| OLD | NEW |
|---|
| (Empty) |
| 1 // Copyright (c) 2012 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. | |
| 4 | |
| 5 #include "content/browser/speech/google_streaming_remote_engine.h" | |
| 6 | |
| 7 #include <stddef.h> | |
| 8 #include <stdint.h> | |
| 9 | |
| 10 #include <memory> | |
| 11 #include <queue> | |
| 12 | |
| 13 #include "base/big_endian.h" | |
| 14 #include "base/message_loop/message_loop.h" | |
| 15 #include "base/numerics/safe_conversions.h" | |
| 16 #include "base/strings/utf_string_conversions.h" | |
| 17 #include "base/sys_byteorder.h" | |
| 18 #include "content/browser/speech/audio_buffer.h" | |
| 19 #include "content/browser/speech/proto/google_streaming_api.pb.h" | |
| 20 #include "content/public/common/speech_recognition_error.h" | |
| 21 #include "content/public/common/speech_recognition_result.h" | |
| 22 #include "net/base/net_errors.h" | |
| 23 #include "net/url_request/test_url_fetcher_factory.h" | |
| 24 #include "net/url_request/url_request_context_getter.h" | |
| 25 #include "net/url_request/url_request_status.h" | |
| 26 #include "testing/gtest/include/gtest/gtest.h" | |
| 27 | |
| 28 using base::HostToNet32; | |
| 29 using base::checked_cast; | |
| 30 using net::URLRequestStatus; | |
| 31 using net::TestURLFetcher; | |
| 32 using net::TestURLFetcherFactory; | |
| 33 | |
| 34 namespace content { | |
| 35 | |
| 36 // Frame types for framed POST data. | |
| 37 static const uint32_t kFrameTypePreamble = 0; | |
| 38 static const uint32_t kFrameTypeRecognitionAudio = 1; | |
| 39 | |
| 40 // Note: the terms upstream and downstream are from the point-of-view of the | |
| 41 // client (engine_under_test_). | |
| 42 | |
| 43 class GoogleStreamingRemoteEngineTest : public SpeechRecognitionEngineDelegate, | |
| 44 public testing::Test { | |
| 45 public: | |
| 46 GoogleStreamingRemoteEngineTest() | |
| 47 : last_number_of_upstream_chunks_seen_(0U), | |
| 48 error_(SPEECH_RECOGNITION_ERROR_NONE), | |
| 49 end_of_utterance_counter_(0) { } | |
| 50 | |
| 51 // Creates a speech recognition request and invokes its URL fetcher delegate | |
| 52 // with the given test data. | |
| 53 void CreateAndTestRequest(bool success, const std::string& http_response); | |
| 54 | |
| 55 // SpeechRecognitionRequestDelegate methods. | |
| 56 void OnSpeechRecognitionEngineResults( | |
| 57 const SpeechRecognitionResults& results) override { | |
| 58 results_.push(results); | |
| 59 } | |
| 60 void OnSpeechRecognitionEngineEndOfUtterance() override { | |
| 61 ++end_of_utterance_counter_; | |
| 62 } | |
| 63 void OnSpeechRecognitionEngineError( | |
| 64 const SpeechRecognitionError& error) override { | |
| 65 error_ = error.code; | |
| 66 } | |
| 67 | |
| 68 // testing::Test methods. | |
| 69 void SetUp() override; | |
| 70 void TearDown() override; | |
| 71 | |
| 72 protected: | |
| 73 enum DownstreamError { | |
| 74 DOWNSTREAM_ERROR_NONE, | |
| 75 DOWNSTREAM_ERROR_HTTP500, | |
| 76 DOWNSTREAM_ERROR_NETWORK, | |
| 77 DOWNSTREAM_ERROR_WEBSERVICE_NO_MATCH | |
| 78 }; | |
| 79 static bool ResultsAreEqual(const SpeechRecognitionResults& a, | |
| 80 const SpeechRecognitionResults& b); | |
| 81 static std::string SerializeProtobufResponse( | |
| 82 const proto::SpeechRecognitionEvent& msg); | |
| 83 | |
| 84 TestURLFetcher* GetUpstreamFetcher(); | |
| 85 TestURLFetcher* GetDownstreamFetcher(); | |
| 86 void StartMockRecognition(); | |
| 87 void EndMockRecognition(); | |
