snappy 0.0.14-java → 0.2.0-java
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +5 -5
- data/.github/workflows/main.yml +34 -0
- data/.github/workflows/publish.yml +34 -0
- data/Gemfile +4 -0
- data/README.md +28 -4
- data/Rakefile +32 -29
- data/ext/api.c +6 -1
- data/ext/extconf.rb +21 -24
- data/lib/snappy.rb +6 -4
- data/lib/snappy/hadoop.rb +22 -0
- data/lib/snappy/hadoop/reader.rb +62 -0
- data/lib/snappy/hadoop/writer.rb +51 -0
- data/lib/snappy/reader.rb +19 -11
- data/lib/snappy/shim.rb +30 -0
- data/lib/snappy/version.rb +3 -1
- data/lib/snappy/writer.rb +8 -9
- data/snappy.gemspec +17 -37
- data/test/hadoop/snappy_hadoop_reader_test.rb +115 -0
- data/test/hadoop/snappy_hadoop_writer_test.rb +48 -0
- data/test/snappy_hadoop_test.rb +26 -0
- data/test/snappy_reader_test.rb +148 -0
- data/test/snappy_test.rb +95 -0
- data/test/snappy_writer_test.rb +55 -0
- data/test/test_helper.rb +7 -0
- data/vendor/snappy/CMakeLists.txt +297 -0
- data/vendor/snappy/CONTRIBUTING.md +26 -0
- data/vendor/snappy/COPYING +1 -1
- data/vendor/snappy/NEWS +60 -0
- data/vendor/snappy/{README → README.md} +29 -16
- data/vendor/snappy/cmake/SnappyConfig.cmake.in +33 -0
- data/vendor/snappy/cmake/config.h.in +62 -0
- data/vendor/snappy/docs/README.md +72 -0
- data/vendor/snappy/snappy-c.h +3 -3
- data/vendor/snappy/snappy-internal.h +113 -32
- data/vendor/snappy/snappy-sinksource.cc +33 -0
- data/vendor/snappy/snappy-sinksource.h +51 -6
- data/vendor/snappy/snappy-stubs-internal.cc +1 -1
- data/vendor/snappy/snappy-stubs-internal.h +160 -45
- data/vendor/snappy/snappy-stubs-public.h.in +23 -47
- data/vendor/snappy/snappy-test.cc +31 -24
- data/vendor/snappy/snappy-test.h +46 -103
- data/vendor/snappy/snappy.cc +786 -431
- data/vendor/snappy/snappy.h +37 -14
- data/vendor/snappy/snappy_compress_fuzzer.cc +59 -0
- data/vendor/snappy/snappy_uncompress_fuzzer.cc +57 -0
- data/vendor/snappy/snappy_unittest.cc +441 -290
- metadata +35 -75
- data/.travis.yml +0 -4
- data/test/test-snappy-reader.rb +0 -129
- data/test/test-snappy-writer.rb +0 -55
- data/test/test-snappy.rb +0 -58
- data/vendor/snappy/ChangeLog +0 -1916
- data/vendor/snappy/Makefile.am +0 -23
- data/vendor/snappy/autogen.sh +0 -7
- data/vendor/snappy/configure.ac +0 -133
- data/vendor/snappy/m4/gtest.m4 +0 -74
- data/vendor/snappy/testdata/alice29.txt +0 -3609
- data/vendor/snappy/testdata/asyoulik.txt +0 -4122
- data/vendor/snappy/testdata/baddata1.snappy +0 -0
- data/vendor/snappy/testdata/baddata2.snappy +0 -0
- data/vendor/snappy/testdata/baddata3.snappy +0 -0
- data/vendor/snappy/testdata/fireworks.jpeg +0 -0
- data/vendor/snappy/testdata/geo.protodata +0 -0
- data/vendor/snappy/testdata/html +0 -1
- data/vendor/snappy/testdata/html_x_4 +0 -1
- data/vendor/snappy/testdata/kppkn.gtb +0 -0
- data/vendor/snappy/testdata/lcet10.txt +0 -7519
- data/vendor/snappy/testdata/paper-100k.pdf +2 -600
- data/vendor/snappy/testdata/plrabn12.txt +0 -10699
- data/vendor/snappy/testdata/urls.10K +0 -10000
data/vendor/snappy/snappy.h
CHANGED
@@ -36,10 +36,10 @@
|
|
36
36
|
// using BMDiff and then compressing the output of BMDiff with
|
37
37
|
// Snappy.
|
38
38
|
|
39
|
-
#ifndef
|
40
|
-
#define
|
39
|
+
#ifndef THIRD_PARTY_SNAPPY_SNAPPY_H__
|
40
|
+
#define THIRD_PARTY_SNAPPY_SNAPPY_H__
|
41
41
|
|
42
|
-
#include <
|
42
|
+
#include <cstddef>
|
43
43
|
#include <string>
|
44
44
|
|
45
45
|
#include "snappy-stubs-public.h"
|
@@ -69,11 +69,12 @@ namespace snappy {
|
|
69
69
|
// Higher-level string based routines (should be sufficient for most users)
|
70
70
|
// ------------------------------------------------------------------------
|
71
71
|
|
72
|
-
// Sets "*
|
73
|
-
// Original contents of *
|
72
|
+
// Sets "*compressed" to the compressed version of "input[0,input_length-1]".
|
73
|
+
// Original contents of *compressed are lost.
|
74
74
|
//
|
75
|
-
// REQUIRES: "input[]" is not an alias of "*
|
76
|
-
size_t Compress(const char* input, size_t input_length,
|
75
|
+
// REQUIRES: "input[]" is not an alias of "*compressed".
|
76
|
+
size_t Compress(const char* input, size_t input_length,
|
77
|
+
std::string* compressed);
|
77
78
|
|
78
79
|
// Decompresses "compressed[0,compressed_length-1]" to "*uncompressed".
|
79
80
|
// Original contents of "*uncompressed" are lost.
|
@@ -82,8 +83,20 @@ namespace snappy {
|
|
82
83
|
//
|
83
84
|
// returns false if the message is corrupted and could not be decompressed
|
84
85
|
bool Uncompress(const char* compressed, size_t compressed_length,
|
85
|
-
string* uncompressed);
|
86
|
+
std::string* uncompressed);
|
86
87
|
|
88
|
+
// Decompresses "compressed" to "*uncompressed".
|
89
|
+
//
|
90
|
+
// returns false if the message is corrupted and could not be decompressed
|
91
|
+
bool Uncompress(Source* compressed, Sink* uncompressed);
|
92
|
+
|
93
|
+
// This routine uncompresses as much of the "compressed" as possible
|
94
|
+
// into sink. It returns the number of valid bytes added to sink
|
95
|
+
// (extra invalid bytes may have been added due to errors; the caller
|
96
|
+
// should ignore those). The emitted data typically has length
|
97
|
+
// GetUncompressedLength(), but may be shorter if an error is
|
98
|
+
// encountered.
|
99
|
+
size_t UncompressAsMuchAsPossible(Source* compressed, Sink* uncompressed);
|
87
100
|
|
88
101
|
// ------------------------------------------------------------------------
|
89
102
|
// Lower-level character array based routines. May be useful for
|
@@ -164,6 +177,14 @@ namespace snappy {
|
|
164
177
|
bool IsValidCompressedBuffer(const char* compressed,
|
165
178
|
size_t compressed_length);
|
166
179
|
|
180
|
+
// Returns true iff the contents of "compressed" can be uncompressed
|
181
|
+
// successfully. Does not return the uncompressed data. Takes
|
182
|
+
// time proportional to *compressed length, but is usually at least
|
183
|
+
// a factor of four faster than actual decompression.
|
184
|
+
// On success, consumes all of *compressed. On failure, consumes an
|
185
|
+
// unspecified prefix of *compressed.
|
186
|
+
bool IsValidCompressed(Source* compressed);
|
187
|
+
|
167
188
|
// The size of a compression block. Note that many parts of the compression
|
168
189
|
// code assumes that kBlockSize <= 65536; in particular, the hash table
|
169
190
|
// can only store 16-bit offsets, and EmitCopy() also assumes the offset
|
@@ -173,12 +194,14 @@ namespace snappy {
|
|
173
194
|
// Note that there might be older data around that is compressed with larger
|
174
195
|
// block sizes, so the decompression code should not rely on the
|
175
196
|
// non-existence of long backreferences.
|
176
|
-
static
|
177
|
-
static
|
197
|
+
static constexpr int kBlockLog = 16;
|
198
|
+
static constexpr size_t kBlockSize = 1 << kBlockLog;
|
178
199
|
|
179
|
-
static
|
180
|
-
static
|
181
|
-
} // end namespace snappy
|
200
|
+
static constexpr int kMinHashTableBits = 8;
|
201
|
+
static constexpr size_t kMinHashTableSize = 1 << kMinHashTableBits;
|
182
202
|
|
203
|
+
static constexpr int kMaxHashTableBits = 14;
|
204
|
+
static constexpr size_t kMaxHashTableSize = 1 << kMaxHashTableBits;
|
205
|
+
} // end namespace snappy
|
183
206
|
|
184
|
-
#endif //
|
207
|
+
#endif // THIRD_PARTY_SNAPPY_SNAPPY_H__
|
@@ -0,0 +1,59 @@
|
|
1
|
+
// Copyright 2019 Google Inc. All Rights Reserved.
|
2
|
+
//
|
3
|
+
// Redistribution and use in source and binary forms, with or without
|
4
|
+
// modification, are permitted provided that the following conditions are
|
5
|
+
// met:
|
6
|
+
//
|
7
|
+
// * Redistributions of source code must retain the above copyright
|
8
|
+
// notice, this list of conditions and the following disclaimer.
|
9
|
+
// * Redistributions in binary form must reproduce the above
|
10
|
+
// copyright notice, this list of conditions and the following disclaimer
|
11
|
+
// in the documentation and/or other materials provided with the
|
12
|
+
// distribution.
|
13
|
+
// * Neither the name of Google Inc. nor the names of its
|
14
|
+
// contributors may be used to endorse or promote products derived from
|
15
|
+
// this software without specific prior written permission.
|
16
|
+
//
|
17
|
+
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
18
|
+
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
19
|
+
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
|
20
|
+
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
|
21
|
+
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
|
22
|
+
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
|
23
|
+
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
|
24
|
+
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
25
|
+
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
26
|
+
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
27
|
+
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
28
|
+
//
|
29
|
+
// libFuzzer harness for fuzzing snappy compression code.
