1 // Copyright 2014 The Chromium Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file.
5 // All data that is passed through a WebSocket with type "Text" needs to be
6 // validated as UTF8. Since this is done on the IO thread, it needs to be
9 // We are only interested in the performance on valid UTF8. Invalid UTF8 will
10 // result in a connection failure, so is unlikely to become a source of
11 // performance issues.
13 #include "base/i18n/streaming_utf8_validator.h"
19 #include "base/bind.h"
20 #include "base/callback.h"
21 #include "base/macros.h"
22 #include "base/strings/string_util.h"
23 #include "base/strings/stringprintf.h"
24 #include "base/test/perf_time_logger.h"
25 #include "testing/gtest/include/gtest/gtest.h"
30 // We want to test ranges of valid UTF-8 sequences. These ranges are inclusive.
31 // They are intended to be large enough that the validator needs to do
32 // meaningful work while being in some sense "realistic" (eg. control characters
34 const char kOneByteSeqRangeStart[] = " "; // U+0020
35 const char kOneByteSeqRangeEnd[] = "~"; // U+007E
37 const char kTwoByteSeqRangeStart[] = "\xc2\xa0"; // U+00A0 non-breaking space
38 const char kTwoByteSeqRangeEnd[] = "\xc9\x8f"; // U+024F small y with stroke
40 const char kThreeByteSeqRangeStart[] = "\xe3\x81\x82"; // U+3042 Hiragana "a"
41 const char kThreeByteSeqRangeEnd[] = "\xe9\xbf\x83"; // U+9FC3 "to blink"
43 const char kFourByteSeqRangeStart[] = "\xf0\xa0\x80\x8b"; // U+2000B
44 const char kFourByteSeqRangeEnd[] = "\xf0\xaa\x9a\xb2"; // U+2A6B2
46 // The different lengths of strings to test.
47 const size_t kTestLengths[] = {1, 32, 256, 32768, 1 << 20};
49 // Simplest possible byte-at-a-time validator, to provide a baseline
50 // for comparison. This is only tried on 1-byte UTF-8 sequences, as
51 // the results will not be meaningful with sequences containing
53 bool IsString7Bit(const std::string& s) {
54 for (std::string::const_iterator it = s.begin(); it != s.end(); ++it) {
61 // Assumes that |previous| is a valid UTF-8 sequence, and attempts to return
62 // the next one. Is just barely smart enough to iterate through the ranges
64 std::string NextUtf8Sequence(const std::string& previous) {
65 DCHECK(StreamingUtf8Validator::Validate(previous));
66 std::string next = previous;
67 for (int i = static_cast<int>(previous.length() - 1); i >= 0; --i) {
68 // All bytes in a UTF-8 sequence except the first one are
69 // constrained to the range 0x80 to 0xbf, inclusive. When we
70 // increment past 0xbf, we carry into the previous byte.
71 if (i > 0 && next[i] == '\xbf') {
78 DCHECK(StreamingUtf8Validator::Validate(next))
79 << "Result \"" << next << "\" failed validation";
83 typedef bool (*TestTargetType)(const std::string&);
85 // Run fuction |target| over |test_string| |times| times, and report the results
86 // using |description|.
87 bool RunTest(const std::string& description,
88 TestTargetType target,
89 const std::string& test_string,
91 base::PerfTimeLogger timer(description.c_str());
93 for (int i = 0; i < times; ++i) {
94 result = target(test_string) && result;
100 // Construct a string by repeating |input| enough times to equal or exceed
102 std::string ConstructRepeatedTestString(const std::string& input,
104 std::string output = input;
105 while (output.length() * 2 < length) {
108 if (output.length() < length) {
109 output += ConstructRepeatedTestString(input, length - output.length());
114 // Construct a string by expanding the range of UTF-8 sequences
115 // between |input_start| and |input_end|, inclusive, and then
116 // repeating the resulting string until it equals or exceeds |length|
117 // bytes. |input_start| and |input_end| must be valid UTF-8
119 std::string ConstructRangedTestString(const std::string& input_start,
120 const std::string& input_end,
122 std::string output = input_start;
123 std::string input = input_start;
124 while (output.length() < length && input != input_end) {
125 input = NextUtf8Sequence(input);
128 if (output.length() < length) {
129 output = ConstructRepeatedTestString(output, length);
134 struct TestFunctionDescription {
135 TestTargetType function;
136 const char* function_name;
139 bool IsStringUTF8(const std::string& str) {
140 return base::IsStringUTF8(base::StringPiece(str));
143 // IsString7Bit is intentionally placed last so it can be excluded easily.