| 88 void InjectDummyAudioChunk(); | |
| 89 size_t UpstreamChunksUploadedFromLastCall(); | |
| 90 std::string LastUpstreamChunkUploaded(); | |
| 91 void ProvideMockProtoResultDownstream( | |
| 92 const proto::SpeechRecognitionEvent& result); | |
| 93 void ProvideMockResultDownstream(const SpeechRecognitionResult& result); | |
| 94 void ExpectResultsReceived(const SpeechRecognitionResults& result); | |
| 95 void ExpectFramedChunk(const std::string& chunk, uint32_t type); | |
| 96 void CloseMockDownstream(DownstreamError error); | |
| 97 | |
| 98 std::unique_ptr<GoogleStreamingRemoteEngine> engine_under_test_; | |
| 99 TestURLFetcherFactory url_fetcher_factory_; | |
| 100 size_t last_number_of_upstream_chunks_seen_; | |
| 101 base::MessageLoop message_loop_; | |
| 102 std::string response_buffer_; | |
| 103 SpeechRecognitionErrorCode error_; | |
| 104 int end_of_utterance_counter_; | |
| 105 std::queue<SpeechRecognitionResults> results_; | |
| 106 }; | |
| 107 | |
| 108 TEST_F(GoogleStreamingRemoteEngineTest, SingleDefinitiveResult) { | |
| 109 StartMockRecognition(); | |
| 110 ASSERT_TRUE(GetUpstreamFetcher()); | |
| 111 ASSERT_EQ(0U, UpstreamChunksUploadedFromLastCall()); | |
| 112 | |
| 113 // Inject some dummy audio chunks and check a corresponding chunked upload | |
| 114 // is performed every time on the server. | |
| 115 for (int i = 0; i < 3; ++i) { | |
| 116 InjectDummyAudioChunk(); | |
| 117 ASSERT_EQ(1U, UpstreamChunksUploadedFromLastCall()); | |
| 118 } | |
| 119 | |
| 120 // Ensure that a final (empty) audio chunk is uploaded on chunks end. | |
| 121 engine_under_test_->AudioChunksEnded(); | |
| 122 ASSERT_EQ(1U, UpstreamChunksUploadedFromLastCall()); | |
| 123 ASSERT_TRUE(engine_under_test_->IsRecognitionPending()); | |
| 124 | |
| 125 // Simulate a protobuf message streamed from the server containing a single | |
| 126 // result with two hypotheses. | |
| 127 SpeechRecognitionResults results; | |
| 128 results.push_back(SpeechRecognitionResult()); | |
| 129 SpeechRecognitionResult& result = results.back(); | |
| 130 result.is_provisional = false; | |
| 131 result.hypotheses.push_back( | |
| 132 SpeechRecognitionHypothesis(base::UTF8ToUTF16("hypothesis 1"), 0.1F)); | |
| 133 result.hypotheses.push_back( | |
| 134 SpeechRecognitionHypothesis(base::UTF8ToUTF16("hypothesis 2"), 0.2F)); | |
| 135 | |
| 136 ProvideMockResultDownstream(result); | |
| 137 ExpectResultsReceived(results); | |
| 138 ASSERT_TRUE(engine_under_test_->IsRecognitionPending()); | |
| 139 | |
| 140 // Ensure everything is closed cleanly after the downstream is closed. | |
| 141 CloseMockDownstream(DOWNSTREAM_ERROR_NONE); | |
| 142 ASSERT_FALSE(engine_under_test_->IsRecognitionPending()); | |
| 143 EndMockRecognition(); | |
| 144 ASSERT_EQ(SPEECH_RECOGNITION_ERROR_NONE, error_); | |
| 145 ASSERT_EQ(0U, results_.size()); | |
| 146 } | |
| 147 | |
| 148 TEST_F(GoogleStreamingRemoteEngineTest, SeveralStreamingResults) { | |
| 149 StartMockRecognition(); | |
| 150 ASSERT_TRUE(GetUpstreamFetcher()); | |
| 151 ASSERT_EQ(0U, UpstreamChunksUploadedFromLastCall()); | |
| 152 | |
| 153 for (int i = 0; i < 4; ++i) { | |
| 154 InjectDummyAudioChunk(); | |
| 155 ASSERT_EQ(1U, UpstreamChunksUploadedFromLastCall()); | |
| 156 | |
| 157 SpeechRecognitionResults results; | |
| 158 results.push_back(SpeechRecognitionResult()); | |
| 159 SpeechRecognitionResult& result = results.back(); | |
