snappy 0.0.13 → 0.1.0

Sign up to get free protection for your applications and to get access to all the features.
Files changed (43) hide show
  1. checksums.yaml +5 -5
  2. data/.travis.yml +28 -1
  3. data/Gemfile +6 -1
  4. data/README.md +28 -4
  5. data/Rakefile +1 -0
  6. data/ext/extconf.rb +21 -24
  7. data/lib/snappy.rb +3 -1
  8. data/lib/snappy/hadoop.rb +22 -0
  9. data/lib/snappy/hadoop/reader.rb +58 -0
  10. data/lib/snappy/hadoop/writer.rb +51 -0
  11. data/lib/snappy/reader.rb +11 -7
  12. data/lib/snappy/shim.rb +30 -0
  13. data/lib/snappy/version.rb +3 -1
  14. data/lib/snappy/writer.rb +8 -9
  15. data/smoke.sh +8 -0
  16. data/snappy.gemspec +6 -30
  17. data/test/hadoop/test-snappy-hadoop-reader.rb +103 -0
  18. data/test/hadoop/test-snappy-hadoop-writer.rb +48 -0
  19. data/test/test-snappy-hadoop.rb +22 -0
  20. data/vendor/snappy/AUTHORS +1 -0
  21. data/vendor/snappy/CMakeLists.txt +174 -0
  22. data/vendor/snappy/CONTRIBUTING.md +26 -0
  23. data/vendor/snappy/COPYING +54 -0
  24. data/vendor/snappy/NEWS +180 -0
  25. data/vendor/snappy/README.md +149 -0
  26. data/vendor/snappy/cmake/SnappyConfig.cmake +1 -0
  27. data/vendor/snappy/cmake/config.h.in +62 -0
  28. data/vendor/snappy/format_description.txt +110 -0
  29. data/vendor/snappy/framing_format.txt +135 -0
  30. data/vendor/snappy/snappy-c.cc +90 -0
  31. data/vendor/snappy/snappy-c.h +138 -0
  32. data/vendor/snappy/snappy-internal.h +224 -0
  33. data/vendor/snappy/snappy-sinksource.cc +104 -0
  34. data/vendor/snappy/snappy-sinksource.h +182 -0
  35. data/vendor/snappy/snappy-stubs-internal.cc +42 -0
  36. data/vendor/snappy/snappy-stubs-internal.h +561 -0
  37. data/vendor/snappy/snappy-stubs-public.h.in +94 -0
  38. data/vendor/snappy/snappy-test.cc +612 -0
  39. data/vendor/snappy/snappy-test.h +573 -0
  40. data/vendor/snappy/snappy.cc +1515 -0
  41. data/vendor/snappy/snappy.h +203 -0
  42. data/vendor/snappy/snappy_unittest.cc +1410 -0
  43. metadata +38 -46
@@ -0,0 +1,104 @@
1
+ // Copyright 2011 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
+ #include <string.h>
30
+
31
+ #include "snappy-sinksource.h"
32
+
33
+ namespace snappy {
34
+
35
+ Source::~Source() { }
36
+
37
+ Sink::~Sink() { }
38
+
39
+ char* Sink::GetAppendBuffer(size_t length, char* scratch) {
40
+ return scratch;
41
+ }
42
+
43
+ char* Sink::GetAppendBufferVariable(
44
+ size_t min_size, size_t desired_size_hint, char* scratch,
45
+ size_t scratch_size, size_t* allocated_size) {
46
+ *allocated_size = scratch_size;
47
+ return scratch;
48
+ }
49
+
50
+ void Sink::AppendAndTakeOwnership(
51
+ char* bytes, size_t n,
52
+ void (*deleter)(void*, const char*, size_t),
53
+ void *deleter_arg) {
54
+ Append(bytes, n);
55
+ (*deleter)(deleter_arg, bytes, n);
56
+ }
57
+
58
+ ByteArraySource::~ByteArraySource() { }
59
+
60
+ size_t ByteArraySource::Available() const { return left_; }
61
+
62
+ const char* ByteArraySource::Peek(size_t* len) {
63
+ *len = left_;
64
+ return ptr_;
65
+ }
66
+
67
+ void ByteArraySource::Skip(size_t n) {
68
+ left_ -= n;
69
+ ptr_ += n;
70
+ }
71
+
72
+ UncheckedByteArraySink::~UncheckedByteArraySink() { }
73
+
74
+ void UncheckedByteArraySink::Append(const char* data, size_t n) {
75
+ // Do no copying if the caller filled in the result of GetAppendBuffer()
76
+ if (data != dest_) {
77
+ memcpy(dest_, data, n);
78
+ }
79
+ dest_ += n;
80
+ }
81
+
82
+ char* UncheckedByteArraySink::GetAppendBuffer(size_t len, char* scratch) {
83
+ return dest_;
84
+ }
85
+
86
+ void UncheckedByteArraySink::AppendAndTakeOwnership(
87
+ char* data, size_t n,
88
+ void (*deleter)(void*, const char*, size_t),
89
+ void *deleter_arg) {
90
+ if (data != dest_) {
91
+ memcpy(dest_, data, n);
