_EQ(table_size & (table_size - 1), 0) << ": table must be power of two"; + const int shift = 32 - Bits::Log2Floor(table_size); + DCHECK_EQ(kuint32max >> shift, table_size - 1); + const char* ip_end = input + input_size; + const char* base_ip = ip; + // Bytes in [next_emit, ip) will be emitted as literal bytes. Or + // [next_emit, ip_end) after the main loop. + const char* next_emit = ip; + + const int kInputMarginBytes = 15; + if (PREDICT_TRUE(input_size >= kInputMarginBytes)) { + const char* ip_limit = input + input_size - kInputMarginBytes; + + for (uint32 next_hash = Hash(++ip, shift); ; ) { + DCHECK_LT(next_emit, ip); + // The body of this loop calls EmitLiteral once and then EmitCopy one or + // more times. (The exception is that when we're close to exhausting + // the input we goto emit_remainder.) + // + // In the first iteration of this loop we're just starting, so + // there's nothing to copy, so calling EmitLiteral once is + // necessary. And we only start a new iteration when the + // current iteration has determined that a call to EmitLiteral will + // precede the next call to EmitCopy (if any). + // + // Step 1: Scan forward in the input looking for a 4-byte-long match. + // If we get close to exhausting the input then goto emit_remainder. + // + // Heuristic match skipping: If 32 bytes are scanned with no matches + // found, start looking only at every other byte. If 32 more bytes are + // scanned, look at every third byte, etc.. When a match is found, + // immediately go back to looking at every byte. This is a small loss + // (~5% performance, ~0.1% density) for compressible data due to more + // bookkeeping, but for non-compressible data (such as JPEG) it's a huge + // win since the compressor quickly "realizes" the data is incompressible + // and doesn't bother looking for matches everywhere. + // + // The "skip" variable keeps track of how many bytes there are since the + // last match; dividing it by 32 (ie. right-shifting by five) gives the + // number of bytes to move ahead for each iteration. + uint32 skip = 32; + + const char* next_ip = ip; + const char* candidate; + do { + ip = next_ip; + uint32 hash = next_hash; + DCHECK_EQ(hash, Hash(ip, shift)); + uint32 bytes_between_hash_lookups = skip++ >> 5; + next_ip = ip + bytes_between_hash_lookups; + if (PREDICT_FALSE(next_ip > ip_limit)) { + goto emit_remainder; + } + next_hash = Hash(next_ip, shift); + candidate = base_ip + table[hash]; + DCHECK_GE(candidate, base_ip); + DCHECK_LT(candidate, ip); + + table[hash] = ip - base_ip; + } while (PREDICT_TRUE(UNALIGNED_LOAD32(ip) != + UNALIGNED_LOAD32(candidate))); + + // Step 2: A 4-byte match has been found. We'll later see if more + // than 4 bytes match. But, prior to the match, input + // bytes [next_emit, ip) are unmatched. Emit them as "literal bytes." + DCHECK_LE(next_emit + 16, ip_end); + op = EmitLiteral(op, next_emit, ip - next_emit, true); + + // Step 3: Call EmitCopy, and then see if another EmitCopy could + // be our next move. Repeat until we find no match for the + // input immediately after what was consumed by the last EmitCopy call. + // + // If we exit this loop normally then we need to call EmitLiteral next, + // though we don't yet know how big the literal will be. We handle that + // by proceeding to the next iteration of the main loop. We also can exit + // this loop via goto if we get close to exhausting the input. + uint64 input_bytes = 0; + uint32 candidate_bytes = 0; + + do { + // We have a 4-byte match at ip, and no need to emit any + // "literal bytes" prior to ip. + const char* base = ip; + int matched = 4 + FindMatchLength(candidate + 4, ip + 4, ip_end); + ip += matched; + int offset = base - candidate; + DCHECK_EQ(0, memcmp(base, candidate, matched)); + op = EmitCopy(op, offset, matched); + // We could immediately start working at ip now, but to improve + // compression we first update table[Hash(ip - 1, ...)]