| 160 result.is_provisional = (i % 2 == 0); // Alternate result types. | |
| 161 float confidence = result.is_provisional ? 0.0F : (i * 0.1F); | |
| 162 result.hypotheses.push_back(SpeechRecognitionHypothesis( | |
| 163 base::UTF8ToUTF16("hypothesis"), confidence)); | |
| 164 | |
| 165 ProvideMockResultDownstream(result); | |
| 166 ExpectResultsReceived(results); | |
| 167 ASSERT_TRUE(engine_under_test_->IsRecognitionPending()); | |
| 168 } | |
| 169 | |
| 170 // Ensure that a final (empty) audio chunk is uploaded on chunks end. | |
| 171 engine_under_test_->AudioChunksEnded(); | |
| 172 ASSERT_EQ(1U, UpstreamChunksUploadedFromLastCall()); | |
| 173 ASSERT_TRUE(engine_under_test_->IsRecognitionPending()); | |
| 174 | |
| 175 // Simulate a final definitive result. | |
| 176 SpeechRecognitionResults results; | |
| 177 results.push_back(SpeechRecognitionResult()); | |
| 178 SpeechRecognitionResult& result = results.back(); | |
| 179 result.is_provisional = false; | |
| 180 result.hypotheses.push_back( | |
| 181 SpeechRecognitionHypothesis(base::UTF8ToUTF16("The final result"), 1.0F)); | |
| 182 ProvideMockResultDownstream(result); | |
| 183 ExpectResultsReceived(results); | |
| 184 ASSERT_TRUE(engine_under_test_->IsRecognitionPending()); | |
| 185 | |
| 186 // Ensure everything is closed cleanly after the downstream is closed. | |
| 187 CloseMockDownstream(DOWNSTREAM_ERROR_NONE); | |
| 188 ASSERT_FALSE(engine_under_test_->IsRecognitionPending()); | |
| 189 EndMockRecognition(); | |
| 190 ASSERT_EQ(SPEECH_RECOGNITION_ERROR_NONE, error_); | |
| 191 ASSERT_EQ(0U, results_.size()); | |
| 192 } | |
| 193 | |
| 194 TEST_F(GoogleStreamingRemoteEngineTest, NoFinalResultAfterAudioChunksEnded) { | |
| 195 StartMockRecognition(); | |
| 196 ASSERT_TRUE(GetUpstreamFetcher()); | |
| 197 ASSERT_EQ(0U, UpstreamChunksUploadedFromLastCall()); | |
| 198 | |
| 199 // Simulate one pushed audio chunk. | |
| 200 InjectDummyAudioChunk(); | |
| 201 ASSERT_EQ(1U, UpstreamChunksUploadedFromLastCall()); | |
| 202 | |
| 203 // Simulate the corresponding definitive result. | |
| 204 SpeechRecognitionResults results; | |
| 205 results.push_back(SpeechRecognitionResult()); | |
| 206 SpeechRecognitionResult& result = results.back(); | |
| 207 result.hypotheses.push_back( | |
| 208 SpeechRecognitionHypothesis(base::UTF8ToUTF16("hypothesis"), 1.0F)); | |
| 209 ProvideMockResultDownstream(result); | |
| 210 ExpectResultsReceived(results); | |
| 211 ASSERT_TRUE(engine_under_test_->IsRecognitionPending()); | |
| 212 | |
| 213 // Simulate a silent downstream closure after |AudioChunksEnded|. | |
| 214 engine_under_test_->AudioChunksEnded(); | |
| 215 ASSERT_EQ(1U, UpstreamChunksUploadedFromLastCall()); | |
| 216 ASSERT_TRUE(engine_under_test_->IsRecognitionPending()); | |
| 217 CloseMockDownstream(DOWNSTREAM_ERROR_NONE); | |
| 218 | |
| 219 // Expect an empty result, aimed at notifying recognition ended with no | |
| 220 // actual results nor errors. | |
| 221 SpeechRecognitionResults empty_results; | |
| 222 ExpectResultsReceived(empty_results); | |
| 223 | |
| 224 // Ensure everything is closed cleanly after the downstream is closed. | |
| 225 ASSERT_FALSE(engine_under_test_->IsRecognitionPending()); | |
| 226 EndMockRecognition(); | |
| 227 ASSERT_EQ(SPEECH_RECOGNITION_ERROR_NONE, error_); | |
| 228 ASSERT_EQ(0U, results_.size()); | |
| 229 } | |
| 230 | |
| 231 TEST_F(GoogleStreamingRemoteEngineTest, NoMatchError) { | |