92
+ (*deleter)(deleter_arg, data, n);
93
+ }
94
+ dest_ += n;
95
+ }
96
+
97
+ char* UncheckedByteArraySink::GetAppendBufferVariable(
98
+ size_t min_size, size_t desired_size_hint, char* scratch,
99
+ size_t scratch_size, size_t* allocated_size) {
100
+ *allocated_size = desired_size_hint;
101
+ return dest_;
102
+ }
103
+
104
+ } // namespace snappy
@@ -0,0 +1,182 @@
1
+ // Copyright 2011 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
+ #ifndef THIRD_PARTY_SNAPPY_SNAPPY_SINKSOURCE_H_
30
+ #define THIRD_PARTY_SNAPPY_SNAPPY_SINKSOURCE_H_
31
+
32
+ #include <stddef.h>
33
+
34
+ namespace snappy {
35
+
36
+ // A Sink is an interface that consumes a sequence of bytes.
37
+ class Sink {
38
+ public:
39
+ Sink() { }
40
+ virtual ~Sink();
41
+
42
+ // Append "bytes[0,n-1]" to this.
43
+ virtual void Append(const char* bytes, size_t n) = 0;
44
+
45
+ // Returns a writable buffer of the specified length for appending.
46
+ // May return a pointer to the caller-owned scratch buffer which
47
+ // must have at least the indicated length. The returned buffer is
48
+ // only valid until the next operation on this Sink.
49
+ //
50
+ // After writing at most "length" bytes, call Append() with the
51
+ // pointer returned from this function and the number of bytes
52
+ // written. Many Append() implementations will avoid copying
53
+ // bytes if this function returned an internal buffer.
54
+ //
55
+ // If a non-scratch buffer is returned, the caller may only pass a
56
+ // prefix of it to Append(). That is, it is not correct to pass an
57
+ // interior pointer of the returned array to Append().
58
+ //
59
+ // The default implementation always returns the scratch buffer.
60
+ virtual char* GetAppendBuffer(size_t length, char* scratch);
61
+
62
+ // For higher performance, Sink implementations can provide custom
63
+ // AppendAndTakeOwnership() and GetAppendBufferVariable() methods.
64
+ // These methods can reduce the number of copies done during
65
+ // compression/decompression.
66
+
67
+ // Append "bytes[0,n-1] to the sink. Takes ownership of "bytes"
68
+ // and calls the deleter function as (*deleter)(deleter_arg, bytes, n)
69
+ // to free the buffer. deleter function must be non NULL.
70
+ //
71
+ // The default implementation just calls Append and frees "bytes".
72
+ // Other implementations may avoid a copy while appending the buffer.
73
+ virtual void AppendAndTakeOwnership(
74
+ char* bytes, size_t n, void (*deleter)(void*, const char*, size_t),
75
+ void *deleter_arg);
76
+
77
+ // Returns a writable buffer for appending and writes the buffer's capacity to
78
+ // *allocated_size. Guarantees *allocated_size >= min_size.
79
+ // May return a pointer to the caller-owned scratch buffer which must have
80
+ // scratch_size >= min_size.
81
+ //
82
+ // The returned buffer is only valid until the next operation
83
+ // on this ByteSink.
84
+ //
85
+ // After writing at most *allocated_size bytes, call Append() with the
86
+ // pointer returned from this function and the number of bytes written.
87
+ // Many Append() implementations will avoid copying bytes if this function
88
+ // returned an internal buffer.
89
+ //
90
+ // If the sink implementation allocates or reallocates an internal buffer,
91
+ // it should use the desired_size_hint if appropriate. If a caller cannot
92
+ // provide a reasonable guess at the desired capacity, it should set
93
+ // desired_size_hint = 0.
94
+ //
95
+ // If a non-scratch buffer is returned, the caller may only pass
96
+ // a prefix to it to Append(). That is, it is not correct to pass an
97
+ // interior pointer to Append().
98
+ //
99
+ // The default implementation always returns the scratch buffer.