|
30
|
+
|
31
|
+
#include <cassert>
|
32
|
+
#include <cstddef>
|
33
|
+
#include <cstdint>
|
34
|
+
#include <string>
|
35
|
+
|
36
|
+
#include "snappy.h"
|
37
|
+
|
38
|
+
// Entry point for LibFuzzer.
|
39
|
+
extern "C" int LLVMFuzzerTestOneInput(const uint8_t* data, size_t size) {
|
40
|
+
std::string input(reinterpret_cast<const char*>(data), size);
|
41
|
+
|
42
|
+
std::string compressed;
|
43
|
+
size_t compressed_size =
|
44
|
+
snappy::Compress(input.data(), input.size(), &compressed);
|
45
|
+
|
46
|
+
(void)compressed_size; // Variable only used in debug builds.
|
47
|
+
assert(compressed_size == compressed.size());
|
48
|
+
assert(compressed.size() <= snappy::MaxCompressedLength(input.size()));
|
49
|
+
assert(snappy::IsValidCompressedBuffer(compressed.data(), compressed.size()));
|
50
|
+
|
51
|
+
std::string uncompressed_after_compress;
|
52
|
+
bool uncompress_succeeded = snappy::Uncompress(
|
53
|
+
compressed.data(), compressed.size(), &uncompressed_after_compress);
|
54
|
+
|
55
|
+
(void)uncompress_succeeded; // Variable only used in debug builds.
|
56
|
+
assert(uncompress_succeeded);
|
57
|
+
assert(input == uncompressed_after_compress);
|
58
|
+
return 0;
|
59
|
+
}
|
@@ -0,0 +1,57 @@
|
|
1
|
+
// Copyright 2019 Google Inc. All Rights Reserved.
|
2
|
+
//
|
3
|
+
// Redistribution and use in source and binary forms, with or without
|
4
|
+
// modification, are permitted provided that the following conditions are
|
5
|
+
// met:
|
6
|
+
//
|
7
|
+
// * Redistributions of source code must retain the above copyright
|
8
|
+
// notice, this list of conditions and the following disclaimer.
|
9
|
+
// * Redistributions in binary form must reproduce the above
|
10
|
+
// copyright notice, this list of conditions and the following disclaimer
|
11
|
+
// in the documentation and/or other materials provided with the
|
12
|
+
// distribution.
|
13
|
+
// * Neither the name of Google Inc. nor the names of its
|
14
|
+
// contributors may be used to endorse or promote products derived from
|
15
|
+
// this software without specific prior written permission.
|
16
|
+
//
|
17
|
+
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
18
|
+
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
19
|
+
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
|
20
|
+
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
|
21
|
+
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
|
22
|
+
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
|
23
|
+
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
|
24
|
+
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
25
|
+
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
26
|
+
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
27
|
+
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
28
|
+
//
|
29
|
+
// libFuzzer harness for fuzzing snappy's decompression code.
|
30
|
+
|
31
|
+
#include <cassert>
|
32
|
+
#include <cstddef>
|
33
|
+
#include <cstdint>
|
34
|
+
#include <string>
|
35
|
+
|
36
|
+
#include "snappy.h"
|
37
|
+
|
38
|
+
// Entry point for LibFuzzer.
|
39
|
+
extern "C" int LLVMFuzzerTestOneInput(const uint8_t* data, size_t size) {
|
40
|
+
std::string input(reinterpret_cast<const char*>(data), size);
|
41
|
+
|
42
|
+
// Avoid self-crafted decompression bombs.
|
43
|
+
size_t uncompressed_size;
|
44
|
+
constexpr size_t kMaxUncompressedSize = 1 << 20;
|
45
|
+
bool get_uncompressed_length_succeeded = snappy::GetUncompressedLength(
|
46
|
+
input.data(), input.size(), &uncompressed_size);
|
47
|
+
if (!get_uncompressed_length_succeeded ||
|
48
|
+
(uncompressed_size > kMaxUncompressedSize)) {
|
49
|
+
return 0;
|
50
|
+
}
|
51
|
+
|
52
|
+
std::string uncompressed;
|
53
|
+
// The return value of snappy::Uncompress() is ignored because decompression
|
54
|
+
// will fail on invalid inputs.
|
55
|
+
snappy::Uncompress(input.data(), input.size(), &uncompressed);
|
56
|
+
return 0;
|
57
|
+
}
|
@@ -29,9 +29,10 @@
|
|
29
29
|
#include <math.h>
|
30
30
|
#include <stdlib.h>
|
31
31
|
|
32
|
-
|
33
32
|
#include <algorithm>
|
33
|
+
#include <random>
|
34
34
|
#include <string>
|
35
|
+
#include <utility>
|
35
36
|
#include <vector>
|
36
37
|
|
37
38
|
#include "snappy.h"
|
@@ -50,25 +51,19 @@ DEFINE_bool(zlib, false,
|
|
50
51
|
"Run zlib compression (http://www.zlib.net)");
|
51
52
|
DEFINE_bool(lzo, false,
|
52
53
|
"Run LZO compression (http://www.oberhumer.com/opensource/lzo/)");
|
53
|
-
DEFINE_bool(quicklz, false,
|
54
|
-
"Run quickLZ compression (http://www.quicklz.com/)");
|
55
|
-
DEFINE_bool(liblzf, false,
|
56
|
-
"Run libLZF compression "
|
57
|
-
"(http://www.goof.com/pcg/marc/liblzf.html)");
|
58
|
-
DEFINE_bool(fastlz, false,
|
59
|
-
"Run FastLZ compression (http://www.fastlz.org/");
|
60
54
|
DEFINE_bool(snappy, true, "Run snappy compression");
|
61
55
|
|
62
|
-
|
63
56
|
DEFINE_bool(write_compressed, false,
|
64
57
|
"Write compressed versions of each file to <file>.comp");
|
65
58
|
DEFINE_bool(write_uncompressed, false,
|
66
59
|
"Write uncompressed versions of each file to <file>.uncomp");
|
67
60
|
|
68
|
-
|
61
|
+
DEFINE_bool(snappy_dump_decompression_table, false,
|
62
|
+
"If true, we print the decompression table during tests.");
|
69
63
|
|
64
|
+
namespace snappy {
|
70
65
|
|
71
|
-
#
|
66
|
+
#if defined(HAVE_FUNC_MMAP) && defined(HAVE_FUNC_SYSCONF)
|
72
67
|
|
73
68
|
// To test against code that reads beyond its input, this class copies a
|
74
69
|
// string to a newly allocated group of pages, the last of which
|
@@ -78,8 +73,8 @@ namespace snappy {
|
|
78
73
|
// be able to read previously allocated memory while doing heap allocations.
|
79
74
|
class DataEndingAtUnreadablePage {
|
80
75
|
public:
|
81
|
-
explicit DataEndingAtUnreadablePage(const string& s) {
|
82
|
-
const size_t page_size =
|
76
|
+
explicit DataEndingAtUnreadablePage(const std::string& s) {
|
77
|
+
const size_t page_size = sysconf(_SC_PAGESIZE);
|
83
78
|
const size_t size = s.size();
|
84
79
|
// Round up space for string to a multiple of page_size.
|
85
80
|
size_t space_for_string = (size + page_size - 1) & ~(page_size - 1);
|
@@ -97,8 +92,9 @@ class DataEndingAtUnreadablePage {
|
|
97
92
|
}
|
98
93
|
|
99
94
|
~DataEndingAtUnreadablePage() {
|
95
|
+
const size_t page_size = sysconf(_SC_PAGESIZE);
|
100
96
|
// Undo the mprotect.
|
101
|
-
CHECK_EQ(0, mprotect(protected_page_,
|
97
|
+
CHECK_EQ(0, mprotect(protected_page_, page_size, PROT_READ|PROT_WRITE));
|
102
98
|
CHECK_EQ(0, munmap(mem_, alloc_size_));
|
103
99
|
}
|
104
100
|
|
@@ -113,19 +109,19 @@ class DataEndingAtUnreadablePage {
|
|
113
109
|
size_t size_;
|
114
110
|
};
|
115
111
|
|
116
|
-
#else // HAVE_FUNC_MMAP
|
112
|
+
#else // defined(HAVE_FUNC_MMAP) && defined(HAVE_FUNC_SYSCONF)
|
117
113
|
|
118
114
|
// Fallback for systems without mmap.
|
119
|
-
|
115
|
+
using DataEndingAtUnreadablePage = std::string;
|
120
116
|
|
121
117
|
#endif
|
122
118
|
|
123
119
|
enum CompressorType {
|
124
|
-
ZLIB, LZO,
|
120
|
+
ZLIB, LZO, SNAPPY
|
125
121
|
};
|
126
122
|
|
127
123
|
const char* names[] = {
|
128
|
-
"ZLIB", "LZO", "
|
124
|
+
"ZLIB", "LZO", "SNAPPY"
|
129
125
|
};
|
130
126
|
|
131
127
|
static size_t MinimumRequiredOutputSpace(size_t input_size,
|
@@ -141,26 +137,12 @@ static size_t MinimumRequiredOutputSpace(size_t input_size,
|
|
141
137
|
return input_size + input_size/64 + 16 + 3;
|
142
138
|
#endif // LZO_VERSION
|
143
139
|
|
144
|
-
#ifdef LZF_VERSION
|
145
|
-
case LIBLZF:
|
146
|
-
return input_size;
|
147
|
-
#endif // LZF_VERSION
|
148
|
-
|
149
|
-
#ifdef QLZ_VERSION_MAJOR
|
150
|
-
case QUICKLZ:
|
151
|
-
return input_size + 36000; // 36000 is used for scratch.
|
152
|
-
#endif // QLZ_VERSION_MAJOR
|
153
|
-
|
154
|
-
#ifdef FASTLZ_VERSION
|
155
|
-
case FASTLZ:
|
156
|
-
return max(static_cast<int>(ceil(input_size * 1.05)), 66);
|
157
|
-
#endif // FASTLZ_VERSION
|
158
|
-
|
159
140
|
case SNAPPY:
|
160
141
|
return snappy::MaxCompressedLength(input_size);
|
161
142
|
|
162
143
|
default:
|
163
144
|
LOG(FATAL) << "Unknown compression type number " << comp;
|
145
|
+
return 0;
|
164
146
|
}
|
165
147
|
}
|
166
148
|
|
@@ -172,7 +154,7 @@ static size_t MinimumRequiredOutputSpace(size_t input_size,
|
|
172
154
|
// "compressed" must be preinitialized to at least MinCompressbufSize(comp)
|
173
155
|
// number of bytes, and may contain junk bytes at the end after return.