| 232 StartMockRecognition(); | |
| 233 ASSERT_TRUE(GetUpstreamFetcher()); | |
| 234 ASSERT_EQ(0U, UpstreamChunksUploadedFromLastCall()); | |
| 235 | |
| 236 for (int i = 0; i < 3; ++i) | |
| 237 InjectDummyAudioChunk(); | |
| 238 engine_under_test_->AudioChunksEnded(); | |
| 239 ASSERT_EQ(4U, UpstreamChunksUploadedFromLastCall()); | |
| 240 ASSERT_TRUE(engine_under_test_->IsRecognitionPending()); | |
| 241 | |
| 242 // Simulate only a provisional result. | |
| 243 SpeechRecognitionResults results; | |
| 244 results.push_back(SpeechRecognitionResult()); | |
| 245 SpeechRecognitionResult& result = results.back(); | |
| 246 result.is_provisional = true; | |
| 247 result.hypotheses.push_back( | |
| 248 SpeechRecognitionHypothesis(base::UTF8ToUTF16("The final result"), 0.0F)); | |
| 249 ProvideMockResultDownstream(result); | |
| 250 ExpectResultsReceived(results); | |
| 251 ASSERT_TRUE(engine_under_test_->IsRecognitionPending()); | |
| 252 | |
| 253 CloseMockDownstream(DOWNSTREAM_ERROR_WEBSERVICE_NO_MATCH); | |
| 254 | |
| 255 // Expect an empty result. | |
| 256 ASSERT_FALSE(engine_under_test_->IsRecognitionPending()); | |
| 257 EndMockRecognition(); | |
| 258 SpeechRecognitionResults empty_result; | |
| 259 ExpectResultsReceived(empty_result); | |
| 260 } | |
| 261 | |
| 262 TEST_F(GoogleStreamingRemoteEngineTest, HTTPError) { | |
| 263 StartMockRecognition(); | |
| 264 ASSERT_TRUE(GetUpstreamFetcher()); | |
| 265 ASSERT_EQ(0U, UpstreamChunksUploadedFromLastCall()); | |
| 266 | |
| 267 InjectDummyAudioChunk(); | |
| 268 ASSERT_EQ(1U, UpstreamChunksUploadedFromLastCall()); | |
| 269 | |
| 270 // Close the downstream with a HTTP 500 error. | |
| 271 CloseMockDownstream(DOWNSTREAM_ERROR_HTTP500); | |
| 272 | |
| 273 // Expect a SPEECH_RECOGNITION_ERROR_NETWORK error to be raised. | |
| 274 ASSERT_FALSE(engine_under_test_->IsRecognitionPending()); | |
| 275 EndMockRecognition(); | |
| 276 ASSERT_EQ(SPEECH_RECOGNITION_ERROR_NETWORK, error_); | |
| 277 ASSERT_EQ(0U, results_.size()); | |
| 278 } | |
| 279 | |
| 280 TEST_F(GoogleStreamingRemoteEngineTest, NetworkError) { | |
| 281 StartMockRecognition(); | |
| 282 ASSERT_TRUE(GetUpstreamFetcher()); | |
| 283 ASSERT_EQ(0U, UpstreamChunksUploadedFromLastCall()); | |
| 284 | |
| 285 InjectDummyAudioChunk(); | |
| 286 ASSERT_EQ(1U, UpstreamChunksUploadedFromLastCall()); | |
| 287 | |
| 288 // Close the downstream fetcher simulating a network failure. | |
| 289 CloseMockDownstream(DOWNSTREAM_ERROR_NETWORK); | |
| 290 | |
| 291 // Expect a SPEECH_RECOGNITION_ERROR_NETWORK error to be raised. | |
| 292 ASSERT_FALSE(engine_under_test_->IsRecognitionPending()); | |
| 293 EndMockRecognition(); | |
| 294 ASSERT_EQ(SPEECH_RECOGNITION_ERROR_NETWORK, error_); | |
| 295 ASSERT_EQ(0U, results_.size()); | |
| 296 } | |
| 297 | |
| 298 TEST_F(GoogleStreamingRemoteEngineTest, Stability) { | |
| 299 StartMockRecognition(); | |
| 300 ASSERT_TRUE(GetUpstreamFetcher()); | |
| 301 ASSERT_EQ(0U, UpstreamChunksUploadedFromLastCall()); | |
| 302 | |
| 303 // Upload a dummy audio chunk. | |
| 304 InjectDummyAudioChunk(); | |
| 305 ASSERT_EQ(1U, UpstreamChunksUploadedFromLastCall()); | |
| 306 engine_under_test_->AudioChunksEnded(); | |
| 307 | |
| 308 // Simulate a protobuf message with an intermediate result without confidence, | |
| 309 // but with stability. | |
| 310 proto::SpeechRecognitionEvent proto_event; | |
| 311 proto_event.set_status(proto::SpeechRecognitionEvent::STATUS_SUCCESS); | |