. + const char* insert_tail = ip - 1; + next_emit = ip; + if (PREDICT_FALSE(ip >= ip_limit)) { + goto emit_remainder; + } + input_bytes = UNALIGNED_LOAD64(insert_tail); + uint32 prev_hash = HashBytes(GetUint32AtOffset(input_bytes, 0), shift); + table[prev_hash] = ip - base_ip - 1; + uint32 cur_hash = HashBytes(GetUint32AtOffset(input_bytes, 1), shift); + candidate = base_ip + table[cur_hash]; + candidate_bytes = UNALIGNED_LOAD32(candidate); + table[cur_hash] = ip - base_ip; + } while (GetUint32AtOffset(input_bytes, 1) == candidate_bytes); + + next_hash = HashBytes(GetUint32AtOffset(input_bytes, 2), shift); + ++ip; + } + } + + emit_remainder: + // Emit the remaining bytes as a literal + if (next_emit < ip_end) { + op = EmitLiteral(op, next_emit, ip_end - next_emit, false); + } + + return op; +} +} // end namespace internal + +// Signature of output types needed by decompression code. +// The decompression code is templatized on a type that obeys this +// signature so that we do not pay virtual function call overhead in +// the middle of a tight decompression loop. +// +// class DecompressionWriter { +// public: +// // Called before decompression +// void SetExpectedLength(size_t length); +// +// // Called after decompression +// bool CheckLength() const; +// +// // Called repeatedly during decompression +// bool Append(const char* ip, uint32 length, bool allow_fast_path); +// bool AppendFromSelf(uint32 offset, uint32 length); +// }; +// +// "allow_fast_path" is a parameter that says if there is at least 16 +// readable bytes in "ip". It is currently only used by SnappyArrayWriter. + +// ----------------------------------------------------------------------- +// Lookup table for decompression code. Generated by ComputeTable() below. +// ----------------------------------------------------------------------- + +// Mapping from i in range [0,4] to a mask to extract the bottom 8*i bits +static const uint32 wordmask[] = { + 0u, 0xffu, 0xffffu, 0xffffffu, 0xffffffffu +}; + +// 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 +}; + +// In debug mode, allow optional computation of the table at startup. +// Also, check that the decompression table is correct. +#ifndef NDEBUG +DEFINE_bool(snappy_dump_decompression_table, false, + "If true, we print the decompression table at startup."); + +static uint16 MakeEntry(unsigned int extra, + unsigned int len, + unsigned int copy_offset) { + // Check that all of the fields fit within the allocated space + DCHECK_EQ(extra, extra & 0x7); // At most 3 bits + DCHECK_EQ(copy_offset, copy_offset & 0x7); // At most 3 bits + DCHECK_EQ(len, len & 0x7f); // At most 7 bits + return len | (copy_offset << 8) | (extra << 11); +} + +static void ComputeTable() { + uint16 dst[256]; + + // Place invalid entries in all places to detect missing initialization + int assigned = 0; + for (int i = 0; i < 256; i++) { + dst[i] = 0xffff; + } + + // Small LITERAL entries. We store (len-1) in the top 6 bits. + for (unsigned int len = 1; len <= 60; len++) { + dst[LITERAL | ((len-1) << 2)] = MakeEntry(0, len, 0); + assigned++; + } + + // Large LITERAL entries. We use 60..63 in the high 6 bits to + // encode the number of bytes of length info that follow the opcode. + for (unsigned int extra_bytes = 1; extra_bytes <= 4; extra_bytes++) { + // We set the length field in the lookup table to 1 because extra + // bytes encode len-1. + dst[LITERAL | ((extra_bytes+59) << 2)] = MakeEntry(extra_bytes, 1, 0); + assigned++; + } + + // COPY_1_BYTE_OFFSET. + // + // The tag byte in the compressed data stores len-4 in 3 bits, and + // offset/256 in 5 bits. offset%256 is stored in the next byte. + // + // This format is used for length in range [4..11] and offset