|
174
156
|
static bool Compress(const char* input, size_t input_size, CompressorType comp,
|
175
|
-
string* compressed, bool compressed_is_preallocated) {
|
157
|
+
std::string* compressed, bool compressed_is_preallocated) {
|
176
158
|
if (!compressed_is_preallocated) {
|
177
159
|
compressed->resize(MinimumRequiredOutputSpace(input_size, comp));
|
178
160
|
}
|
@@ -214,58 +196,6 @@ static bool Compress(const char* input, size_t input_size, CompressorType comp,
|
|
214
196
|
}
|
215
197
|
#endif // LZO_VERSION
|
216
198
|
|
217
|
-
#ifdef LZF_VERSION
|
218
|
-
case LIBLZF: {
|
219
|
-
int destlen = lzf_compress(input,
|
220
|
-
input_size,
|
221
|
-
string_as_array(compressed),
|
222
|
-
input_size);
|
223
|
-
if (destlen == 0) {
|
224
|
-
// lzf *can* cause lots of blowup when compressing, so they
|
225
|
-
// recommend to limit outsize to insize, and just not compress
|
226
|
-
// if it's bigger. Ideally, we'd just swap input and output.
|
227
|
-
compressed->assign(input, input_size);
|
228
|
-
destlen = input_size;
|
229
|
-
}
|
230
|
-
if (!compressed_is_preallocated) {
|
231
|
-
compressed->resize(destlen);
|
232
|
-
}
|
233
|
-
break;
|
234
|
-
}
|
235
|
-
#endif // LZF_VERSION
|
236
|
-
|
237
|
-
#ifdef QLZ_VERSION_MAJOR
|
238
|
-
case QUICKLZ: {
|
239
|
-
qlz_state_compress *state_compress = new qlz_state_compress;
|
240
|
-
int destlen = qlz_compress(input,
|
241
|
-
string_as_array(compressed),
|
242
|
-
input_size,
|
243
|
-
state_compress);
|
244
|
-
delete state_compress;
|
245
|
-
CHECK_NE(0, destlen);
|
246
|
-
if (!compressed_is_preallocated) {
|
247
|
-
compressed->resize(destlen);
|
248
|
-
}
|
249
|
-
break;
|
250
|
-
}
|
251
|
-
#endif // QLZ_VERSION_MAJOR
|
252
|
-
|
253
|
-
#ifdef FASTLZ_VERSION
|
254
|
-
case FASTLZ: {
|
255
|
-
// Use level 1 compression since we mostly care about speed.
|
256
|
-
int destlen = fastlz_compress_level(
|
257
|
-
1,
|
258
|
-
input,
|
259
|
-
input_size,
|
260
|
-
string_as_array(compressed));
|
261
|
-
if (!compressed_is_preallocated) {
|
262
|
-
compressed->resize(destlen);
|
263
|
-
}
|
264
|
-
CHECK_NE(destlen, 0);
|
265
|
-
break;
|
266
|
-
}
|
267
|
-
#endif // FASTLZ_VERSION
|
268
|
-
|
269
199
|
case SNAPPY: {
|
270
200
|
size_t destlen;
|
271
201
|
snappy::RawCompress(input, input_size,
|
@@ -278,7 +208,6 @@ static bool Compress(const char* input, size_t input_size, CompressorType comp,
|
|
278
208
|
break;
|
279
209
|
}
|
280
210
|
|
281
|
-
|
282
211
|
default: {
|
283
212
|
return false; // the asked-for library wasn't compiled in
|
284
213
|
}
|
@@ -286,8 +215,8 @@ static bool Compress(const char* input, size_t input_size, CompressorType comp,
|
|
286
215
|
return true;
|
287
216
|
}
|
288
217
|
|
289
|
-
static bool Uncompress(const string& compressed, CompressorType comp,
|
290
|
-
int size, string* output) {
|
218
|
+
static bool Uncompress(const std::string& compressed, CompressorType comp,
|
219
|
+
int size, std::string* output) {
|
291
220
|
switch (comp) {
|
292
221
|
#ifdef ZLIB_VERSION
|
293
222
|
case ZLIB: {
|
@@ -321,56 +250,12 @@ static bool Uncompress(const string& compressed, CompressorType comp,
|
|
321
250
|
}
|
322
251
|
#endif // LZO_VERSION
|
323
252
|
|
324
|
-
#ifdef LZF_VERSION
|
325
|
-
case LIBLZF: {
|
326
|
-
output->resize(size);
|
327
|
-
int destlen = lzf_decompress(compressed.data(),
|
328
|
-
compressed.size(),
|
329
|
-
string_as_array(output),
|
330
|
-
output->size());
|
331
|
-
if (destlen == 0) {
|
332
|
-
// This error probably means we had decided not to compress,
|
333
|
-
// and thus have stored input in output directly.
|
334
|
-
output->assign(compressed.data(), compressed.size());
|
335
|
-
destlen = compressed.size();
|
336
|
-
}
|
337
|
-
CHECK_EQ(destlen, size);
|
338
|
-
break;
|
339
|
-
}
|
340
|
-
#endif // LZF_VERSION
|
341
|
-
|
342
|
-
#ifdef QLZ_VERSION_MAJOR
|
343
|
-
case QUICKLZ: {
|
344
|
-
output->resize(size);
|
345
|
-
qlz_state_decompress *state_decompress = new qlz_state_decompress;
|
346
|
-
int destlen = qlz_decompress(compressed.data(),
|
347
|
-
string_as_array(output),
|
348
|
-
state_decompress);
|
349
|
-
delete state_decompress;
|
350
|
-
CHECK_EQ(destlen, size);
|
351
|
-
break;
|
352
|
-
}
|
353
|
-
#endif // QLZ_VERSION_MAJOR
|
354
|
-
|
355
|
-
#ifdef FASTLZ_VERSION
|
356
|
-
case FASTLZ: {
|
357
|
-
output->resize(size);
|
358
|
-
int destlen = fastlz_decompress(compressed.data(),
|
359
|
-
compressed.length(),
|
360
|
-
string_as_array(output),
|
361
|
-
size);
|
362
|
-
CHECK_EQ(destlen, size);
|
363
|
-
break;
|
364
|
-
}
|
365
|
-
#endif // FASTLZ_VERSION
|
366
|
-
|
367
253
|
case SNAPPY: {
|
368
254
|
snappy::RawUncompress(compressed.data(), compressed.size(),
|
369
255
|
string_as_array(output));
|
370
256
|
break;
|
371
257
|
}
|
372
258
|
|
373
|
-
|
374
259
|
default: {
|
375
260
|
return false; // the asked-for library wasn't compiled in
|
376
261
|
}
|
@@ -392,13 +277,13 @@ static void Measure(const char* data,
|
|
392
277
|
{
|
393
278
|
// Chop the input into blocks
|
394
279
|
int num_blocks = (length + block_size - 1) / block_size;
|
395
|
-
vector<const char*> input(num_blocks);
|
396
|
-
vector<size_t> input_length(num_blocks);
|
397
|
-
vector<string> compressed(num_blocks);
|
398
|
-
vector<string> output(num_blocks);
|
280
|
+
std::vector<const char*> input(num_blocks);
|
281
|
+
std::vector<size_t> input_length(num_blocks);
|
282
|
+
std::vector<std::string> compressed(num_blocks);
|
283
|
+
std::vector<std::string> output(num_blocks);
|
399
284
|
for (int b = 0; b < num_blocks; b++) {
|
400
285
|
int input_start = b * block_size;
|
401
|
-
int input_limit = min<int>((b+1)*block_size, length);
|
286
|
+
int input_limit = std::min<int>((b+1)*block_size, length);
|
402
287
|
input[b] = data+input_start;
|
403
288
|
input_length[b] = input_limit-input_start;
|
404
289
|
|
@@ -448,35 +333,33 @@ static void Measure(const char* data,
|
|
448
333
|
}
|
449
334
|
|
450
335
|
compressed_size = 0;
|
451
|
-
for (
|
336
|
+
for (size_t i = 0; i < compressed.size(); i++) {
|
452
337
|
compressed_size += compressed[i].size();
|
453
338
|
}
|
454
339
|
}
|
455
340
|
|
456
|
-
sort(ctime, ctime + kRuns);
|
457
|
-
sort(utime, utime + kRuns);
|
341
|
+
std::sort(ctime, ctime + kRuns);
|
342
|
+
std::sort(utime, utime + kRuns);
|
458
343
|
const int med = kRuns/2;
|
459
344
|
|
460
345
|
float comp_rate = (length / ctime[med]) * repeats / 1048576.0;
|
461
346
|
float uncomp_rate = (length / utime[med]) * repeats / 1048576.0;
|
462
|
-
string x = names[comp];
|
347
|
+
std::string x = names[comp];
|
463
348
|
x += ":";
|
464
|
-
string urate = (uncomp_rate >= 0)
|
465
|
-
|
466
|
-
: string("?");
|
349
|
+
std::string urate = (uncomp_rate >= 0) ? StrFormat("%.1f", uncomp_rate)
|
350
|
+
: std::string("?");
|
467
351
|
printf("%-7s [b %dM] bytes %6d -> %6d %4.1f%% "
|
468
352
|
"comp %5.1f MB/s uncomp %5s MB/s\n",
|
469
353
|
x.c_str(),
|
470
354
|
block_size/(1<<20),
|
471
355
|
static_cast<int>(length), static_cast<uint32>(compressed_size),
|
472
|
-
(compressed_size * 100.0) / max<int>(1, length),
|
356
|
+
(compressed_size * 100.0) / std::max<int>(1, length),
|
473
357
|
comp_rate,
|
474
358
|
urate.c_str());
|
475
359
|
}
|
476
360
|
|
477
|
-
|
478
|
-
|
479
|
-
string compressed;
|
361
|
+
static int VerifyString(const std::string& input) {
|
362
|
+
std::string compressed;
|
480
363
|
DataEndingAtUnreadablePage i(input);
|
481
364
|
const size_t written = snappy::Compress(i.data(), i.size(), &compressed);
|
482
365
|
CHECK_EQ(written, compressed.size());
|
@@ -484,16 +367,33 @@ static int VerifyString(const string& input) {
|
|
484
367
|
snappy::MaxCompressedLength(input.size()));
|
485
368
|
CHECK(snappy::IsValidCompressedBuffer(compressed.data(), compressed.size()));
|
486
369
|
|
487
|
-
string uncompressed;
|
370
|
+
std::string uncompressed;
|
488
371
|
DataEndingAtUnreadablePage c(compressed);
|
489
372
|
CHECK(snappy::Uncompress(c.data(), c.size(), &uncompressed));
|
490
373
|
CHECK_EQ(uncompressed, input);
|
491
374
|
return uncompressed.size();
|
492
375
|
}
|
493
376
|
|
377
|
+
static void VerifyStringSink(const std::string& input) {
|
378
|
+
std::string compressed;
|
379
|
+
DataEndingAtUnreadablePage i(input);
|
380
|
+
const size_t written = snappy::Compress(i.data(), i.size(), &compressed);
|
381
|
+
CHECK_EQ(written, compressed.size());
|
382
|
+
CHECK_LE(compressed.size(),
|
383
|
+
snappy::MaxCompressedLength(input.size()));
|
384
|
+
CHECK(snappy::IsValidCompressedBuffer(compressed.data(), compressed.size()));
|
494
385
|
|
495
|
-
|
496
|
-
|
386
|
+
std::string uncompressed;
|
387
|
+
uncompressed.resize(input.size());
|
388
|
+
snappy::UncheckedByteArraySink sink(string_as_array(&uncompressed));
|
389
|
+
DataEndingAtUnreadablePage c(compressed);
|
390
|
+
snappy::ByteArraySource source(c.data(), c.size());
|
391
|
+
CHECK(snappy::Uncompress(&source, &sink));
|
392
|
+
CHECK_EQ(uncompressed, input);
|
393
|
+
}
|
394
|
+
|
395
|
+
static void VerifyIOVec(const std::string& input) {
|
396
|
+
std::string compressed;
|
497
397
|
DataEndingAtUnreadablePage i(input);
|
498
398
|
const size_t written = snappy::Compress(i.data(), i.size(), &compressed);
|
499
399
|
CHECK_EQ(written, compressed.size());
|
@@ -504,23 +404,28 @@ static void VerifyIOVec(const string& input) {
|
|
504
404
|
// Try uncompressing into an iovec containing a random number of entries
|
505
405
|
// ranging from 1 to 10.