| 312 proto::SpeechRecognitionResult* proto_result = proto_event.add_result(); | |
| 313 proto_result->set_stability(0.5); | |
| 314 proto::SpeechRecognitionAlternative *proto_alternative = | |
| 315 proto_result->add_alternative(); | |
| 316 proto_alternative->set_transcript("foo"); | |
| 317 ProvideMockProtoResultDownstream(proto_event); | |
| 318 | |
| 319 // Set up expectations. | |
| 320 SpeechRecognitionResults results; | |
| 321 results.push_back(SpeechRecognitionResult()); | |
| 322 SpeechRecognitionResult& result = results.back(); | |
| 323 result.is_provisional = true; | |
| 324 result.hypotheses.push_back( | |
| 325 SpeechRecognitionHypothesis(base::UTF8ToUTF16("foo"), 0.5)); | |
| 326 | |
| 327 // Check that the protobuf generated the expected result. | |
| 328 ExpectResultsReceived(results); | |
| 329 | |
| 330 // Since it was a provisional result, recognition is still pending. | |
| 331 ASSERT_TRUE(engine_under_test_->IsRecognitionPending()); | |
| 332 | |
| 333 // Shut down. | |
| 334 CloseMockDownstream(DOWNSTREAM_ERROR_NONE); | |
| 335 ASSERT_FALSE(engine_under_test_->IsRecognitionPending()); | |
| 336 EndMockRecognition(); | |
| 337 | |
| 338 // Since there was no final result, we get an empty "no match" result. | |
| 339 SpeechRecognitionResults empty_result; | |
| 340 ExpectResultsReceived(empty_result); | |
| 341 ASSERT_EQ(SPEECH_RECOGNITION_ERROR_NONE, error_); | |
| 342 ASSERT_EQ(0U, results_.size()); | |
| 343 } | |
| 344 | |
| 345 TEST_F(GoogleStreamingRemoteEngineTest, EndOfUtterance) { | |
| 346 StartMockRecognition(); | |
| 347 ASSERT_TRUE(GetUpstreamFetcher()); | |
| 348 | |
| 349 // Simulate a END_OF_UTTERANCE proto event with continuous true. | |
| 350 SpeechRecognitionEngine::Config config; | |
| 351 config.continuous = true; | |
| 352 engine_under_test_->SetConfig(config); | |
| 353 proto::SpeechRecognitionEvent proto_event; | |
| 354 proto_event.set_endpoint(proto::SpeechRecognitionEvent::END_OF_UTTERANCE); | |
| 355 ASSERT_EQ(0, end_of_utterance_counter_); | |
| 356 ProvideMockProtoResultDownstream(proto_event); | |
| 357 ASSERT_EQ(0, end_of_utterance_counter_); | |
| 358 | |
| 359 // Simulate a END_OF_UTTERANCE proto event with continuous false. | |
| 360 config.continuous = false; | |
| 361 engine_under_test_->SetConfig(config); | |
| 362 ProvideMockProtoResultDownstream(proto_event); | |
| 363 ASSERT_EQ(1, end_of_utterance_counter_); | |
| 364 | |
| 365 // Shut down. | |
| 366 CloseMockDownstream(DOWNSTREAM_ERROR_NONE); | |
| 367 EndMockRecognition(); | |
| 368 } | |
| 369 | |
| 370 TEST_F(GoogleStreamingRemoteEngineTest, SendPreamble) { | |
| 371 const size_t kPreambleLength = 100; | |
| 372 scoped_refptr<SpeechRecognitionSessionPreamble> preamble = | |
| 373 new SpeechRecognitionSessionPreamble(); | |
| 374 preamble->sample_rate = 16000; | |
| 375 preamble->sample_depth = 2; | |
| 376 preamble->sample_data.assign(kPreambleLength, 0); | |
| 377 SpeechRecognitionEngine::Config config; | |
| 378 config.auth_token = "foo"; | |
| 379 config.auth_scope = "bar"; | |
| 380 config.preamble = preamble; | |
| 381 engine_under_test_->SetConfig(config); | |
| 382 | |
| 383 StartMockRecognition(); | |
| 384 ASSERT_TRUE(GetUpstreamFetcher()); | |
| 385 // First chunk uploaded should be the preamble. | |
| 386 ASSERT_EQ(1U, UpstreamChunksUploadedFromLastCall()); | |
| 387 std::string chunk = LastUpstreamChunkUploaded(); | |