100
+ virtual char* GetAppendBufferVariable(
101
+ size_t min_size, size_t desired_size_hint, char* scratch,
102
+ size_t scratch_size, size_t* allocated_size);
103
+
104
+ private:
105
+ // No copying
106
+ Sink(const Sink&);
107
+ void operator=(const Sink&);
108
+ };
109
+
110
+ // A Source is an interface that yields a sequence of bytes
111
+ class Source {
112
+ public:
113
+ Source() { }
114
+ virtual ~Source();
115
+
116
+ // Return the number of bytes left to read from the source
117
+ virtual size_t Available() const = 0;
118
+
119
+ // Peek at the next flat region of the source. Does not reposition
120
+ // the source. The returned region is empty iff Available()==0.
121
+ //
122
+ // Returns a pointer to the beginning of the region and store its
123
+ // length in *len.
124
+ //
125
+ // The returned region is valid until the next call to Skip() or
126
+ // until this object is destroyed, whichever occurs first.
127
+ //
128
+ // The returned region may be larger than Available() (for example
129
+ // if this ByteSource is a view on a substring of a larger source).
130
+ // The caller is responsible for ensuring that it only reads the
131
+ // Available() bytes.
132
+ virtual const char* Peek(size_t* len) = 0;
133
+
134
+ // Skip the next n bytes. Invalidates any buffer returned by
135
+ // a previous call to Peek().
136
+ // REQUIRES: Available() >= n
137
+ virtual void Skip(size_t n) = 0;
138
+
139
+ private:
140
+ // No copying
141
+ Source(const Source&);
142
+ void operator=(const Source&);
143
+ };
144
+
145
+ // A Source implementation that yields the contents of a flat array
146
+ class ByteArraySource : public Source {
147
+ public:
148
+ ByteArraySource(const char* p, size_t n) : ptr_(p), left_(n) { }
149
+ virtual ~ByteArraySource();
150
+ virtual size_t Available() const;
151
+ virtual const char* Peek(size_t* len);
152
+ virtual void Skip(size_t n);
153
+ private:
154
+ const char* ptr_;
155
+ size_t left_;
156
+ };
157
+
158
+ // A Sink implementation that writes to a flat array without any bound checks.
159
+ class UncheckedByteArraySink : public Sink {
160
+ public:
161
+ explicit UncheckedByteArraySink(char* dest) : dest_(dest) { }
162
+ virtual ~UncheckedByteArraySink();
163
+ virtual void Append(const char* data, size_t n);
164
+ virtual char* GetAppendBuffer(size_t len, char* scratch);
165
+ virtual char* GetAppendBufferVariable(
166
+ size_t min_size, size_t desired_size_hint, char* scratch,
167
+ size_t scratch_size, size_t* allocated_size);
168
+ virtual void AppendAndTakeOwnership(
169
+ char* bytes, size_t n, void (*deleter)(void*, const char*, size_t),
170
+ void *deleter_arg);
171
+
172
+ // Return the current output pointer so that a caller can see how
173
+ // many bytes were produced.
174
+ // Note: this is not a Sink method.
175
+ char* CurrentDestination() const { return dest_; }
176
+ private:
177
+ char* dest_;
178
+ };
179
+
180
+ } // namespace snappy
181
+
182
+ #endif // THIRD_PARTY_SNAPPY_SNAPPY_SINKSOURCE_H_
@@ -0,0 +1,42 @@
1
+ // Copyright 2011 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
+ #include <algorithm>
30
+ #include <string>
31
+
32
+ #include "snappy-stubs-internal.h"
33
+
34
+ namespace snappy {
35
+
36
+ void Varint::Append32(string* s, uint32 value) {
37
+ char buf[Varint::kMax32];
38
+ const char* p = Varint::Encode32(buf, value);
39
+ s->append(buf, p - buf);
40
+ }
41
+
42
+ } // namespace snappy
@@ -0,0 +1,561 @@
1
+ // Copyright 2011 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
+ // Various stubs for the open-source version of Snappy.
30
+
31
+ #ifndef THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_INTERNAL_H_
32
+ #define THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_INTERNAL_H_
33
+
34
+ #ifdef HAVE_CONFIG_H
35
+ #include "config.h"
36
+ #endif
37
+
38
+ #include <string>
39
+
40
+ #include <assert.h>
41
+ #include <stdlib.h>
42
+ #include <string.h>
43
+
44
+ #ifdef HAVE_SYS_MMAN_H
45
+ #include <sys/mman.h>
46
+ #endif
47
+
48
+ #ifdef HAVE_UNISTD_H
49
+ #include <unistd.h>
50
+ #endif
51
+
52
+ #if defined(_MSC_VER)
53
+ #include <intrin.h>
54
+ #endif // defined(_MSC_VER)
55
+
56
+ #include "snappy-stubs-public.h"
57
+
58
+ #if defined(__x86_64__)
59
+
60
+ // Enable 64-bit optimized versions of some routines.