|
506
406
|
char* buf = new char[input.size()];
|
507
|
-
|
508
|
-
|
407
|
+
std::minstd_rand0 rng(input.size());
|
408
|
+
std::uniform_int_distribution<size_t> uniform_1_to_10(1, 10);
|
409
|
+
size_t num = uniform_1_to_10(rng);
|
509
410
|
if (input.size() < num) {
|
510
411
|
num = input.size();
|
511
412
|
}
|
512
413
|
struct iovec* iov = new iovec[num];
|
513
414
|
int used_so_far = 0;
|
514
|
-
|
415
|
+
std::bernoulli_distribution one_in_five(1.0 / 5);
|
416
|
+
for (size_t i = 0; i < num; ++i) {
|
417
|
+
assert(used_so_far < input.size());
|
515
418
|
iov[i].iov_base = buf + used_so_far;
|
516
419
|
if (i == num - 1) {
|
517
420
|
iov[i].iov_len = input.size() - used_so_far;
|
518
421
|
} else {
|
519
422
|
// Randomly choose to insert a 0 byte entry.
|
520
|
-
if (
|
423
|
+
if (one_in_five(rng)) {
|
521
424
|
iov[i].iov_len = 0;
|
522
425
|
} else {
|
523
|
-
|
426
|
+
std::uniform_int_distribution<size_t> uniform_not_used_so_far(
|
427
|
+
0, input.size() - used_so_far - 1);
|
428
|
+
iov[i].iov_len = uniform_not_used_so_far(rng);
|
524
429
|
}
|
525
430
|
}
|
526
431
|
used_so_far += iov[i].iov_len;
|
@@ -534,22 +439,22 @@ static void VerifyIOVec(const string& input) {
|
|
534
439
|
|
535
440
|
// Test that data compressed by a compressor that does not
|
536
441
|
// obey block sizes is uncompressed properly.
|
537
|
-
static void VerifyNonBlockedCompression(const string& input) {
|
442
|
+
static void VerifyNonBlockedCompression(const std::string& input) {
|
538
443
|
if (input.length() > snappy::kBlockSize) {
|
539
444
|
// We cannot test larger blocks than the maximum block size, obviously.
|
540
445
|
return;
|
541
446
|
}
|
542
447
|
|
543
|
-
string prefix;
|
448
|
+
std::string prefix;
|
544
449
|
Varint::Append32(&prefix, input.size());
|
545
450
|
|
546
451
|
// Setup compression table
|
547
|
-
snappy::internal::WorkingMemory wmem;
|
452
|
+
snappy::internal::WorkingMemory wmem(input.size());
|
548
453
|
int table_size;
|
549
454
|
uint16* table = wmem.GetHashTable(input.size(), &table_size);
|
550
455
|
|
551
456
|
// Compress entire input in one shot
|
552
|
-
string compressed;
|
457
|
+
std::string compressed;
|
553
458
|
compressed += prefix;
|
554
459
|
compressed.resize(prefix.size()+snappy::MaxCompressedLength(input.size()));
|
555
460
|
char* dest = string_as_array(&compressed) + prefix.size();
|
@@ -557,57 +462,79 @@ static void VerifyNonBlockedCompression(const string& input) {
|
|
557
462
|
dest, table, table_size);
|
558
463
|
compressed.resize(end - compressed.data());
|
559
464
|
|
560
|
-
// Uncompress into string
|
561
|
-
string uncomp_str;
|
465
|
+
// Uncompress into std::string
|
466
|
+
std::string uncomp_str;
|
562
467
|
CHECK(snappy::Uncompress(compressed.data(), compressed.size(), &uncomp_str));
|
563
468
|
CHECK_EQ(uncomp_str, input);
|
564
469
|
|
470
|
+
// Uncompress using source/sink
|
471
|
+
std::string uncomp_str2;
|
472
|
+
uncomp_str2.resize(input.size());
|
473
|
+
snappy::UncheckedByteArraySink sink(string_as_array(&uncomp_str2));
|
474
|
+
snappy::ByteArraySource source(compressed.data(), compressed.size());
|
475
|
+
CHECK(snappy::Uncompress(&source, &sink));
|
476
|
+
CHECK_EQ(uncomp_str2, input);
|
477
|
+
|
478
|
+
// Uncompress into iovec
|
479
|
+
{
|
480
|
+
static const int kNumBlocks = 10;
|
481
|
+
struct iovec vec[kNumBlocks];
|
482
|
+
const int block_size = 1 + input.size() / kNumBlocks;
|
483
|
+
std::string iovec_data(block_size * kNumBlocks, 'x');
|
484
|
+
for (int i = 0; i < kNumBlocks; i++) {
|
485
|
+
vec[i].iov_base = string_as_array(&iovec_data) + i * block_size;
|
486
|
+
vec[i].iov_len = block_size;
|
487
|
+
}
|
488
|
+
CHECK(snappy::RawUncompressToIOVec(compressed.data(), compressed.size(),
|
489
|
+
vec, kNumBlocks));
|
490
|
+
CHECK_EQ(std::string(iovec_data.data(), input.size()), input);
|
491
|
+
}
|
565
492
|
}
|
566
493
|
|
567
494
|
// Expand the input so that it is at least K times as big as block size
|
568
|
-
static string Expand(const string& input) {
|
495
|
+
static std::string Expand(const std::string& input) {
|
569
496
|
static const int K = 3;
|
570
|
-
string data = input;
|
497
|
+
std::string data = input;
|
571
498
|
while (data.size() < K * snappy::kBlockSize) {
|
572
499
|
data += input;
|
573
500
|
}
|
574
501
|
return data;
|
575
502
|
}
|
576
503
|
|
577
|
-
static int Verify(const string& input) {
|
504
|
+
static int Verify(const std::string& input) {
|
578
505
|
VLOG(1) << "Verifying input of size " << input.size();
|
579
506
|
|
580
507
|
// Compress using string based routines
|
581
508
|
const int result = VerifyString(input);
|
582
509
|
|
510
|
+
// Verify using sink based routines
|
511
|
+
VerifyStringSink(input);
|
583
512
|
|
584
513
|
VerifyNonBlockedCompression(input);
|
585
514
|
VerifyIOVec(input);
|
586
515
|
if (!input.empty()) {
|
587
|
-
const string expanded = Expand(input);
|
516
|
+
const std::string expanded = Expand(input);
|
588
517
|
VerifyNonBlockedCompression(expanded);
|
589
518
|
VerifyIOVec(input);
|
590
519
|
}
|
591
520
|
|
592
|
-
|
593
521
|
return result;
|
594
522
|
}
|
595
523
|
|
596
|
-
|
597
|
-
// corrupted data.
|
598
|
-
|
599
|
-
static bool IsValidCompressedBuffer(const string& c) {
|
524
|
+
static bool IsValidCompressedBuffer(const std::string& c) {
|
600
525
|
return snappy::IsValidCompressedBuffer(c.data(), c.size());
|
601
526
|
}
|
602
|
-
static bool Uncompress(const string& c, string* u) {
|
527
|
+
static bool Uncompress(const std::string& c, std::string* u) {
|
603
528
|
return snappy::Uncompress(c.data(), c.size(), u);
|
604
529
|
}
|
605
530
|
|
606
|
-
|
607
|
-
|
531
|
+
// This test checks to ensure that snappy doesn't coredump if it gets
|
532
|
+
// corrupted data.
|
533
|
+
TEST(CorruptedTest, VerifyCorrupted) {
|
534
|
+
std::string source = "making sure we don't crash with corrupted input";
|
608
535
|
VLOG(1) << source;
|
609
|
-
string dest;
|
610
|
-
|
536
|
+
std::string dest;
|
537
|
+
std::string uncmp;
|
611
538
|
snappy::Compress(source.data(), source.size(), &dest);
|
612
539
|
|
613
540
|
// Mess around with the data. It's hard to simulate all possible
|
@@ -616,19 +543,19 @@ TYPED_TEST(CorruptedTest, VerifyCorrupted) {
|
|
616
543
|
dest[1]--;
|
617
544
|
dest[3]++;
|
618
545
|
// this really ought to fail.