| 388 ExpectFramedChunk(chunk, kFrameTypePreamble); | |
| 389 | |
| 390 for (int i = 0; i < 3; ++i) { | |
| 391 InjectDummyAudioChunk(); | |
| 392 ASSERT_EQ(1U, UpstreamChunksUploadedFromLastCall()); | |
| 393 chunk = LastUpstreamChunkUploaded(); | |
| 394 ExpectFramedChunk(chunk, kFrameTypeRecognitionAudio); | |
| 395 } | |
| 396 engine_under_test_->AudioChunksEnded(); | |
| 397 ASSERT_TRUE(engine_under_test_->IsRecognitionPending()); | |
| 398 | |
| 399 // Simulate a protobuf message streamed from the server containing a single | |
| 400 // result with one hypotheses. | |
| 401 SpeechRecognitionResults results; | |
| 402 results.push_back(SpeechRecognitionResult()); | |
| 403 SpeechRecognitionResult& result = results.back(); | |
| 404 result.is_provisional = false; | |
| 405 result.hypotheses.push_back( | |
| 406 SpeechRecognitionHypothesis(base::UTF8ToUTF16("hypothesis 1"), 0.1F)); | |
| 407 | |
| 408 ProvideMockResultDownstream(result); | |
| 409 ExpectResultsReceived(results); | |
| 410 ASSERT_TRUE(engine_under_test_->IsRecognitionPending()); | |
| 411 | |
| 412 // Ensure everything is closed cleanly after the downstream is closed. | |
| 413 CloseMockDownstream(DOWNSTREAM_ERROR_NONE); | |
| 414 ASSERT_FALSE(engine_under_test_->IsRecognitionPending()); | |
| 415 EndMockRecognition(); | |
| 416 ASSERT_EQ(SPEECH_RECOGNITION_ERROR_NONE, error_); | |
| 417 ASSERT_EQ(0U, results_.size()); | |
| 418 } | |
| 419 | |
| 420 void GoogleStreamingRemoteEngineTest::SetUp() { | |
| 421 engine_under_test_.reset( | |
| 422 new GoogleStreamingRemoteEngine(NULL /*URLRequestContextGetter*/)); | |
| 423 engine_under_test_->set_delegate(this); | |
| 424 } | |
| 425 | |
| 426 void GoogleStreamingRemoteEngineTest::TearDown() { | |
| 427 engine_under_test_.reset(); | |
| 428 } | |
| 429 | |
| 430 TestURLFetcher* GoogleStreamingRemoteEngineTest::GetUpstreamFetcher() { | |
| 431 return url_fetcher_factory_.GetFetcherByID( | |
| 432 GoogleStreamingRemoteEngine::kUpstreamUrlFetcherIdForTesting); | |
| 433 } | |
| 434 | |
| 435 TestURLFetcher* GoogleStreamingRemoteEngineTest::GetDownstreamFetcher() { | |
| 436 return url_fetcher_factory_.GetFetcherByID( | |
| 437 GoogleStreamingRemoteEngine::kDownstreamUrlFetcherIdForTesting); | |
| 438 } | |
| 439 | |
| 440 // Starts recognition on the engine, ensuring that both stream fetchers are | |
| 441 // created. | |
| 442 void GoogleStreamingRemoteEngineTest::StartMockRecognition() { | |
| 443 DCHECK(engine_under_test_.get()); | |
| 444 | |
| 445 ASSERT_FALSE(engine_under_test_->IsRecognitionPending()); | |
| 446 | |
| 447 engine_under_test_->StartRecognition(); | |
| 448 ASSERT_TRUE(engine_under_test_->IsRecognitionPending()); | |
| 449 | |
| 450 TestURLFetcher* upstream_fetcher = GetUpstreamFetcher(); | |
| 451 ASSERT_TRUE(upstream_fetcher); | |
| 452 upstream_fetcher->set_url(upstream_fetcher->GetOriginalURL()); | |
| 453 | |
| 454 TestURLFetcher* downstream_fetcher = GetDownstreamFetcher(); | |
| 455 ASSERT_TRUE(downstream_fetcher); | |
| 456 downstream_fetcher->set_url(downstream_fetcher->GetOriginalURL()); | |
| 457 } | |
| 458 | |
| 459 void GoogleStreamingRemoteEngineTest::EndMockRecognition() { | |
| 460 DCHECK(engine_under_test_.get()); | |
| 461 engine_under_test_->EndRecognition(); | |
| 462 ASSERT_FALSE(engine_under_test_->IsRecognitionPending()); | |
| 463 | |
| 464 // TODO(primiano): In order to be very pedantic we should check that both the | |