61
+ #define ARCH_K8 1
62
+
63
+ #elif defined(__ppc64__)
64
+
65
+ #define ARCH_PPC 1
66
+
67
+ #elif defined(__aarch64__)
68
+
69
+ #define ARCH_ARM 1
70
+
71
+ #endif
72
+
73
+ // Needed by OS X, among others.
74
+ #ifndef MAP_ANONYMOUS
75
+ #define MAP_ANONYMOUS MAP_ANON
76
+ #endif
77
+
78
+ // The size of an array, if known at compile-time.
79
+ // Will give unexpected results if used on a pointer.
80
+ // We undefine it first, since some compilers already have a definition.
81
+ #ifdef ARRAYSIZE
82
+ #undef ARRAYSIZE
83
+ #endif
84
+ #define ARRAYSIZE(a) (sizeof(a) / sizeof(*(a)))
85
+
86
+ // Static prediction hints.
87
+ #ifdef HAVE_BUILTIN_EXPECT
88
+ #define SNAPPY_PREDICT_FALSE(x) (__builtin_expect(x, 0))
89
+ #define SNAPPY_PREDICT_TRUE(x) (__builtin_expect(!!(x), 1))
90
+ #else
91
+ #define SNAPPY_PREDICT_FALSE(x) x
92
+ #define SNAPPY_PREDICT_TRUE(x) x
93
+ #endif
94
+
95
+ // This is only used for recomputing the tag byte table used during
96
+ // decompression; for simplicity we just remove it from the open-source
97
+ // version (anyone who wants to regenerate it can just do the call
98
+ // themselves within main()).
99
+ #define DEFINE_bool(flag_name, default_value, description) \
100
+ bool FLAGS_ ## flag_name = default_value
101
+ #define DECLARE_bool(flag_name) \
102
+ extern bool FLAGS_ ## flag_name
103
+
104
+ namespace snappy {
105
+
106
+ static const uint32 kuint32max = static_cast<uint32>(0xFFFFFFFF);
107
+ static const int64 kint64max = static_cast<int64>(0x7FFFFFFFFFFFFFFFLL);
108
+
109
+ // Potentially unaligned loads and stores.
110
+
111
+ // x86, PowerPC, and ARM64 can simply do these loads and stores native.
112
+
113
+ #if defined(__i386__) || defined(__x86_64__) || defined(__powerpc__) || \
114
+ defined(__aarch64__)
115
+
116
+ #define UNALIGNED_LOAD16(_p) (*reinterpret_cast<const uint16 *>(_p))
117
+ #define UNALIGNED_LOAD32(_p) (*reinterpret_cast<const uint32 *>(_p))
118
+ #define UNALIGNED_LOAD64(_p) (*reinterpret_cast<const uint64 *>(_p))
119
+
120
+ #define UNALIGNED_STORE16(_p, _val) (*reinterpret_cast<uint16 *>(_p) = (_val))
121
+ #define UNALIGNED_STORE32(_p, _val) (*reinterpret_cast<uint32 *>(_p) = (_val))
122
+ #define UNALIGNED_STORE64(_p, _val) (*reinterpret_cast<uint64 *>(_p) = (_val))
123
+
124
+ // ARMv7 and newer support native unaligned accesses, but only of 16-bit
125
+ // and 32-bit values (not 64-bit); older versions either raise a fatal signal,
126
+ // do an unaligned read and rotate the words around a bit, or do the reads very
127
+ // slowly (trip through kernel mode). There's no simple #define that says just
128
+ // “ARMv7 or higher”, so we have to filter away all ARMv5 and ARMv6
129
+ // sub-architectures.
130
+ //
131
+ // This is a mess, but there's not much we can do about it.
132
+ //
133
+ // To further complicate matters, only LDR instructions (single reads) are
134
+ // allowed to be unaligned, not LDRD (two reads) or LDM (many reads). Unless we
135
+ // explicitly tell the compiler that these accesses can be unaligned, it can and
136
+ // will combine accesses. On armcc, the way to signal this is done by accessing
137
+ // through the type (uint32 __packed *), but GCC has no such attribute
138
+ // (it ignores __attribute__((packed)) on individual variables). However,
139
+ // we can tell it that a _struct_ is unaligned, which has the same effect,
140
+ // so we do that.