|
619
|
-
CHECK(!IsValidCompressedBuffer(
|
620
|
-
CHECK(!Uncompress(
|
546
|
+
CHECK(!IsValidCompressedBuffer(dest));
|
547
|
+
CHECK(!Uncompress(dest, &uncmp));
|
621
548
|
|
622
549
|
// This is testing for a security bug - a buffer that decompresses to 100k
|
623
550
|
// but we lie in the snappy header and only reserve 0 bytes of memory :)
|
624
551
|
source.resize(100000);
|
625
|
-
for (
|
552
|
+
for (size_t i = 0; i < source.length(); ++i) {
|
626
553
|
source[i] = 'A';
|
627
554
|
}
|
628
555
|
snappy::Compress(source.data(), source.size(), &dest);
|
629
556
|
dest[0] = dest[1] = dest[2] = dest[3] = 0;
|
630
|
-
CHECK(!IsValidCompressedBuffer(
|
631
|
-
CHECK(!Uncompress(
|
557
|
+
CHECK(!IsValidCompressedBuffer(dest));
|
558
|
+
CHECK(!Uncompress(dest, &uncmp));
|
632
559
|
|
633
560
|
if (sizeof(void *) == 4) {
|
634
561
|
// Another security check; check a crazy big length can't DoS us with an
|
@@ -637,26 +564,26 @@ TYPED_TEST(CorruptedTest, VerifyCorrupted) {
|
|
637
564
|
// where 3 GB might be an acceptable allocation size, Uncompress()
|
638
565
|
// attempts to decompress, and sometimes causes the test to run out of
|
639
566
|
// memory.
|
640
|
-
dest[0] = dest[1] = dest[2] = dest[3] =
|
567
|
+
dest[0] = dest[1] = dest[2] = dest[3] = '\xff';
|
641
568
|
// This decodes to a really large size, i.e., about 3 GB.
|
642
569
|
dest[4] = 'k';
|
643
|
-
CHECK(!IsValidCompressedBuffer(
|
644
|
-
CHECK(!Uncompress(
|
570
|
+
CHECK(!IsValidCompressedBuffer(dest));
|
571
|
+
CHECK(!Uncompress(dest, &uncmp));
|
645
572
|
} else {
|
646
573
|
LOG(WARNING) << "Crazy decompression lengths not checked on 64-bit build";
|
647
574
|
}
|
648
575
|
|
649
576
|
// This decodes to about 2 MB; much smaller, but should still fail.
|
650
|
-
dest[0] = dest[1] = dest[2] =
|
577
|
+
dest[0] = dest[1] = dest[2] = '\xff';
|
651
578
|
dest[3] = 0x00;
|
652
|
-
CHECK(!IsValidCompressedBuffer(
|
653
|
-
CHECK(!Uncompress(
|
579
|
+
CHECK(!IsValidCompressedBuffer(dest));
|
580
|
+
CHECK(!Uncompress(dest, &uncmp));
|
654
581
|
|
655
582
|
// try reading stuff in from a bad file.
|
656
583
|
for (int i = 1; i <= 3; ++i) {
|
657
|
-
string data =
|
658
|
-
|
659
|
-
string uncmp;
|
584
|
+
std::string data =
|
585
|
+
ReadTestDataFile(StrFormat("baddata%d.snappy", i).c_str(), 0);
|
586
|
+
std::string uncmp;
|
660
587
|
// check that we don't return a crazy length
|
661
588
|
size_t ulen;
|
662
589
|
CHECK(!snappy::GetUncompressedLength(data.data(), data.size(), &ulen)
|
@@ -665,8 +592,8 @@ TYPED_TEST(CorruptedTest, VerifyCorrupted) {
|
|
665
592
|
snappy::ByteArraySource source(data.data(), data.size());
|
666
593
|
CHECK(!snappy::GetUncompressedLength(&source, &ulen2) ||
|
667
594
|
(ulen2 < (1<<20)));
|
668
|
-
CHECK(!IsValidCompressedBuffer(
|
669
|
-
CHECK(!Uncompress(
|
595
|
+
CHECK(!IsValidCompressedBuffer(data));
|
596
|
+
CHECK(!Uncompress(data, &uncmp));
|
670
597
|
}
|
671
598
|
}
|
672
599
|
|
@@ -674,7 +601,7 @@ TYPED_TEST(CorruptedTest, VerifyCorrupted) {
|
|
674
601
|
// These mirror the compression code in snappy.cc, but are copied
|
675
602
|
// here so that we can bypass some limitations in the how snappy.cc
|
676
603
|
// invokes these routines.
|
677
|
-
static void AppendLiteral(string* dst, const string& literal) {
|
604
|
+
static void AppendLiteral(std::string* dst, const std::string& literal) {
|
678
605
|
if (literal.empty()) return;
|
679
606
|
int n = literal.size() - 1;
|
680
607
|
if (n < 60) {
|
@@ -689,12 +616,12 @@ static void AppendLiteral(string* dst, const string& literal) {
|
|
689
616
|
n >>= 8;
|
690
617
|
}
|
691
618
|
dst->push_back(0 | ((59+count) << 2));
|
692
|
-
*dst += string(number, count);
|
619
|
+
*dst += std::string(number, count);
|
693
620
|
}
|
694
621
|
*dst += literal;
|
695
622
|
}
|
696
623
|
|
697
|
-
static void AppendCopy(string* dst, int offset, int length) {
|
624
|
+
static void AppendCopy(std::string* dst, int offset, int length) {
|
698
625
|
while (length > 0) {
|
699
626
|
// Figure out how much to copy in one shot
|
700
627
|
int to_copy;
|
@@ -731,51 +658,67 @@ TEST(Snappy, SimpleTests) {
|
|
731
658
|
Verify("ab");
|
732
659
|
Verify("abc");
|
733
660
|
|
734
|
-
Verify("aaaaaaa" + string(16, 'b') + string("aaaaa") + "abc");
|
735
|
-
Verify("aaaaaaa" + string(256, 'b') + string("aaaaa") + "abc");
|
736
|
-
Verify("aaaaaaa" + string(2047, 'b') + string("aaaaa") + "abc");
|
737
|
-
Verify("aaaaaaa" + string(65536, 'b') + string("aaaaa") + "abc");
|
738
|
-
Verify("abcaaaaaaa" + string(65536, 'b') + string("aaaaa") + "abc");
|
661
|
+
Verify("aaaaaaa" + std::string(16, 'b') + std::string("aaaaa") + "abc");
|
662
|
+
Verify("aaaaaaa" + std::string(256, 'b') + std::string("aaaaa") + "abc");
|
663
|
+
Verify("aaaaaaa" + std::string(2047, 'b') + std::string("aaaaa") + "abc");
|
664
|
+
Verify("aaaaaaa" + std::string(65536, 'b') + std::string("aaaaa") + "abc");
|
665
|
+
Verify("abcaaaaaaa" + std::string(65536, 'b') + std::string("aaaaa") + "abc");
|
739
666
|
}
|
740
667
|
|
741
668
|
// Verify max blowup (lots of four-byte copies)
|
742
669
|
TEST(Snappy, MaxBlowup) {
|
743
|
-
|
744
|
-
|
745
|
-
|
746
|
-
|
747
|
-
input.
|
748
|
-
|
749
|
-
for (int i =
|
750
|
-
|
751
|
-
|
752
|
-
input.append(reinterpret_cast<char*>(&bytes), sizeof(bytes));
|
670
|
+
std::mt19937 rng;
|
671
|
+
std::uniform_int_distribution<int> uniform_byte(0, 255);
|
672
|
+
std::string input;
|
673
|
+
for (int i = 0; i < 80000; ++i)
|
674
|
+
input.push_back(static_cast<char>(uniform_byte(rng)));
|
675
|
+
|
676
|
+
for (int i = 0; i < 80000; i += 4) {
|
677
|
+
std::string four_bytes(input.end() - i - 4, input.end() - i);
|
678
|
+
input.append(four_bytes);
|
753
679
|
}
|
754
680
|
Verify(input);
|
755
681
|
}
|
756
682
|
|
757
683
|
TEST(Snappy, RandomData) {
|
758
|
-
|
759
|
-
|
760
|
-
|
684
|
+
std::minstd_rand0 rng(FLAGS_test_random_seed);
|
685
|
+
std::uniform_int_distribution<int> uniform_0_to_3(0, 3);
|
686
|
+
std::uniform_int_distribution<int> uniform_0_to_8(0, 8);
|
687
|
+
std::uniform_int_distribution<int> uniform_byte(0, 255);
|
688
|
+
std::uniform_int_distribution<size_t> uniform_4k(0, 4095);
|
689
|
+
std::uniform_int_distribution<size_t> uniform_64k(0, 65535);
|
690
|
+
std::bernoulli_distribution one_in_ten(1.0 / 10);
|
691
|
+
|
692
|
+
constexpr int num_ops = 20000;
|
761
693
|
for (int i = 0; i < num_ops; i++) {
|
762
694
|
if ((i % 1000) == 0) {
|
763
695
|
VLOG(0) << "Random op " << i << " of " << num_ops;
|
764
696
|
}
|
765
697
|
|
766
|
-
string x;
|
767
|
-
|
698
|
+
std::string x;
|
699
|
+
size_t len = uniform_4k(rng);
|
768
700
|
if (i < 100) {
|
769
|
-
len = 65536 +
|
701
|
+
len = 65536 + uniform_64k(rng);
|
770
702
|
}
|
771
703
|
while (x.size() < len) {
|
772
704
|
int run_len = 1;
|
773
|
-
if (
|
774
|
-
|
705
|
+
if (one_in_ten(rng)) {
|
706
|
+
int skewed_bits = uniform_0_to_8(rng);
|
707
|
+
// int is guaranteed to hold at least 16 bits, this uses at most 8 bits.
|
708
|
+
std::uniform_int_distribution<int> skewed_low(0,
|
709
|
+
(1 << skewed_bits) - 1);
|
710
|
+
run_len = skewed_low(rng);
|
711
|
+
}
|
712
|
+
char c = static_cast<char>(uniform_byte(rng));
|
713
|
+
if (i >= 100) {
|
714
|
+
int skewed_bits = uniform_0_to_3(rng);
|
715
|
+
// int is guaranteed to hold at least 16 bits, this uses at most 3 bits.
|
716
|
+
std::uniform_int_distribution<int> skewed_low(0,
|
717
|
+
(1 << skewed_bits) - 1);
|
718
|
+
c = static_cast<char>(skewed_low(rng));
|
775
719
|
}
|
776
|
-
char c = (i < 100) ? rnd.Uniform(256) : rnd.Skewed(3);
|
777
720
|
while (run_len-- > 0 && x.size() < len) {
|
778
|
-
x
|
721
|
+
x.push_back(c);
|
779
722
|
}
|
780
723
|
}
|
781
724
|
|
@@ -789,19 +732,19 @@ TEST(Snappy, FourByteOffset) {
|
|
789
732
|
// copy manually.