in + // range [0..2047] + for (unsigned int len = 4; len < 12; len++) { + for (unsigned int offset = 0; offset < 2048; offset += 256) { + dst[COPY_1_BYTE_OFFSET | ((len-4)<<2) | ((offset>>8)<<5)] = + MakeEntry(1, len, offset>>8); + assigned++; + } + } + + // COPY_2_BYTE_OFFSET. + // Tag contains len-1 in top 6 bits, and offset in next two bytes. + for (unsigned int len = 1; len <= 64; len++) { + dst[COPY_2_BYTE_OFFSET | ((len-1)<<2)] = MakeEntry(2, len, 0); + assigned++; + } + + // COPY_4_BYTE_OFFSET. + // Tag contents len-1 in top 6 bits, and offset in next four bytes. + for (unsigned int len = 1; len <= 64; len++) { + dst[COPY_4_BYTE_OFFSET | ((len-1)<<2)] = MakeEntry(4, len, 0); + assigned++; + } + + // Check that each entry was initialized exactly once. + CHECK_EQ(assigned, 256); + for (int i = 0; i < 256; i++) { + CHECK_NE(dst[i], 0xffff); + } + + if (FLAGS_snappy_dump_decompression_table) { + printf("static const uint16 char_table[256] = {\n "); + for (int i = 0; i < 256; i++) { + printf("0x%04x%s", + dst[i], + ((i == 255) ? "\n" : (((i%8) == 7) ? ",\n " : ", "))); + } + printf("};\n"); + } + + // Check that computed table matched recorded table + for (int i = 0; i < 256; i++) { + CHECK_EQ(dst[i], char_table[i]); + } +} +REGISTER_MODULE_INITIALIZER(snappy, ComputeTable()); +#endif /* !NDEBUG */ + +// Helper class for decompression +class SnappyDecompressor { + private: + Source* reader_; // Underlying source of bytes to decompress + const char* ip_; // Points to next buffered byte + const char* ip_limit_; // Points just past buffered bytes + uint32 peeked_; // Bytes peeked from reader (need to skip) + bool eof_; // Hit end of input without an error? + char scratch_[5]; // Temporary buffer for PeekFast() boundaries + + // Ensure that all of the tag metadata for the next tag is available + // in [ip_..ip_limit_-1]. Also ensures that [ip,ip+4] is readable even + // if (ip_limit_ - ip_ < 5). + // + // Returns true on success, false on error or end of input. + bool RefillTag(); + + public: + explicit SnappyDecompressor(Source* reader) + : reader_(reader), + ip_(NULL), + ip_limit_(NULL), + peeked_(0), + eof_(false) { + } + + ~SnappyDecompressor() { + // Advance past any bytes we peeked at from the reader + reader_->Skip(peeked_); + } + + // Returns true iff we have hit the end of the input without an error. + bool eof() const { + return eof_; + } + + // Read the uncompressed length stored at the start of the compressed data. + // On succcess, stores the length in *result and returns true. + // On failure, returns false. + bool ReadUncompressedLength(uint32* result) { + DCHECK(ip_ == NULL); // Must not have read anything yet + // Length is encoded in 1..5 bytes + *result = 0; + uint32 shift = 0; + while (true) { + if (shift >= 32) return false; + size_t n; + const char* ip = reader_->Peek(&n); + if (n == 0) return false; + const unsigned char c = *(reinterpret_cast<const unsigned char*>(ip)); + reader_->Skip(1); + *result |= static_cast<uint32>(c & 0x7f) << shift; + if (c < 128) { + break; + } + shift += 7; + } + return true; + } + + // Process the next item found in the input. + // Returns true if successful, false on error or end of input. + template <class Writer> + void DecompressAllTags(Writer* writer) { + const char* ip = ip_; + for ( ;; ) { + if (ip_limit_ - ip < 5) { + ip_ = ip; + if (!RefillTag()) return; + ip = ip_; + } + + const unsigned char c = *(reinterpret_cast<const unsigned char*>(ip++)); + const uint32 entry = char_table[c]; + const uint32 trailer = LittleEndian::Load32(ip) & wordmask[entry >> 11]; + ip += entry >> 11; + const uint32 length = entry & 0xff; + + if ((c & 0x3) == LITERAL) { + uint32 literal_length = length + trailer; + uint32 avail = ip_limit_ - ip; + while (avail < literal_length) { + bool allow_fast_path = (avail >= 16); + if (!writer->Append(ip, avail, allow_fast_path)) return; + literal_length -= avail; + reader_->Skip(peeked_); + size_t n; + ip = reader_->Peek(&n); + avail = n; + peeked_ = avail; + if (avail == 0) return; // Premature end of input + ip_limit_ = ip + avail; + } + bool allow_fast_path = (avail >= 16); + if (!writer->Append(ip, literal_length, allow_fast_path)) { + return; + } + ip += literal_length; + } else { + // copy_offset/256 is encoded in bits 8..10. By just fetching + // those bits, we get copy_offset (since the bit-field starts at + // bit 8). + const uint32 copy_offset = entry & 0x700; + if (!writer->AppendFromSelf(copy_offset + trailer, length)) { + return; + } + } + } + } +}; + +bool SnappyDecompressor::RefillTag() { + const char* ip = ip_; + if (ip == ip_limit_) { + // Fetch a new fragment from the reader + reader_->Skip(peeked_); // All peeked bytes are used up + size_t n; + ip = reader_->Peek(&n); + peeked_ = n; + if (n == 0) { + eof_ = true; + return false; + } + ip_limit_ = ip + n; + } + + // Read the tag character + DCHECK_LT(ip, ip_limit_); + const unsigned char c = *(reinterpret_cast<const unsigned char*>(ip)); + const uint32 entry = char_table[c]; + const uint32 needed = (entry >> 11) + 1; // +1 byte for 'c' + DCHECK_LE(needed, sizeof(scratch_)); + + // Read more bytes from reader if needed + uint32 nbuf = ip_limit_ - ip; + if (nbuf < needed) { + // Stitch together bytes from ip and reader to form the word + // contents. We store the needed bytes in "scratch_". They + // will be consumed immediately by the caller since we do not + // read more than we need. + memmove(scratch_, ip, nbuf); + reader_->Skip(peeked_); // All peeked bytes are used up + peeked_ = 0; + while (nbuf < needed) { + size_t length; + const char* src = reader_->Peek(&length); + if (length == 0) return false; + uint32 to_add = min<uint32>(needed - nbuf, length); + memcpy(scratch_ + nbuf, src, to_add); + nbuf += to_add; + reader_->Skip(to_add); + } + DCHECK_EQ(nbuf, needed); + ip_ = scratch_; + ip_limit_ = scratch_ + needed; + } else if (nbuf < 5) { + // Have enough bytes, but move into scratch_ so that we do not + // read past end of input + memmove(scratch_, ip, nbuf); + reader_->Skip(peeked_); // All peeked bytes are used up + peeked_ = 0; + ip_ = scratch_; + ip_limit_ = scratch_ + nbuf; + } else { + // Pass pointer to buffer returned by reader_. + ip_ = ip; + } + return true; +} + +template <typename Writer> +static bool InternalUncompress(Source* r, + Writer* writer, + uint32 max_len) { + // Read the uncompressed length from the front of the compressed input + SnappyDecompressor decompressor(r); + uint32 uncompressed_len = 0; + if (!decompressor.ReadUncompressedLength(&uncompressed_len)) return false; + // Protect against possible DoS attack + if (static_cast<uint64>(uncompressed_len) > max_len) { + return false; + } + + writer->SetExpectedLength(uncompressed_len); + + // Process the entire input + decompressor.DecompressAllTags(writer); + return (decompressor.eof() && writer->CheckLength()); +} + +bool GetUncompressedLength(Source* source, uint32* result) { + SnappyDecompressor decompressor(source); + return decompressor.ReadUncompressedLength(result); +} + +size_t Compress(Source* reader, Sink* writer) { + size_t written = 0; + int N = reader->Available(); + char ulength[Varint::kMax32]; + char* p = Varint::Encode32(ulength, N); + writer->Append(ulength, p-ulength); + written += (p - ulength); + + internal::WorkingMemory wmem; + char* scratch = NULL; + char* scratch_output = NULL; + + while (N > 0) { + // Get next block to compress (without copying if possible) + size_t fragment_size; + const char* fragment = reader->Peek(&fragment_size); + DCHECK_NE(fragment_size, 0) << ": premature end of input"; + const int num_to_read = min(N, kBlockSize); + size_t bytes_read = fragment_size; + + int pending_advance = 0; + if (bytes_read >= num_to_read) { + // Buffer returned by reader is large enough + pending_advance = num_to_read; + fragment_size = num_to_read; + } else { + // Read into scratch buffer + if (scratch == NULL) { + // If this is the last iteration, we want to allocate N bytes + // of space, otherwise the max possible kBlockSize space. + // num_to_read contains exactly the correct value + scratch = new char[num_to_read]; + } + memcpy(scratch, fragment, bytes_read); + reader->Skip(bytes_read); + + while (bytes_read < num_to_read) { + fragment = reader->Peek(&fragment_size); + size_t n = min<size_t>(fragment_size, num_to_read - bytes_read); + memcpy(scratch + bytes_read, fragment, n); + bytes_read += n; + reader->Skip(n); + } + DCHECK_EQ(bytes_read, num_to_read); + fragment = scratch; + fragment_size = num_to_read; + } + DCHECK_EQ(fragment_size, num_to_read); + + // Get encoding table for compression + int table_size; + uint16* table = wmem.GetHashTable(num_to_read, &table_size); + + // Compress input_fragment and append to dest + const int max_output = MaxCompressedLength(num_to_read); + + // Need a scratch buffer for the output, in case the byte sink doesn't + // have room for us directly. + if (scratch_output == NULL) { + scratch_output = new char[max_output]; + } else { + // Since we encode kBlockSize regions followed by a region + // which is <= kBlockSize in length, a previously allocated + // scratch_output[] region is big enough for this iteration. + } + char* dest = writer->GetAppendBuffer(max_output, scratch_output); + char* end = internal::CompressFragment(fragment, fragment_size, + dest, table, table_size); + writer->Append(dest, end - dest); + written += (end - dest); + + N -= num_to_read; + reader->Skip(pending_advance); + } + + delete[] scratch; + delete[] scratch_output; + + return written; +} + +// ----------------------------------------------------------------------- +// Flat array interfaces +// ----------------------------------------------------------------------- + +// A type that writes to a flat array. +// Note that this is not a "ByteSink", but a type that matches the +// Writer template argument to SnappyDecompressor::DecompressAllTags(). +class SnappyArrayWriter { + private: + char* base_; + char* op_; + char* op_limit_; + + public: + inline explicit SnappyArrayWriter(char* dst) + : base_(dst), + op_(dst) { + } + + inline void SetExpectedLength(size_t len) { + op_limit_ = op_ + len; + } + + inline bool CheckLength() const { + return op_ == op_limit_; + } + + inline bool Append(const char* ip, uint32 len, bool allow_fast_path) { + char* op = op_; + const int space_left = op_limit_ - op; + if (allow_fast_path && len <= 16 && space_left >= 16) { + // Fast path, used for the majority (about 90%) of dynamic invocations. + UNALIGNED_STORE64(op, UNALIGNED_LOAD64(ip)); + UNALIGNED_STORE64(op + 8, UNALIGNED_LOAD64(ip + 8)); + } else { + if (space_left < len) { + return false; + } + memcpy(op, ip, len); + } + op_ = op + len; + return true; + } + + inline bool AppendFromSelf(uint32 offset, uint32 len) { + char* op = op_; + const int space_left = op_limit_ - op; + + if (op - base_ <= offset - 1u) { // -1u catches offset==0 + return false; + } + if (len <= 16 && offset >= 8 && space_left >= 16) { + // Fast path, used for the majority (70-80%) of dynamic invocations. + UNALIGNED_STORE64(op, UNALIGNED_LOAD64(op - offset)); + UNALIGNED_STORE64(op + 8, UNALIGNED_LOAD64(op - offset + 8)); + } else { + if (space_left >= len + kMaxIncrementCopyOverflow) { + IncrementalCopyFastPath(op - offset, op, len); + } else { + if (space_left < len) { + return false; + } + IncrementalCopy(op - offset, op, len); + } + } + + op_ = op + len; + return true; + } +}; + +bool RawUncompress(const char* compressed, size_t