| 465 // upstream and downstream URL fetchers have been disposed at this time. | |
| 466 // Unfortunately it seems that there is no direct way to detect (in tests) | |
| 467 // if a url_fetcher has been freed or not, since they are not automatically | |
| 468 // de-registered from the TestURLFetcherFactory on destruction. | |
| 469 } | |
| 470 | |
| 471 void GoogleStreamingRemoteEngineTest::InjectDummyAudioChunk() { | |
| 472 unsigned char dummy_audio_buffer_data[2] = {'\0', '\0'}; | |
| 473 scoped_refptr<AudioChunk> dummy_audio_chunk( | |
| 474 new AudioChunk(&dummy_audio_buffer_data[0], | |
| 475 sizeof(dummy_audio_buffer_data), | |
| 476 2 /* bytes per sample */)); | |
| 477 DCHECK(engine_under_test_.get()); | |
| 478 engine_under_test_->TakeAudioChunk(*dummy_audio_chunk.get()); | |
| 479 } | |
| 480 | |
| 481 size_t GoogleStreamingRemoteEngineTest::UpstreamChunksUploadedFromLastCall() { | |
| 482 TestURLFetcher* upstream_fetcher = GetUpstreamFetcher(); | |
| 483 DCHECK(upstream_fetcher); | |
| 484 const size_t number_of_chunks = upstream_fetcher->upload_chunks().size(); | |
| 485 DCHECK_GE(number_of_chunks, last_number_of_upstream_chunks_seen_); | |
| 486 const size_t new_chunks = number_of_chunks - | |
| 487 last_number_of_upstream_chunks_seen_; | |
| 488 last_number_of_upstream_chunks_seen_ = number_of_chunks; | |
| 489 return new_chunks; | |
| 490 } | |
| 491 | |
| 492 std::string GoogleStreamingRemoteEngineTest::LastUpstreamChunkUploaded() { | |
| 493 TestURLFetcher* upstream_fetcher = GetUpstreamFetcher(); | |
| 494 DCHECK(upstream_fetcher); | |
| 495 DCHECK(!upstream_fetcher->upload_chunks().empty()); | |
| 496 return upstream_fetcher->upload_chunks().back(); | |
| 497 } | |
| 498 | |
| 499 void GoogleStreamingRemoteEngineTest::ProvideMockProtoResultDownstream( | |
| 500 const proto::SpeechRecognitionEvent& result) { | |
| 501 TestURLFetcher* downstream_fetcher = GetDownstreamFetcher(); | |
| 502 | |
| 503 ASSERT_TRUE(downstream_fetcher); | |
| 504 downstream_fetcher->set_status(URLRequestStatus(/* default=SUCCESS */)); | |
| 505 downstream_fetcher->set_response_code(200); | |
| 506 | |
| 507 std::string response_string = SerializeProtobufResponse(result); | |
| 508 response_buffer_.append(response_string); | |
| 509 downstream_fetcher->SetResponseString(response_buffer_); | |
| 510 downstream_fetcher->delegate()->OnURLFetchDownloadProgress( | |
| 511 downstream_fetcher, | |
| 512 response_buffer_.size(), | |
| 513 -1 /* total response length not used */); | |
| 514 } | |
| 515 | |
| 516 void GoogleStreamingRemoteEngineTest::ProvideMockResultDownstream( | |
| 517 const SpeechRecognitionResult& result) { | |
| 518 proto::SpeechRecognitionEvent proto_event; | |
| 519 proto_event.set_status(proto::SpeechRecognitionEvent::STATUS_SUCCESS); | |
| 520 proto::SpeechRecognitionResult* proto_result = proto_event.add_result(); | |
| 521 proto_result->set_final(!result.is_provisional); | |
| 522 for (size_t i = 0; i < result.hypotheses.size(); ++i) { | |
| 523 proto::SpeechRecognitionAlternative* proto_alternative = | |
| 524 proto_result->add_alternative(); | |
| 525 const SpeechRecognitionHypothesis& hypothesis = result.hypotheses[i]; | |
| 526 proto_alternative->set_confidence(hypothesis.confidence); | |
| 527 proto_alternative->set_transcript(base::UTF16ToUTF8(hypothesis.utterance)); | |
| 528 } | |
| 529 ProvideMockProtoResultDownstream(proto_event); | |
| 530 } | |
