2 * Copyright 2016 Facebook, Inc.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
17 #include <folly/io/Compression.h>
21 #include <unordered_map>
23 #include <boost/noncopyable.hpp>
24 #include <glog/logging.h>
25 #include <gtest/gtest.h>
27 #include <folly/Benchmark.h>
28 #include <folly/Hash.h>
29 #include <folly/Random.h>
30 #include <folly/Varint.h>
31 #include <folly/io/IOBufQueue.h>
33 namespace folly { namespace io { namespace test {
35 class DataHolder : private boost::noncopyable {
37 uint64_t hash(size_t size) const;
38 ByteRange data(size_t size) const;
41 explicit DataHolder(size_t sizeLog2);
43 std::unique_ptr<uint8_t[]> data_;
44 mutable std::unordered_map<uint64_t, uint64_t> hashCache_;
47 DataHolder::DataHolder(size_t sizeLog2)
48 : size_(size_t(1) << sizeLog2),
49 data_(new uint8_t[size_]) {
52 uint64_t DataHolder::hash(size_t size) const {
53 CHECK_LE(size, size_);
54 auto p = hashCache_.find(size);
55 if (p != hashCache_.end()) {
59 uint64_t h = folly::hash::fnv64_buf(data_.get(), size);
64 ByteRange DataHolder::data(size_t size) const {
65 CHECK_LE(size, size_);
66 return ByteRange(data_.get(), size);
69 uint64_t hashIOBuf(const IOBuf* buf) {
70 uint64_t h = folly::hash::FNV_64_HASH_START;
71 for (auto& range : *buf) {
72 h = folly::hash::fnv64_buf(range.data(), range.size(), h);
77 class RandomDataHolder : public DataHolder {
79 explicit RandomDataHolder(size_t sizeLog2);
82 RandomDataHolder::RandomDataHolder(size_t sizeLog2)
83 : DataHolder(sizeLog2) {
84 constexpr size_t numThreadsLog2 = 3;
85 constexpr size_t numThreads = size_t(1) << numThreadsLog2;
87 uint32_t seed = randomNumberSeed();
89 std::vector<std::thread> threads;
90 threads.reserve(numThreads);
91 for (size_t t = 0; t < numThreads; ++t) {
93 [this, seed, t, numThreadsLog2, sizeLog2] () {
94 std::mt19937 rng(seed + t);
95 size_t countLog2 = sizeLog2 - numThreadsLog2;
96 size_t start = size_t(t) << countLog2;
97 for (size_t i = 0; i < countLog2; ++i) {
98 this->data_[start + i] = rng();
103 for (auto& t : threads) {
108 class ConstantDataHolder : public DataHolder {
110 explicit ConstantDataHolder(size_t sizeLog2);
113 ConstantDataHolder::ConstantDataHolder(size_t sizeLog2)
114 : DataHolder(sizeLog2) {
115 memset(data_.get(), 'a', size_);
118 constexpr size_t dataSizeLog2 = 27; // 128MiB
119 RandomDataHolder randomDataHolder(dataSizeLog2);
120 ConstantDataHolder constantDataHolder(dataSizeLog2);
122 TEST(CompressionTestNeedsUncompressedLength, Simple) {
123 EXPECT_FALSE(getCodec(CodecType::NO_COMPRESSION)->needsUncompressedLength());
124 EXPECT_TRUE(getCodec(CodecType::LZ4)->needsUncompressedLength());
125 EXPECT_FALSE(getCodec(CodecType::SNAPPY)->needsUncompressedLength());
126 EXPECT_FALSE(getCodec(CodecType::ZLIB)->needsUncompressedLength());
127 EXPECT_FALSE(getCodec(CodecType::LZ4_VARINT_SIZE)->needsUncompressedLength());
128 EXPECT_TRUE(getCodec(CodecType::LZMA2)->needsUncompressedLength());
129 EXPECT_FALSE(getCodec(CodecType::LZMA2_VARINT_SIZE)
130 ->needsUncompressedLength());
131 EXPECT_TRUE(getCodec(CodecType::ZSTD_BETA)->needsUncompressedLength());
132 EXPECT_FALSE(getCodec(CodecType::GZIP)->needsUncompressedLength());
135 class CompressionTest
136 : public testing::TestWithParam<std::tr1::tuple<int, CodecType>> {
138 void SetUp() override {
139 auto tup = GetParam();
140 uncompressedLength_ = uint64_t(1) << std::tr1::get<0>(tup);
141 codec_ = getCodec(std::tr1::get<1>(tup));
144 void runSimpleTest(const DataHolder& dh);
146 uint64_t uncompressedLength_;
147 std::unique_ptr<Codec> codec_;
150 void CompressionTest::runSimpleTest(const DataHolder& dh) {
151 auto original = IOBuf::wrapBuffer(dh.data(uncompressedLength_));
152 auto compressed = codec_->compress(original.get());
153 if (!codec_->needsUncompressedLength()) {
154 auto uncompressed = codec_->uncompress(compressed.get());
156 EXPECT_EQ(uncompressedLength_, uncompressed->computeChainDataLength());
157 EXPECT_EQ(dh.hash(uncompressedLength_), hashIOBuf(uncompressed.get()));
160 auto uncompressed = codec_->uncompress(compressed.get(),