141
+
142
+ #elif defined(__arm__) && \
143
+ !defined(__ARM_ARCH_4__) && \
144
+ !defined(__ARM_ARCH_4T__) && \
145
+ !defined(__ARM_ARCH_5__) && \
146
+ !defined(__ARM_ARCH_5T__) && \
147
+ !defined(__ARM_ARCH_5TE__) && \
148
+ !defined(__ARM_ARCH_5TEJ__) && \
149
+ !defined(__ARM_ARCH_6__) && \
150
+ !defined(__ARM_ARCH_6J__) && \
151
+ !defined(__ARM_ARCH_6K__) && \
152
+ !defined(__ARM_ARCH_6Z__) && \
153
+ !defined(__ARM_ARCH_6ZK__) && \
154
+ !defined(__ARM_ARCH_6T2__)
155
+
156
+ #if __GNUC__
157
+ #define ATTRIBUTE_PACKED __attribute__((__packed__))
158
+ #else
159
+ #define ATTRIBUTE_PACKED
160
+ #endif
161
+
162
+ namespace base {
163
+ namespace internal {
164
+
165
+ struct Unaligned16Struct {
166
+ uint16 value;
167
+ uint8 dummy; // To make the size non-power-of-two.
168
+ } ATTRIBUTE_PACKED;
169
+
170
+ struct Unaligned32Struct {
171
+ uint32 value;
172
+ uint8 dummy; // To make the size non-power-of-two.
173
+ } ATTRIBUTE_PACKED;
174
+
175
+ } // namespace internal
176
+ } // namespace base
177
+
178
+ #define UNALIGNED_LOAD16(_p) \
179
+ ((reinterpret_cast<const ::snappy::base::internal::Unaligned16Struct *>(_p))->value)
180
+ #define UNALIGNED_LOAD32(_p) \
181
+ ((reinterpret_cast<const ::snappy::base::internal::Unaligned32Struct *>(_p))->value)
182
+
183
+ #define UNALIGNED_STORE16(_p, _val) \
184
+ ((reinterpret_cast< ::snappy::base::internal::Unaligned16Struct *>(_p))->value = \
185
+ (_val))
186
+ #define UNALIGNED_STORE32(_p, _val) \
187
+ ((reinterpret_cast< ::snappy::base::internal::Unaligned32Struct *>(_p))->value = \
188
+ (_val))
189
+
190
+ // TODO(user): NEON supports unaligned 64-bit loads and stores.
191
+ // See if that would be more efficient on platforms supporting it,
192
+ // at least for copies.
193
+
194
+ inline uint64 UNALIGNED_LOAD64(const void *p) {
195
+ uint64 t;
196
+ memcpy(&t, p, sizeof t);
197
+ return t;
198
+ }
199
+
200
+ inline void UNALIGNED_STORE64(void *p, uint64 v) {
201
+ memcpy(p, &v, sizeof v);
202
+ }
203
+
204
+ #else
205
+
206
+ // These functions are provided for architectures that don't support
207
+ // unaligned loads and stores.
208
+
209
+ inline uint16 UNALIGNED_LOAD16(const void *p) {
210
+ uint16 t;
211
+ memcpy(&t, p, sizeof t);
212
+ return t;
213
+ }
214
+
215
+ inline uint32 UNALIGNED_LOAD32(const void *p) {
216
+ uint32 t;
217
+ memcpy(&t, p, sizeof t);
218
+ return t;
219
+ }
220
+
221
+ inline uint64 UNALIGNED_LOAD64(const void *p) {
222
+ uint64 t;
223
+ memcpy(&t, p, sizeof t);
224
+ return t;
225
+ }
226
+
227
+ inline void UNALIGNED_STORE16(void *p, uint16 v) {
228
+ memcpy(p, &v, sizeof v);
229
+ }
230
+
231
+ inline void UNALIGNED_STORE32(void *p, uint32 v) {
232
+ memcpy(p, &v, sizeof v);
233
+ }
234
+
235
+ inline void UNALIGNED_STORE64(void *p, uint64 v) {
236
+ memcpy(p, &v, sizeof v);
237
+ }
238
+
239
+ #endif
240
+
241
+ // The following guarantees declaration of the byte swap functions.
242
+ #if defined(SNAPPY_IS_BIG_ENDIAN)
243
+
244
+ #ifdef HAVE_SYS_BYTEORDER_H
245
+ #include <sys/byteorder.h>
246
+ #endif
247
+
248
+ #ifdef HAVE_SYS_ENDIAN_H
249
+ #include <sys/endian.h>
250
+ #endif
251
+
252
+ #ifdef _MSC_VER
253
+ #include <stdlib.h>
254
+ #define bswap_16(x) _byteswap_ushort(x)
255
+ #define bswap_32(x) _byteswap_ulong(x)
256
+ #define bswap_64(x) _byteswap_uint64(x)
257
+
258
+ #elif defined(__APPLE__)
259
+ // Mac OS X / Darwin features
260
+ #include <libkern/OSByteOrder.h>
261
+ #define bswap_16(x) OSSwapInt16(x)
262
+ #define bswap_32(x) OSSwapInt32(x)
263
+ #define bswap_64(x) OSSwapInt64(x)
264
+
265
+ #elif defined(HAVE_BYTESWAP_H)
266
+ #include <byteswap.h>
267
+
268
+ #elif defined(bswap32)
269
+ // FreeBSD defines bswap{16,32,64} in <sys/endian.h> (already #included).