|
790
733
|
|
791
734
|
// The two fragments that make up the input string.
|
792
|
-
string fragment1 = "012345689abcdefghijklmnopqrstuvwxyz";
|
793
|
-
string fragment2 = "some other string";
|
735
|
+
std::string fragment1 = "012345689abcdefghijklmnopqrstuvwxyz";
|
736
|
+
std::string fragment2 = "some other string";
|
794
737
|
|
795
738
|
// How many times each fragment is emitted.
|
796
739
|
const int n1 = 2;
|
797
740
|
const int n2 = 100000 / fragment2.size();
|
798
741
|
const int length = n1 * fragment1.size() + n2 * fragment2.size();
|
799
742
|
|
800
|
-
string compressed;
|
743
|
+
std::string compressed;
|
801
744
|
Varint::Append32(&compressed, length);
|
802
745
|
|
803
746
|
AppendLiteral(&compressed, fragment1);
|
804
|
-
string src = fragment1;
|
747
|
+
std::string src = fragment1;
|
805
748
|
for (int i = 0; i < n2; i++) {
|
806
749
|
AppendLiteral(&compressed, fragment2);
|
807
750
|
src += fragment2;
|
@@ -810,7 +753,7 @@ TEST(Snappy, FourByteOffset) {
|
|
810
753
|
src += fragment1;
|
811
754
|
CHECK_EQ(length, src.size());
|
812
755
|
|
813
|
-
string uncompressed;
|
756
|
+
std::string uncompressed;
|
814
757
|
CHECK(snappy::IsValidCompressedBuffer(compressed.data(), compressed.size()));
|
815
758
|
CHECK(snappy::Uncompress(compressed.data(), compressed.size(),
|
816
759
|
&uncompressed));
|
@@ -832,7 +775,7 @@ TEST(Snappy, IOVecEdgeCases) {
|
|
832
775
|
iov[i].iov_len = kLengths[i];
|
833
776
|
}
|
834
777
|
|
835
|
-
string compressed;
|
778
|
+
std::string compressed;
|
836
779
|
Varint::Append32(&compressed, 22);
|
837
780
|
|
838
781
|
// A literal whose output crosses three blocks.
|
@@ -893,7 +836,7 @@ TEST(Snappy, IOVecLiteralOverflow) {
|
|
893
836
|
iov[i].iov_len = kLengths[i];
|
894
837
|
}
|
895
838
|
|
896
|
-
string compressed;
|
839
|
+
std::string compressed;
|
897
840
|
Varint::Append32(&compressed, 8);
|
898
841
|
|
899
842
|
AppendLiteral(&compressed, "12345678");
|
@@ -915,7 +858,7 @@ TEST(Snappy, IOVecCopyOverflow) {
|
|
915
858
|
iov[i].iov_len = kLengths[i];
|
916
859
|
}
|
917
860
|
|
918
|
-
string compressed;
|
861
|
+
std::string compressed;
|
919
862
|
Varint::Append32(&compressed, 8);
|
920
863
|
|
921
864
|
AppendLiteral(&compressed, "123");
|
@@ -929,8 +872,7 @@ TEST(Snappy, IOVecCopyOverflow) {
|
|
929
872
|
}
|
930
873
|
}
|
931
874
|
|
932
|
-
|
933
|
-
static bool CheckUncompressedLength(const string& compressed,
|
875
|
+
static bool CheckUncompressedLength(const std::string& compressed,
|
934
876
|
size_t* ulength) {
|
935
877
|
const bool result1 = snappy::GetUncompressedLength(compressed.data(),
|
936
878
|
compressed.size(),
|
@@ -944,7 +886,7 @@ static bool CheckUncompressedLength(const string& compressed,
|
|
944
886
|
}
|
945
887
|
|
946
888
|
TEST(SnappyCorruption, TruncatedVarint) {
|
947
|
-
string compressed, uncompressed;
|
889
|
+
std::string compressed, uncompressed;
|
948
890
|
size_t ulength;
|
949
891
|
compressed.push_back('\xf0');
|
950
892
|
CHECK(!CheckUncompressedLength(compressed, &ulength));
|
@@ -954,13 +896,13 @@ TEST(SnappyCorruption, TruncatedVarint) {
|
|
954
896
|
}
|
955
897
|
|
956
898
|
TEST(SnappyCorruption, UnterminatedVarint) {
|
957
|
-
string compressed, uncompressed;
|
899
|
+
std::string compressed, uncompressed;
|
958
900
|
size_t ulength;
|
959
|
-
compressed.push_back(
|
960
|
-
compressed.push_back(
|
961
|
-
compressed.push_back(
|
962
|
-
compressed.push_back(
|
963
|
-
compressed.push_back(
|
901
|
+
compressed.push_back('\x80');
|
902
|
+
compressed.push_back('\x80');
|
903
|
+
compressed.push_back('\x80');
|
904
|
+
compressed.push_back('\x80');
|
905
|
+
compressed.push_back('\x80');
|
964
906
|
compressed.push_back(10);
|
965
907
|
CHECK(!CheckUncompressedLength(compressed, &ulength));
|
966
908
|
CHECK(!snappy::IsValidCompressedBuffer(compressed.data(), compressed.size()));
|
@@ -968,18 +910,32 @@ TEST(SnappyCorruption, UnterminatedVarint) {
|
|
968
910
|
&uncompressed));
|
969
911
|
}
|
970
912
|
|
913
|
+
TEST(SnappyCorruption, OverflowingVarint) {
|
914
|
+
std::string compressed, uncompressed;
|
915
|
+
size_t ulength;
|
916
|
+
compressed.push_back('\xfb');
|
917
|
+
compressed.push_back('\xff');
|
918
|
+
compressed.push_back('\xff');
|
919
|
+
compressed.push_back('\xff');
|
920
|
+
compressed.push_back('\x7f');
|
921
|
+
CHECK(!CheckUncompressedLength(compressed, &ulength));
|
922
|
+
CHECK(!snappy::IsValidCompressedBuffer(compressed.data(), compressed.size()));
|
923
|
+
CHECK(!snappy::Uncompress(compressed.data(), compressed.size(),
|
924
|
+
&uncompressed));
|
925
|
+
}
|
926
|
+
|
971
927
|
TEST(Snappy, ReadPastEndOfBuffer) {
|
972
928
|
// Check that we do not read past end of input
|
973
929
|
|
974
930
|
// Make a compressed string that ends with a single-byte literal
|
975
|
-
string compressed;
|
931
|
+
std::string compressed;
|
976
932
|
Varint::Append32(&compressed, 1);
|
977
933
|
AppendLiteral(&compressed, "x");
|
978
934
|
|
979
|
-
string uncompressed;
|
935
|
+
std::string uncompressed;
|
980
936
|
DataEndingAtUnreadablePage c(compressed);
|
981
937
|
CHECK(snappy::Uncompress(c.data(), c.size(), &uncompressed));
|
982
|
-
CHECK_EQ(uncompressed, string("x"));
|
938
|
+
CHECK_EQ(uncompressed, std::string("x"));
|
983
939
|
}
|
984
940
|
|
985
941
|
// Check for an infinite loop caused by a copy with offset==0
|
@@ -998,11 +954,13 @@ TEST(Snappy, ZeroOffsetCopyValidation) {
|
|
998
954
|
EXPECT_FALSE(snappy::IsValidCompressedBuffer(compressed, 4));
|
999
955
|
}
|
1000
956
|
|
1001
|
-
|
1002
957
|
namespace {
|
1003
958
|
|
1004
959
|
int TestFindMatchLength(const char* s1, const char *s2, unsigned length) {
|
1005
|
-
|
960
|
+
std::pair<size_t, bool> p =
|
961
|
+
snappy::internal::FindMatchLength(s1, s2, s2 + length);
|
962
|
+
CHECK_EQ(p.first < 8, p.second);
|
963
|
+
return p.first;
|
1006
964
|
}
|
1007
965
|
|
1008
966
|
} // namespace
|
@@ -1098,22 +1056,24 @@ TEST(Snappy, FindMatchLength) {
|
|
1098
1056
|
}
|
1099
1057
|
|
1100
1058
|
TEST(Snappy, FindMatchLengthRandom) {
|
1101
|
-
|
1102
|
-
|
1103
|
-
|
1059
|
+
constexpr int kNumTrials = 10000;
|
1060
|
+
constexpr int kTypicalLength = 10;
|
1061
|
+
std::minstd_rand0 rng(FLAGS_test_random_seed);
|
1062
|
+
std::uniform_int_distribution<int> uniform_byte(0, 255);
|
1063
|
+
std::bernoulli_distribution one_in_two(1.0 / 2);
|
1064
|
+
std::bernoulli_distribution one_in_typical_length(1.0 / kTypicalLength);
|
1104
1065
|
|
1105
1066
|
for (int i = 0; i < kNumTrials; i++) {
|
1106
|
-
string s, t;
|
1107
|
-
char a =
|
1108
|
-
char b =
|
1109
|
-
while (!
|
1110
|
-
s.push_back(
|
1111
|
-
t.push_back(
|
1067
|
+
std::string s, t;
|
1068
|
+
char a = static_cast<char>(uniform_byte(rng));
|
1069
|
+
char b = static_cast<char>(uniform_byte(rng));
|
1070
|
+
while (!one_in_typical_length(rng)) {
|
1071
|
+
s.push_back(one_in_two(rng) ? a : b);
|
1072
|
+
t.push_back(one_in_two(rng) ? a : b);
|
1112
1073
|
}
|
1113
1074
|
DataEndingAtUnreadablePage u(s);
|
1114
1075
|
DataEndingAtUnreadablePage v(t);
|
1115
|
-
int matched =
|
1116
|
-
u.data(), v.data(), v.data() + t.size());
|
1076
|
+
int matched = TestFindMatchLength(u.data(), v.data(), t.size());
|
1117
1077
|
if (matched == t.size()) {
|
1118
1078
|
EXPECT_EQ(s, t);
|
1119
1079
|
} else {
|
@@ -1125,51 +1085,140 @@ TEST(Snappy, FindMatchLengthRandom) {
|
|
1125
1085
|
}
|
1126
1086
|
}
|
1127
1087
|
|
1088
|
+
static uint16 MakeEntry(unsigned int extra,
|
1089
|
+
unsigned int len,
|
1090
|
+
unsigned int copy_offset) {
|
1091
|
+
// Check that all of the fields fit within the allocated space
|
1092
|
+
assert(extra == (extra & 0x7)); // At most 3 bits
|
1093
|
+
assert(copy_offset == (copy_offset & 0x7)); // At most 3 bits
|
1094
|
+
assert(len == (len & 0x7f)); // At most 7 bits
|
1095
|
+
return len | (copy_offset << 8) | (extra << 11);
|
1096
|
+
}
|
1097
|
+
|
1098
|
+
// Check that the decompression table is correct, and optionally print out
|
1099
|
+
// the computed one.