n, char* uncompressed) { + ByteArraySource reader(compressed, n); + return RawUncompress(&reader, uncompressed); +} + +bool RawUncompress(Source* compressed, char* uncompressed) { + SnappyArrayWriter output(uncompressed); + return InternalUncompress(compressed, &output, kuint32max); +} + +bool Uncompress(const char* compressed, size_t n, string* uncompressed) { + size_t ulength; + if (!GetUncompressedLength(compressed, n, &ulength)) { + return false; + } + // Protect against possible DoS attack + if ((static_cast<uint64>(ulength) + uncompressed->size()) > + uncompressed->max_size()) { + return false; + } + STLStringResizeUninitialized(uncompressed, ulength); + return RawUncompress(compressed, n, string_as_array(uncompressed)); +} + + +// A Writer that drops everything on the floor and just does validation +class SnappyDecompressionValidator { + private: + size_t expected_; + size_t produced_; + + public: + inline SnappyDecompressionValidator() : produced_(0) { } + inline void SetExpectedLength(size_t len) { + expected_ = len; + } + inline bool CheckLength() const { + return expected_ == produced_; + } + inline bool Append(const char* ip, uint32 len, bool allow_fast_path) { + produced_ += len; + return produced_ <= expected_; + } + inline bool AppendFromSelf(uint32 offset, uint32 len) { + if (produced_ <= offset - 1u) return false; // -1u catches offset==0 + produced_ += len; + return produced_ <= expected_; + } +}; + +bool IsValidCompressedBuffer(const char* compressed, size_t n) { + ByteArraySource reader(compressed, n); + SnappyDecompressionValidator writer; + return InternalUncompress(&reader, &writer, kuint32max); +} + +void RawCompress(const char* input, + size_t input_length, + char* compressed, + size_t* compressed_length) { + ByteArraySource reader(input, input_length); + UncheckedByteArraySink writer(compressed); + Compress(&reader, &writer); + + // Compute how many bytes were added + *compressed_length = (writer.CurrentDestination() - compressed); +} + +size_t Compress(const char* input, size_t input_length, string* compressed) { + // Pre-grow the buffer to the max length of the compressed output + compressed->resize(MaxCompressedLength(input_length)); + + size_t compressed_length; + RawCompress(input, input_length, string_as_array(compressed), + &compressed_length); + compressed->resize(compressed_length); + return compressed_length; +} + + +} // end namespace snappy + diff --git a/third_party/snappy/snappy.h b/third_party/snappy/snappy.h new file mode 100755 index 00000000000..8d6ef2294f5 --- /dev/null +++ b/third_party/snappy/snappy.h @@ -0,0 +1,155 @@ +// 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 UTIL_SNAPPY_SNAPPY_H__ +#define UTIL_SNAPPY_SNAPPY_H__ + +#include <stddef.h> +#include <string> + +#include "snappy-stubs-public.h" + +namespace 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); + + 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); + + + // ------------------------------------------------------------------------ + // 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); + + // 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); + + // *** DO NOT CHANGE THE VALUE OF kBlockSize *** + // + // New Compression code chops up the input into blocks of at most + // the following size. This ensures that back-references in the + // output never cross kBlockSize block boundaries. This can be + // helpful in implementing blocked decompression. However the + // decompression code should not rely on this guarantee since older + // compression code may not obey it. + static const int kBlockLog = 15; + static const int kBlockSize = 1 << kBlockLog; + + static const int kMaxHashTableBits = 14; + static const int kMaxHashTableSize = 1 << kMaxHashTableBits; + +} // end namespace snappy + + +#endif // UTIL_SNAPPY_SNAPPY_H__ |