270
+ #define bswap_16(x) bswap16(x)
271
+ #define bswap_32(x) bswap32(x)
272
+ #define bswap_64(x) bswap64(x)
273
+
274
+ #elif defined(BSWAP_64)
275
+ // Solaris 10 defines BSWAP_{16,32,64} in <sys/byteorder.h> (already #included).
276
+ #define bswap_16(x) BSWAP_16(x)
277
+ #define bswap_32(x) BSWAP_32(x)
278
+ #define bswap_64(x) BSWAP_64(x)
279
+
280
+ #else
281
+
282
+ inline uint16 bswap_16(uint16 x) {
283
+ return (x << 8) | (x >> 8);
284
+ }
285
+
286
+ inline uint32 bswap_32(uint32 x) {
287
+ x = ((x & 0xff00ff00UL) >> 8) | ((x & 0x00ff00ffUL) << 8);
288
+ return (x >> 16) | (x << 16);
289
+ }
290
+
291
+ inline uint64 bswap_64(uint64 x) {
292
+ x = ((x & 0xff00ff00ff00ff00ULL) >> 8) | ((x & 0x00ff00ff00ff00ffULL) << 8);
293
+ x = ((x & 0xffff0000ffff0000ULL) >> 16) | ((x & 0x0000ffff0000ffffULL) << 16);
294
+ return (x >> 32) | (x << 32);
295
+ }
296
+
297
+ #endif
298
+
299
+ #endif // defined(SNAPPY_IS_BIG_ENDIAN)
300
+
301
+ // Convert to little-endian storage, opposite of network format.
302
+ // Convert x from host to little endian: x = LittleEndian.FromHost(x);
303
+ // convert x from little endian to host: x = LittleEndian.ToHost(x);
304
+ //
305
+ // Store values into unaligned memory converting to little endian order:
306
+ // LittleEndian.Store16(p, x);
307
+ //
308
+ // Load unaligned values stored in little endian converting to host order:
309
+ // x = LittleEndian.Load16(p);
310
+ class LittleEndian {
311
+ public:
312
+ // Conversion functions.
313
+ #if defined(SNAPPY_IS_BIG_ENDIAN)
314
+
315
+ static uint16 FromHost16(uint16 x) { return bswap_16(x); }
316
+ static uint16 ToHost16(uint16 x) { return bswap_16(x); }
317
+
318
+ static uint32 FromHost32(uint32 x) { return bswap_32(x); }
319
+ static uint32 ToHost32(uint32 x) { return bswap_32(x); }
320
+
321
+ static bool IsLittleEndian() { return false; }
322
+
323
+ #else // !defined(SNAPPY_IS_BIG_ENDIAN)
324
+
325
+ static uint16 FromHost16(uint16 x) { return x; }
326
+ static uint16 ToHost16(uint16 x) { return x; }
327
+
328
+ static uint32 FromHost32(uint32 x) { return x; }
329
+ static uint32 ToHost32(uint32 x) { return x; }
330
+
331
+ static bool IsLittleEndian() { return true; }
332
+
333
+ #endif // !defined(SNAPPY_IS_BIG_ENDIAN)
334
+
335
+ // Functions to do unaligned loads and stores in little-endian order.
336
+ static uint16 Load16(const void *p) {
337
+ return ToHost16(UNALIGNED_LOAD16(p));
338
+ }
339
+
340
+ static void Store16(void *p, uint16 v) {
341
+ UNALIGNED_STORE16(p, FromHost16(v));
342
+ }
343
+
344
+ static uint32 Load32(const void *p) {
345
+ return ToHost32(UNALIGNED_LOAD32(p));
346
+ }
347
+
348
+ static void Store32(void *p, uint32 v) {
349
+ UNALIGNED_STORE32(p, FromHost32(v));
350
+ }
351
+ };
352
+
353
+ // Some bit-manipulation functions.
354
+ class Bits {
355
+ public:
356
+ // Return floor(log2(n)) for positive integer n. Returns -1 iff n == 0.