144 const TestFunctionDescription kTestFunctions[] = {
145 {&StreamingUtf8Validator::Validate, "StreamingUtf8Validator"},
146 {&IsStringUTF8, "IsStringUTF8"}, {&IsString7Bit, "IsString7Bit"}};
148 // Construct a test string from |construct_test_string| for each of the lengths
149 // in |kTestLengths| in turn. For each string, run each test in |test_functions|
150 // for a number of iterations such that the total number of bytes validated
154 base::Callback<std::string(size_t length)> construct_test_string,
155 const TestFunctionDescription* test_functions,
157 for (size_t i = 0; i < arraysize(kTestLengths); ++i) {
158 const size_t length = kTestLengths[i];
159 const std::string test_string = construct_test_string.Run(length);
160 const int real_length = static_cast<int>(test_string.length());
161 const int times = (1 << 24) / real_length;
162 for (size_t test_index = 0; test_index < test_count; ++test_index) {
163 EXPECT_TRUE(RunTest(StringPrintf(format,
164 test_functions[test_index].function_name,
167 test_functions[test_index].function,
174 TEST(StreamingUtf8ValidatorPerfTest, OneByteRepeated) {
175 RunSomeTests("%s: bytes=1 repeated length=%d repeat=%d",
176 base::Bind(ConstructRepeatedTestString, kOneByteSeqRangeStart),
181 TEST(StreamingUtf8ValidatorPerfTest, OneByteRange) {
182 RunSomeTests("%s: bytes=1 ranged length=%d repeat=%d",
183 base::Bind(ConstructRangedTestString,
184 kOneByteSeqRangeStart,
185 kOneByteSeqRangeEnd),
190 TEST(StreamingUtf8ValidatorPerfTest, TwoByteRepeated) {
191 RunSomeTests("%s: bytes=2 repeated length=%d repeat=%d",
192 base::Bind(ConstructRepeatedTestString, kTwoByteSeqRangeStart),
197 TEST(StreamingUtf8ValidatorPerfTest, TwoByteRange) {
198 RunSomeTests("%s: bytes=2 ranged length=%d repeat=%d",
199 base::Bind(ConstructRangedTestString,
200 kTwoByteSeqRangeStart,
201 kTwoByteSeqRangeEnd),
206 TEST(StreamingUtf8ValidatorPerfTest, ThreeByteRepeated) {
208 "%s: bytes=3 repeated length=%d repeat=%d",
209 base::Bind(ConstructRepeatedTestString, kThreeByteSeqRangeStart),
214 TEST(StreamingUtf8ValidatorPerfTest, ThreeByteRange) {
215 RunSomeTests("%s: bytes=3 ranged length=%d repeat=%d",
216 base::Bind(ConstructRangedTestString,
217 kThreeByteSeqRangeStart,
218 kThreeByteSeqRangeEnd),
223 TEST(StreamingUtf8ValidatorPerfTest, FourByteRepeated) {
224 RunSomeTests("%s: bytes=4 repeated length=%d repeat=%d",
225 base::Bind(ConstructRepeatedTestString, kFourByteSeqRangeStart),
230 TEST(StreamingUtf8ValidatorPerfTest, FourByteRange) {
231 RunSomeTests("%s: bytes=4 ranged length=%d repeat=%d",
232 base::Bind(ConstructRangedTestString,
233 kFourByteSeqRangeStart,
234 kFourByteSeqRangeEnd),