| 531 | |
| 532 void GoogleStreamingRemoteEngineTest::CloseMockDownstream( | |
| 533 DownstreamError error) { | |
| 534 TestURLFetcher* downstream_fetcher = GetDownstreamFetcher(); | |
| 535 ASSERT_TRUE(downstream_fetcher); | |
| 536 | |
| 537 const net::Error net_error = | |
| 538 (error == DOWNSTREAM_ERROR_NETWORK) ? net::ERR_FAILED : net::OK; | |
| 539 downstream_fetcher->set_status(URLRequestStatus::FromError(net_error)); | |
| 540 downstream_fetcher->set_response_code( | |
| 541 (error == DOWNSTREAM_ERROR_HTTP500) ? 500 : 200); | |
| 542 | |
| 543 if (error == DOWNSTREAM_ERROR_WEBSERVICE_NO_MATCH) { | |
| 544 // Send empty response. | |
| 545 proto::SpeechRecognitionEvent response; | |
| 546 response_buffer_.append(SerializeProtobufResponse(response)); | |
| 547 } | |
| 548 downstream_fetcher->SetResponseString(response_buffer_); | |
| 549 downstream_fetcher->delegate()->OnURLFetchComplete(downstream_fetcher); | |
| 550 } | |
| 551 | |
| 552 void GoogleStreamingRemoteEngineTest::ExpectResultsReceived( | |
| 553 const SpeechRecognitionResults& results) { | |
| 554 ASSERT_GE(1U, results_.size()); | |
| 555 ASSERT_TRUE(ResultsAreEqual(results, results_.front())); | |
| 556 results_.pop(); | |
| 557 } | |
| 558 | |
| 559 bool GoogleStreamingRemoteEngineTest::ResultsAreEqual( | |
| 560 const SpeechRecognitionResults& a, const SpeechRecognitionResults& b) { | |
| 561 if (a.size() != b.size()) | |
| 562 return false; | |
| 563 | |
| 564 SpeechRecognitionResults::const_iterator it_a = a.begin(); | |
| 565 SpeechRecognitionResults::const_iterator it_b = b.begin(); | |
| 566 for (; it_a != a.end() && it_b != b.end(); ++it_a, ++it_b) { | |
| 567 if (it_a->is_provisional != it_b->is_provisional || | |
| 568 it_a->hypotheses.size() != it_b->hypotheses.size()) { | |
| 569 return false; | |
| 570 } | |
| 571 for (size_t i = 0; i < it_a->hypotheses.size(); ++i) { | |
| 572 const SpeechRecognitionHypothesis& hyp_a = it_a->hypotheses[i]; | |
| 573 const SpeechRecognitionHypothesis& hyp_b = it_b->hypotheses[i]; | |
| 574 if (hyp_a.utterance != hyp_b.utterance || | |
| 575 hyp_a.confidence != hyp_b.confidence) { | |
| 576 return false; | |
| 577 } | |
| 578 } | |
| 579 } | |
| 580 | |
| 581 return true; | |
| 582 } | |
| 583 | |
| 584 void GoogleStreamingRemoteEngineTest::ExpectFramedChunk( | |
| 585 const std::string& chunk, uint32_t type) { | |
| 586 uint32_t value; | |
| 587 base::ReadBigEndian(&chunk[0], &value); | |
| 588 EXPECT_EQ(chunk.size() - 8, value); | |
| 589 base::ReadBigEndian(&chunk[4], &value); | |
| 590 EXPECT_EQ(type, value); | |
| 591 } | |
| 592 | |
| 593 std::string GoogleStreamingRemoteEngineTest::SerializeProtobufResponse( | |
| 594 const proto::SpeechRecognitionEvent& msg) { | |
| 595 std::string msg_string; | |
| 596 msg.SerializeToString(&msg_string); | |
| 597 | |
| 598 // Prepend 4 byte prefix length indication to the protobuf message as | |
| 599 // envisaged by the google streaming recognition webservice protocol. | |
| 600 uint32_t prefix = HostToNet32(checked_cast<uint32_t>(msg_string.size())); | |
| 601 msg_string.insert(0, reinterpret_cast<char*>(&prefix), sizeof(prefix)); | |
| 602 | |
| 603 return msg_string; | |
| 604 } | |
| 605 | |
| 606 } // namespace content | |
| OLD | NEW |
|---|
Chromium Code Reviews
Issue 1891543002:
Devirtualize SpeechRecognitionEngine (Closed)
Base URL: https://chromium.googlesource.com/chromium/src.git@kill_one_shot_engine