357
+ static int Log2Floor(uint32 n);
358
+
359
+ // Return the first set least / most significant bit, 0-indexed. Returns an
360
+ // undefined value if n == 0. FindLSBSetNonZero() is similar to ffs() except
361
+ // that it's 0-indexed.
362
+ static int FindLSBSetNonZero(uint32 n);
363
+
364
+ #if defined(ARCH_K8) || defined(ARCH_PPC) || defined(ARCH_ARM)
365
+ static int FindLSBSetNonZero64(uint64 n);
366
+ #endif // defined(ARCH_K8) || defined(ARCH_PPC) || defined(ARCH_ARM)
367
+
368
+ private:
369
+ // No copying
370
+ Bits(const Bits&);
371
+ void operator=(const Bits&);
372
+ };
373
+
374
+ #ifdef HAVE_BUILTIN_CTZ
375
+
376
+ inline int Bits::Log2Floor(uint32 n) {
377
+ return n == 0 ? -1 : 31 ^ __builtin_clz(n);
378
+ }
379
+
380
+ inline int Bits::FindLSBSetNonZero(uint32 n) {
381
+ return __builtin_ctz(n);
382
+ }
383
+
384
+ #if defined(ARCH_K8) || defined(ARCH_PPC) || defined(ARCH_ARM)
385
+ inline int Bits::FindLSBSetNonZero64(uint64 n) {
386
+ return __builtin_ctzll(n);
387
+ }
388
+ #endif // defined(ARCH_K8) || defined(ARCH_PPC) || defined(ARCH_ARM)
389
+
390
+ #elif defined(_MSC_VER)
391
+
392
+ inline int Bits::Log2Floor(uint32 n) {
393
+ unsigned long where;
394
+ if (_BitScanReverse(&where, n)) {
395
+ return where;
396
+ } else {
397
+ return -1;
398
+ }
399
+ }
400
+
401
+ inline int Bits::FindLSBSetNonZero(uint32 n) {
402
+ unsigned long where;
403
+ if (_BitScanForward(&where, n)) return static_cast<int>(where);
404
+ return 32;
405
+ }
406
+
407
+ #if defined(ARCH_K8) || defined(ARCH_PPC) || defined(ARCH_ARM)
408
+ inline int Bits::FindLSBSetNonZero64(uint64 n) {
409
+ unsigned long where;
410
+ if (_BitScanForward64(&where, n)) return static_cast<int>(where);
411
+ return 64;
412
+ }
413
+ #endif // defined(ARCH_K8) || defined(ARCH_PPC) || defined(ARCH_ARM)
414
+
415
+ #else // Portable versions.
416
+
417
+ inline int Bits::Log2Floor(uint32 n) {
418
+ if (n == 0)
419
+ return -1;
420
+ int log = 0;
421
+ uint32 value = n;
422
+ for (int i = 4; i >= 0; --i) {
423
+ int shift = (1 << i);
424
+ uint32 x = value >> shift;
425
+ if (x != 0) {
426
+ value = x;
427
+ log += shift;
428
+ }
429
+ }
430
+ assert(value == 1);
431
+ return log;
432
+ }
433
+
434
+ inline int Bits::FindLSBSetNonZero(uint32 n) {
435
+ int rc = 31;
436
+ for (int i = 4, shift = 1 << 4; i >= 0; --i) {
437
+ const uint32 x = n << shift;
438
+ if (x != 0) {
439
+ n = x;
440
+ rc -= shift;
441
+ }
442
+ shift >>= 1;
443
+ }
444
+ return rc;
445
+ }
446
+
447
+ #if defined(ARCH_K8) || defined(ARCH_PPC) || defined(ARCH_ARM)
448
+ // FindLSBSetNonZero64() is defined in terms of FindLSBSetNonZero().
449
+ inline int Bits::FindLSBSetNonZero64(uint64 n) {
450
+ const uint32 bottombits = static_cast<uint32>(n);
451
+ if (bottombits == 0) {
452
+ // Bottom bits are zero, so scan in top bits
453
+ return 32 + FindLSBSetNonZero(static_cast<uint32>(n >> 32));
454
+ } else {
455
+ return FindLSBSetNonZero(bottombits);
456
+ }
457
+ }
458
+ #endif // defined(ARCH_K8) || defined(ARCH_PPC) || defined(ARCH_ARM)
459
+
460
+ #endif // End portable versions.
461
+
462
+ // Variable-length integer encoding.
463
+ class Varint {
464
+ public:
465
+ // Maximum lengths of varint encoding of uint32.
466
+ static const int kMax32 = 5;
467
+
468
+ // Attempts to parse a varint32 from a prefix of the bytes in [ptr,limit-1].