|
1100
|
+
TEST(Snappy, VerifyCharTable) {
|
1101
|
+
using snappy::internal::LITERAL;
|
1102
|
+
using snappy::internal::COPY_1_BYTE_OFFSET;
|
1103
|
+
using snappy::internal::COPY_2_BYTE_OFFSET;
|
1104
|
+
using snappy::internal::COPY_4_BYTE_OFFSET;
|
1105
|
+
using snappy::internal::char_table;
|
1106
|
+
|
1107
|
+
uint16 dst[256];
|
1108
|
+
|
1109
|
+
// Place invalid entries in all places to detect missing initialization
|
1110
|
+
int assigned = 0;
|
1111
|
+
for (int i = 0; i < 256; i++) {
|
1112
|
+
dst[i] = 0xffff;
|
1113
|
+
}
|
1114
|
+
|
1115
|
+
// Small LITERAL entries. We store (len-1) in the top 6 bits.
|
1116
|
+
for (unsigned int len = 1; len <= 60; len++) {
|
1117
|
+
dst[LITERAL | ((len-1) << 2)] = MakeEntry(0, len, 0);
|
1118
|
+
assigned++;
|
1119
|
+
}
|
1120
|
+
|
1121
|
+
// Large LITERAL entries. We use 60..63 in the high 6 bits to
|
1122
|
+
// encode the number of bytes of length info that follow the opcode.
|
1123
|
+
for (unsigned int extra_bytes = 1; extra_bytes <= 4; extra_bytes++) {
|
1124
|
+
// We set the length field in the lookup table to 1 because extra
|
1125
|
+
// bytes encode len-1.
|
1126
|
+
dst[LITERAL | ((extra_bytes+59) << 2)] = MakeEntry(extra_bytes, 1, 0);
|
1127
|
+
assigned++;
|
1128
|
+
}
|
1129
|
+
|
1130
|
+
// COPY_1_BYTE_OFFSET.
|
1131
|
+
//
|
1132
|
+
// The tag byte in the compressed data stores len-4 in 3 bits, and
|
1133
|
+
// offset/256 in 5 bits. offset%256 is stored in the next byte.
|
1134
|
+
//
|
1135
|
+
// This format is used for length in range [4..11] and offset in
|
1136
|
+
// range [0..2047]
|
1137
|
+
for (unsigned int len = 4; len < 12; len++) {
|
1138
|
+
for (unsigned int offset = 0; offset < 2048; offset += 256) {
|
1139
|
+
dst[COPY_1_BYTE_OFFSET | ((len-4)<<2) | ((offset>>8)<<5)] =
|
1140
|
+
MakeEntry(1, len, offset>>8);
|
1141
|
+
assigned++;
|
1142
|
+
}
|
1143
|
+
}
|
1144
|
+
|
1145
|
+
// COPY_2_BYTE_OFFSET.
|
1146
|
+
// Tag contains len-1 in top 6 bits, and offset in next two bytes.
|
1147
|
+
for (unsigned int len = 1; len <= 64; len++) {
|
1148
|
+
dst[COPY_2_BYTE_OFFSET | ((len-1)<<2)] = MakeEntry(2, len, 0);
|
1149
|
+
assigned++;
|
1150
|
+
}
|
1151
|
+
|
1152
|
+
// COPY_4_BYTE_OFFSET.
|
1153
|
+
// Tag contents len-1 in top 6 bits, and offset in next four bytes.
|
1154
|
+
for (unsigned int len = 1; len <= 64; len++) {
|
1155
|
+
dst[COPY_4_BYTE_OFFSET | ((len-1)<<2)] = MakeEntry(4, len, 0);
|
1156
|
+
assigned++;
|
1157
|
+
}
|
1158
|
+
|
1159
|
+
// Check that each entry was initialized exactly once.
|
1160
|
+
EXPECT_EQ(256, assigned) << "Assigned only " << assigned << " of 256";
|
1161
|
+
for (int i = 0; i < 256; i++) {
|
1162
|
+
EXPECT_NE(0xffff, dst[i]) << "Did not assign byte " << i;
|
1163
|
+
}
|
1164
|
+
|
1165
|
+
if (FLAGS_snappy_dump_decompression_table) {
|
1166
|
+
printf("static const uint16 char_table[256] = {\n ");
|
1167
|
+
for (int i = 0; i < 256; i++) {
|
1168
|
+
printf("0x%04x%s",
|
1169
|
+
dst[i],
|
1170
|
+
((i == 255) ? "\n" : (((i%8) == 7) ? ",\n " : ", ")));
|
1171
|
+
}
|
1172
|
+
printf("};\n");
|
1173
|
+
}
|
1174
|
+
|
1175
|
+
// Check that computed table matched recorded table.
|
1176
|
+
for (int i = 0; i < 256; i++) {
|
1177
|
+
EXPECT_EQ(dst[i], char_table[i]) << "Mismatch in byte " << i;
|
1178
|
+
}
|
1179
|
+
}
|
1128
1180
|
|
1129
1181
|
static void CompressFile(const char* fname) {
|
1130
|
-
string fullinput;
|
1131
|
-
file::GetContents(fname, &fullinput, file::Defaults())
|
1182
|
+
std::string fullinput;
|
1183
|
+
CHECK_OK(file::GetContents(fname, &fullinput, file::Defaults()));
|
1132
1184
|
|
1133
|
-
string compressed;
|
1185
|
+
std::string compressed;
|
1134
1186
|
Compress(fullinput.data(), fullinput.size(), SNAPPY, &compressed, false);
|
1135
1187
|
|
1136
|
-
file::SetContents(string(fname).append(".comp"), compressed,
|
1137
|
-
|
1188
|
+
CHECK_OK(file::SetContents(std::string(fname).append(".comp"), compressed,
|
1189
|
+
file::Defaults()));
|
1138
1190
|
}
|
1139
1191
|
|
1140
1192
|
static void UncompressFile(const char* fname) {
|
1141
|
-
string fullinput;
|
1142
|
-
file::GetContents(fname, &fullinput, file::Defaults())
|
1193
|
+
std::string fullinput;
|
1194
|
+
CHECK_OK(file::GetContents(fname, &fullinput, file::Defaults()));
|
1143
1195
|
|
1144
1196
|
size_t uncompLength;
|
1145
1197
|
CHECK(CheckUncompressedLength(fullinput, &uncompLength));
|
1146
1198
|
|
1147
|
-
string uncompressed;
|
1199
|
+
std::string uncompressed;
|
1148
1200
|
uncompressed.resize(uncompLength);
|
1149
1201
|
CHECK(snappy::Uncompress(fullinput.data(), fullinput.size(), &uncompressed));
|
1150
1202
|
|
1151
|
-
file::SetContents(string(fname).append(".uncomp"), uncompressed,
|
1152
|
-
|
1203
|
+
CHECK_OK(file::SetContents(std::string(fname).append(".uncomp"), uncompressed,
|
1204
|
+
file::Defaults()));
|
1153
1205
|
}
|
1154
1206
|
|
1155
1207
|
static void MeasureFile(const char* fname) {
|
1156
|
-
string fullinput;
|
1157
|
-
file::GetContents(fname, &fullinput, file::Defaults())
|
1208
|
+
std::string fullinput;
|
1209
|
+
CHECK_OK(file::GetContents(fname, &fullinput, file::Defaults()));
|
1158
1210
|
printf("%-40s :\n", fname);
|
1159
1211
|
|
1160
1212
|
int start_len = (FLAGS_start_len < 0) ? fullinput.size() : FLAGS_start_len;
|
1161
1213
|
int end_len = fullinput.size();
|
1162
1214
|
if (FLAGS_end_len >= 0) {
|
1163
|
-
end_len = min<int>(fullinput.size(), FLAGS_end_len);
|
1215
|
+
end_len = std::min<int>(fullinput.size(), FLAGS_end_len);
|
1164
1216
|
}
|
1165
1217
|
for (int len = start_len; len <= end_len; len++) {
|
1166
1218
|
const char* const input = fullinput.data();
|
1167
1219
|
int repeats = (FLAGS_bytes + len) / (len + 1);
|
1168
1220
|
if (FLAGS_zlib) Measure(input, len, ZLIB, repeats, 1024<<10);
|
1169
1221
|
if (FLAGS_lzo) Measure(input, len, LZO, repeats, 1024<<10);
|
1170
|
-
if (FLAGS_liblzf) Measure(input, len, LIBLZF, repeats, 1024<<10);
|
1171
|
-
if (FLAGS_quicklz) Measure(input, len, QUICKLZ, repeats, 1024<<10);
|
1172
|
-
if (FLAGS_fastlz) Measure(input, len, FASTLZ, repeats, 1024<<10);
|
1173
1222
|
if (FLAGS_snappy) Measure(input, len, SNAPPY, repeats, 4096<<10);
|
1174
1223
|
|
1175
1224
|
// For block-size based measurements
|
@@ -1209,10 +1258,10 @@ static void BM_UFlat(int iters, int arg) {
|
|
1209
1258
|
// Pick file to process based on "arg"
|
1210
1259
|
CHECK_GE(arg, 0);
|
1211
1260
|
CHECK_LT(arg, ARRAYSIZE(files));
|
1212
|
-
string contents =
|
1213
|
-
|
1261
|
+
std::string contents =
|
1262
|
+
ReadTestDataFile(files[arg].filename, files[arg].size_limit);
|
1214
1263
|
|
1215
|
-
string zcontents;
|
1264
|
+
std::string zcontents;
|
1216
1265
|
snappy::Compress(contents.data(), contents.size(), &zcontents);
|
1217
1266
|
char* dst = new char[contents.size()];
|
1218
1267
|
|
@@ -1235,10 +1284,10 @@ static void BM_UValidate(int iters, int arg) {
|
|
1235
1284
|
// Pick file to process based on "arg"
|
1236
1285
|
CHECK_GE(arg, 0);
|
1237
1286
|
CHECK_LT(arg, ARRAYSIZE(files));
|
1238
|
-
string contents =
|
1239
|
-
|
1287
|
+
std::string contents =
|
1288
|
+
ReadTestDataFile(files[arg].filename, files[arg].size_limit);
|
1240
1289
|
|
1241
|
-
string zcontents;
|
1290
|
+
std::string zcontents;
|
1242
1291
|
snappy::Compress(contents.data(), contents.size(), &zcontents);
|
1243
1292
|
|
1244
1293
|
SetBenchmarkBytesProcessed(static_cast<int64>(iters) *
|
@@ -1258,10 +1307,10 @@ static void BM_UIOVec(int iters, int arg) {
|
|
1258
1307
|
// Pick file to process based on "arg"
|
1259
1308
|
CHECK_GE(arg, 0);
|
1260
1309
|
CHECK_LT(arg, ARRAYSIZE(files));
|
1261
|
-
string contents =
|
1262
|
-
|
1310
|
+
std::string contents =
|
1311
|
+
ReadTestDataFile(files[arg].filename, files[arg].size_limit);
|
1263
1312
|
|
1264
|
-
string zcontents;
|
1313
|
+
std::string zcontents;
|
1265
1314
|
snappy::Compress(contents.data(), contents.size(), &zcontents);
|
1266
1315
|
|
1267
1316
|
// Uncompress into an iovec containing ten entries.