161 uncompressedLength_);
162 EXPECT_EQ(uncompressedLength_, uncompressed->computeChainDataLength());
163 EXPECT_EQ(dh.hash(uncompressedLength_), hashIOBuf(uncompressed.get()));
167 TEST_P(CompressionTest, RandomData) {
168 runSimpleTest(randomDataHolder);
171 TEST_P(CompressionTest, ConstantData) {
172 runSimpleTest(constantDataHolder);
175 INSTANTIATE_TEST_CASE_P(
178 testing::Combine(testing::Values(0, 1, 12, 22, 25, 27),
179 testing::Values(CodecType::NO_COMPRESSION,
183 CodecType::LZ4_VARINT_SIZE,
185 CodecType::LZMA2_VARINT_SIZE,
186 CodecType::ZSTD_BETA,
189 class CompressionVarintTest
190 : public testing::TestWithParam<std::tr1::tuple<int, CodecType>> {
192 void SetUp() override {
193 auto tup = GetParam();
194 uncompressedLength_ = uint64_t(1) << std::tr1::get<0>(tup);
195 codec_ = getCodec(std::tr1::get<1>(tup));
198 void runSimpleTest(const DataHolder& dh);
200 uint64_t uncompressedLength_;
201 std::unique_ptr<Codec> codec_;
204 inline uint64_t oneBasedMsbPos(uint64_t number) {
206 for (; number > 0; ++pos, number >>= 1) {
211 void CompressionVarintTest::runSimpleTest(const DataHolder& dh) {
212 auto original = IOBuf::wrapBuffer(dh.data(uncompressedLength_));
213 auto compressed = codec_->compress(original.get());
216 Random::rand64(std::max(9UL, oneBasedMsbPos(uncompressedLength_)) / 9UL);
217 auto tinyBuf = IOBuf::copyBuffer(compressed->data(),
218 std::min(compressed->length(), breakPoint));
219 compressed->trimStart(breakPoint);
220 tinyBuf->prependChain(std::move(compressed));
221 compressed = std::move(tinyBuf);
223 auto uncompressed = codec_->uncompress(compressed.get());
225 EXPECT_EQ(uncompressedLength_, uncompressed->computeChainDataLength());
226 EXPECT_EQ(dh.hash(uncompressedLength_), hashIOBuf(uncompressed.get()));
229 TEST_P(CompressionVarintTest, RandomData) { runSimpleTest(randomDataHolder); }
231 TEST_P(CompressionVarintTest, ConstantData) {
232 runSimpleTest(constantDataHolder);
235 INSTANTIATE_TEST_CASE_P(
236 CompressionVarintTest,
237 CompressionVarintTest,
238 testing::Combine(testing::Values(0, 1, 12, 22, 25, 27),
239 testing::Values(CodecType::LZ4_VARINT_SIZE,
240 CodecType::LZMA2_VARINT_SIZE)));
242 class CompressionCorruptionTest : public testing::TestWithParam<CodecType> {
244 void SetUp() override { codec_ = getCodec(GetParam()); }
246 void runSimpleTest(const DataHolder& dh);
248 std::unique_ptr<Codec> codec_;
251 void CompressionCorruptionTest::runSimpleTest(const DataHolder& dh) {
252 constexpr uint64_t uncompressedLength = 42;
253 auto original = IOBuf::wrapBuffer(dh.data(uncompressedLength));
254 auto compressed = codec_->compress(original.get());
256 if (!codec_->needsUncompressedLength()) {
257 auto uncompressed = codec_->uncompress(compressed.get());
258 EXPECT_EQ(uncompressedLength, uncompressed->computeChainDataLength());
259 EXPECT_EQ(dh.hash(uncompressedLength), hashIOBuf(uncompressed.get()));
262 auto uncompressed = codec_->uncompress(compressed.get(),
264 EXPECT_EQ(uncompressedLength, uncompressed->computeChainDataLength());
265 EXPECT_EQ(dh.hash(uncompressedLength), hashIOBuf(uncompressed.get()));
268 EXPECT_THROW(codec_->uncompress(compressed.get(), uncompressedLength + 1),
271 // Corrupt the first character
272 ++(compressed->writableData()[0]);
274 if (!codec_->needsUncompressedLength()) {
275 EXPECT_THROW(codec_->uncompress(compressed.get()),
279 EXPECT_THROW(codec_->uncompress(compressed.get(), uncompressedLength),
283 TEST_P(CompressionCorruptionTest, RandomData) {
284 runSimpleTest(randomDataHolder);
287 TEST_P(CompressionCorruptionTest, ConstantData) {
288 runSimpleTest(constantDataHolder);
291 INSTANTIATE_TEST_CASE_P(
292 CompressionCorruptionTest,
293 CompressionCorruptionTest,
295 // NO_COMPRESSION can't detect corruption
296 // LZ4 can't detect corruption reliably (sigh)
302 int main(int argc, char *argv[]) {
303 testing::InitGoogleTest(&argc, argv);
304 gflags::ParseCommandLineFlags(&argc, &argv, true);
306 auto ret = RUN_ALL_TESTS();
308 folly::runBenchmarksOnFlag();