469
+ // Never reads a character at or beyond limit. If a valid/terminated varint32
470
+ // was found in the range, stores it in *OUTPUT and returns a pointer just
471
+ // past the last byte of the varint32. Else returns NULL. On success,
472
+ // "result <= limit".
473
+ static const char* Parse32WithLimit(const char* ptr, const char* limit,
474
+ uint32* OUTPUT);
475
+
476
+ // REQUIRES "ptr" points to a buffer of length sufficient to hold "v".
477
+ // EFFECTS Encodes "v" into "ptr" and returns a pointer to the
478
+ // byte just past the last encoded byte.
479
+ static char* Encode32(char* ptr, uint32 v);
480
+
481
+ // EFFECTS Appends the varint representation of "value" to "*s".
482
+ static void Append32(string* s, uint32 value);
483
+ };
484
+
485
+ inline const char* Varint::Parse32WithLimit(const char* p,
486
+ const char* l,
487
+ uint32* OUTPUT) {
488
+ const unsigned char* ptr = reinterpret_cast<const unsigned char*>(p);
489
+ const unsigned char* limit = reinterpret_cast<const unsigned char*>(l);
490
+ uint32 b, result;
491
+ if (ptr >= limit) return NULL;
492
+ b = *(ptr++); result = b & 127; if (b < 128) goto done;
493
+ if (ptr >= limit) return NULL;
494
+ b = *(ptr++); result |= (b & 127) << 7; if (b < 128) goto done;
495
+ if (ptr >= limit) return NULL;
496
+ b = *(ptr++); result |= (b & 127) << 14; if (b < 128) goto done;
497
+ if (ptr >= limit) return NULL;
498
+ b = *(ptr++); result |= (b & 127) << 21; if (b < 128) goto done;
499
+ if (ptr >= limit) return NULL;
500
+ b = *(ptr++); result |= (b & 127) << 28; if (b < 16) goto done;
501
+ return NULL; // Value is too long to be a varint32
502
+ done:
503
+ *OUTPUT = result;
504
+ return reinterpret_cast<const char*>(ptr);
505
+ }
506
+
507
+ inline char* Varint::Encode32(char* sptr, uint32 v) {
508
+ // Operate on characters as unsigneds
509
+ unsigned char* ptr = reinterpret_cast<unsigned char*>(sptr);
510
+ static const int B = 128;
511
+ if (v < (1<<7)) {
512
+ *(ptr++) = v;
513
+ } else if (v < (1<<14)) {
514
+ *(ptr++) = v | B;
515
+ *(ptr++) = v>>7;
516
+ } else if (v < (1<<21)) {
517
+ *(ptr++) = v | B;
518
+ *(ptr++) = (v>>7) | B;
519
+ *(ptr++) = v>>14;
520
+ } else if (v < (1<<28)) {
521
+ *(ptr++) = v | B;
522
+ *(ptr++) = (v>>7) | B;
523
+ *(ptr++) = (v>>14) | B;
524
+ *(ptr++) = v>>21;
525
+ } else {
526
+ *(ptr++) = v | B;
527
+ *(ptr++) = (v>>7) | B;
528
+ *(ptr++) = (v>>14) | B;
529
+ *(ptr++) = (v>>21) | B;
530
+ *(ptr++) = v>>28;
531
+ }
532
+ return reinterpret_cast<char*>(ptr);
533
+ }
534
+
535
+ // If you know the internal layout of the std::string in use, you can
536
+ // replace this function with one that resizes the string without
537
+ // filling the new space with zeros (if applicable) --
538
+ // it will be non-portable but faster.
539
+ inline void STLStringResizeUninitialized(string* s, size_t new_size) {
540
+ s->resize(new_size);
541
+ }
542
+
543
+ // Return a mutable char* pointing to a string's internal buffer,
544
+ // which may not be null-terminated. Writing through this pointer will
545
+ // modify the string.
546
+ //
547
+ // string_as_array(&str)[i] is valid for 0 <= i < str.size() until the
548
+ // next call to a string method that invalidates iterators.
549
+ //
550
+ // As of 2006-04, there is no standard-blessed way of getting a
551
+ // mutable reference to a string's internal buffer. However, issue 530
552
+ // (http://www.open-std.org/JTC1/SC22/WG21/docs/lwg-defects.html#530)
553
+ // proposes this as the method. It will officially be part of the standard
554
+ // for C++0x. This should already work on all current implementations.
555
+ inline char* string_as_array(string* str) {
556
+ return str->empty() ? NULL : &*str->begin();
557
+ }
558
+
559
+ } // namespace snappy
560
+
561
+ #endif // THIRD_PARTY_SNAPPY_OPENSOURCE_SNAPPY_STUBS_INTERNAL_H_