|
@@ -1298,6 +1347,37 @@ static void BM_UIOVec(int iters, int arg) {
|
|
1298
1347
|
}
|
1299
1348
|
BENCHMARK(BM_UIOVec)->DenseRange(0, 4);
|
1300
1349
|
|
1350
|
+
static void BM_UFlatSink(int iters, int arg) {
|
1351
|
+
StopBenchmarkTiming();
|
1352
|
+
|
1353
|
+
// Pick file to process based on "arg"
|
1354
|
+
CHECK_GE(arg, 0);
|
1355
|
+
CHECK_LT(arg, ARRAYSIZE(files));
|
1356
|
+
std::string contents =
|
1357
|
+
ReadTestDataFile(files[arg].filename, files[arg].size_limit);
|
1358
|
+
|
1359
|
+
std::string zcontents;
|
1360
|
+
snappy::Compress(contents.data(), contents.size(), &zcontents);
|
1361
|
+
char* dst = new char[contents.size()];
|
1362
|
+
|
1363
|
+
SetBenchmarkBytesProcessed(static_cast<int64>(iters) *
|
1364
|
+
static_cast<int64>(contents.size()));
|
1365
|
+
SetBenchmarkLabel(files[arg].label);
|
1366
|
+
StartBenchmarkTiming();
|
1367
|
+
while (iters-- > 0) {
|
1368
|
+
snappy::ByteArraySource source(zcontents.data(), zcontents.size());
|
1369
|
+
snappy::UncheckedByteArraySink sink(dst);
|
1370
|
+
CHECK(snappy::Uncompress(&source, &sink));
|
1371
|
+
}
|
1372
|
+
StopBenchmarkTiming();
|
1373
|
+
|
1374
|
+
std::string s(dst, contents.size());
|
1375
|
+
CHECK_EQ(contents, s);
|
1376
|
+
|
1377
|
+
delete[] dst;
|
1378
|
+
}
|
1379
|
+
|
1380
|
+
BENCHMARK(BM_UFlatSink)->DenseRange(0, ARRAYSIZE(files) - 1);
|
1301
1381
|
|
1302
1382
|
static void BM_ZFlat(int iters, int arg) {
|
1303
1383
|
StopBenchmarkTiming();
|
@@ -1305,8 +1385,8 @@ static void BM_ZFlat(int iters, int arg) {
|
|
1305
1385
|
// Pick file to process based on "arg"
|
1306
1386
|
CHECK_GE(arg, 0);
|
1307
1387
|
CHECK_LT(arg, ARRAYSIZE(files));
|
1308
|
-
string contents =
|
1309
|
-
|
1388
|
+
std::string contents =
|
1389
|
+
ReadTestDataFile(files[arg].filename, files[arg].size_limit);
|
1310
1390
|
|
1311
1391
|
char* dst = new char[snappy::MaxCompressedLength(contents.size())];
|
1312
1392
|
|
@@ -1321,31 +1401,102 @@ static void BM_ZFlat(int iters, int arg) {
|
|
1321
1401
|
StopBenchmarkTiming();
|
1322
1402
|
const double compression_ratio =
|
1323
1403
|
static_cast<double>(zsize) / std::max<size_t>(1, contents.size());
|
1324
|
-
SetBenchmarkLabel(
|
1325
|
-
|
1326
|
-
VLOG(0) <<
|
1327
|
-
|
1404
|
+
SetBenchmarkLabel(StrFormat("%s (%.2f %%)", files[arg].label,
|
1405
|
+
100.0 * compression_ratio));
|
1406
|
+
VLOG(0) << StrFormat("compression for %s: %zd -> %zd bytes",
|
1407
|
+
files[arg].label, static_cast<int>(contents.size()),
|
1408
|
+
static_cast<int>(zsize));
|
1328
1409
|
delete[] dst;
|
1329
1410
|
}
|
1330
1411
|
BENCHMARK(BM_ZFlat)->DenseRange(0, ARRAYSIZE(files) - 1);
|
1331
1412
|
|
1413
|
+
static void BM_ZFlatAll(int iters, int arg) {
|
1414
|
+
StopBenchmarkTiming();
|
1332
1415
|
|
1333
|
-
|
1416
|
+
CHECK_EQ(arg, 0);
|
1417
|
+
const int num_files = ARRAYSIZE(files);
|
1334
1418
|
|
1419
|
+
std::vector<std::string> contents(num_files);
|
1420
|
+
std::vector<char*> dst(num_files);
|
1421
|
+
|
1422
|
+
int64 total_contents_size = 0;
|
1423
|
+
for (int i = 0; i < num_files; ++i) {
|
1424
|
+
contents[i] = ReadTestDataFile(files[i].filename, files[i].size_limit);
|
1425
|
+
dst[i] = new char[snappy::MaxCompressedLength(contents[i].size())];
|
1426
|
+
total_contents_size += contents[i].size();
|
1427
|
+
}
|
1428
|
+
|
1429
|
+
SetBenchmarkBytesProcessed(static_cast<int64>(iters) * total_contents_size);
|
1430
|
+
StartBenchmarkTiming();
|
1431
|
+
|
1432
|
+
size_t zsize = 0;
|
1433
|
+
while (iters-- > 0) {
|
1434
|
+
for (int i = 0; i < num_files; ++i) {
|
1435
|
+
snappy::RawCompress(contents[i].data(), contents[i].size(), dst[i],
|
1436
|
+
&zsize);
|
1437
|
+
}
|
1438
|
+
}
|
1439
|
+
StopBenchmarkTiming();
|
1440
|
+
|
1441
|
+
for (int i = 0; i < num_files; ++i) {
|
1442
|
+
delete[] dst[i];
|
1443
|
+
}
|
1444
|
+
SetBenchmarkLabel(StrFormat("%d files", num_files));
|
1445
|
+
}
|
1446
|
+
BENCHMARK(BM_ZFlatAll)->DenseRange(0, 0);
|
1447
|
+
|
1448
|
+
static void BM_ZFlatIncreasingTableSize(int iters, int arg) {
|
1449
|
+
StopBenchmarkTiming();
|
1450
|
+
|
1451
|
+
CHECK_EQ(arg, 0);
|
1452
|
+
CHECK_GT(ARRAYSIZE(files), 0);
|
1453
|
+
const std::string base_content =
|
1454
|
+
ReadTestDataFile(files[0].filename, files[0].size_limit);
|
1455
|
+
|
1456
|
+
std::vector<std::string> contents;
|
1457
|
+
std::vector<char*> dst;
|
1458
|
+
int64 total_contents_size = 0;
|
1459
|
+
for (int table_bits = kMinHashTableBits; table_bits <= kMaxHashTableBits;
|
1460
|
+
++table_bits) {
|
1461
|
+
std::string content = base_content;
|
1462
|
+
content.resize(1 << table_bits);
|
1463
|
+
dst.push_back(new char[snappy::MaxCompressedLength(content.size())]);
|
1464
|
+
total_contents_size += content.size();
|
1465
|
+
contents.push_back(std::move(content));
|
1466
|
+
}
|
1467
|
+
|
1468
|
+
size_t zsize = 0;
|
1469
|
+
SetBenchmarkBytesProcessed(static_cast<int64>(iters) * total_contents_size);
|
1470
|
+
StartBenchmarkTiming();
|
1471
|
+
while (iters-- > 0) {
|
1472
|
+
for (int i = 0; i < contents.size(); ++i) {
|
1473
|
+
snappy::RawCompress(contents[i].data(), contents[i].size(), dst[i],
|
1474
|
+
&zsize);
|
1475
|
+
}
|
1476
|
+
}
|
1477
|
+
StopBenchmarkTiming();
|
1478
|
+
|
1479
|
+
for (int i = 0; i < dst.size(); ++i) {
|
1480
|
+
delete[] dst[i];
|
1481
|
+
}
|
1482
|
+
SetBenchmarkLabel(StrFormat("%zd tables", contents.size()));
|
1483
|
+
}
|
1484
|
+
BENCHMARK(BM_ZFlatIncreasingTableSize)->DenseRange(0, 0);
|
1485
|
+
|
1486
|
+
} // namespace snappy
|
1335
1487
|
|
1336
1488
|
int main(int argc, char** argv) {
|
1337
1489
|
InitGoogle(argv[0], &argc, &argv, true);
|
1338
1490
|
RunSpecifiedBenchmarks();
|
1339
1491
|
|
1340
|
-
|
1341
1492
|
if (argc >= 2) {
|
1342
1493
|
for (int arg = 1; arg < argc; arg++) {
|
1343
1494
|
if (FLAGS_write_compressed) {
|
1344
|
-
CompressFile(argv[arg]);
|
1495
|
+
snappy::CompressFile(argv[arg]);
|
1345
1496
|
} else if (FLAGS_write_uncompressed) {
|
1346
|
-
UncompressFile(argv[arg]);
|
1497
|
+
snappy::UncompressFile(argv[arg]);
|
1347
1498
|
} else {
|
1348
|
-
MeasureFile(argv[arg]);
|
1499
|
+
snappy::MeasureFile(argv[arg]);
|
1349
1500
|
}
|
1350
1501
|
}
|
1351
1502
|
return 0;
|