google-cloud-speech 0.23.0 → 0.24.0

data/lib/google/cloud/speech/result.rb

@@ -13,7 +13,7 @@
 # limitations under the License.
 
 
-require "google/cloud/speech/v1beta1"
+require "google/cloud/speech/v1"
 
 module Google
   module Cloud
@@ -23,9 +23,9 @@ module Google
       #
       # A speech recognition result corresponding to a portion of the audio.
       #
-      # See {Project#recognize} and {Job#results}.
+      # See {Project#recognize} and {Operation#results}.
       #
-      # @see https://cloud.google.com/speech/reference/rpc/google.cloud.speech.v1beta1#google.cloud.speech.v1beta1.SpeechRecognitionResult
+      # @see https://cloud.google.com/speech/reference/rpc/google.cloud.speech.V1#google.cloud.speech.V1.SpeechRecognitionResult
       #   SpeechRecognitionResult
       #
       # @attr_reader [String] transcript Transcript text representing the words
@@ -46,7 +46,9 @@ module Google
       #   speech = Google::Cloud::Speech.new
       #
       #   audio = speech.audio "path/to/audio.raw",
-      #                        encoding: :raw, sample_rate: 16000
+      #                        encoding: :raw,
+      #                        language: "en-US",
+      #                        sample_rate: 16000
       #   results = audio.recognize
       #
       #   result = results.first
@@ -94,7 +96,9 @@ module Google
         #   speech = Google::Cloud::Speech.new
         #
         #   audio = speech.audio "path/to/audio.raw",
-        #                        encoding: :raw, sample_rate: 16000
+        #                        encoding: :raw,
+        #                        language: "en-US",
+        #                        sample_rate: 16000
         #   results = audio.recognize
         #
         #   result = results.first
@@ -124,9 +128,9 @@ module Google
       #
       # See {Project#stream} and {Stream#on_interim}.
       #
-      # @see https://cloud.google.com/speech/reference/rpc/google.cloud.speech.v1beta1#google.cloud.speech.v1beta1.SpeechRecognitionResult
+      # @see https://cloud.google.com/speech/reference/rpc/google.cloud.speech.V1#google.cloud.speech.V1.SpeechRecognitionResult
       #   SpeechRecognitionResult
-      # @see https://cloud.google.com/speech/reference/rpc/google.cloud.speech.v1beta1#google.cloud.speech.v1beta1.StreamingRecognitionResult
+      # @see https://cloud.google.com/speech/reference/rpc/google.cloud.speech.V1#google.cloud.speech.V1.StreamingRecognitionResult
       #   StreamingRecognitionResult
       #
       # @attr_reader [String] transcript Transcript text representing the words
@@ -150,7 +154,9 @@ module Google
       #
       #   speech = Google::Cloud::Speech.new
       #
-      #   stream = speech.stream encoding: :raw, sample_rate: 16000
+      #   stream = speech.stream encoding: :raw,
+      #                          language: "en-US",
+      #                          sample_rate: 16000
       #
       #   # register callback for when an interim result is returned
       #   stream.on_interim do |final_results, interim_results|

data/lib/google/cloud/speech/service.rb

@@ -16,7 +16,7 @@
 require "google/cloud/errors"
 require "google/cloud/speech/credentials"
 require "google/cloud/speech/version"
-require "google/cloud/speech/v1beta1"
+require "google/cloud/speech/v1"
 
 module Google
   module Cloud
@@ -33,7 +33,7 @@ module Google
                        client_config: nil
           @project = project
           @credentials = credentials
-          @host = host || V1beta1::SpeechClient::SERVICE_ADDRESS
+          @host = host || V1::SpeechClient::SERVICE_ADDRESS
           @timeout = timeout
           @client_config = client_config || {}
         end
@@ -53,7 +53,7 @@ module Google
         def service
           return mocked_service if mocked_service
           @service ||= \
-            V1beta1::SpeechClient.new(
+            V1::SpeechClient.new(
               service_path: host,
               channel: channel,
               timeout: timeout,
@@ -82,13 +82,14 @@ module Google
 
         def recognize_sync audio, config
           execute do
-            service.sync_recognize config, audio, options: default_options
+            service.recognize config, audio, options: default_options
           end
         end
 
         def recognize_async audio, config
           execute do
-            service.async_recognize config, audio, options: default_options
+            service.long_running_recognize \
+              config, audio, options: default_options
           end
         end
 
@@ -98,7 +99,12 @@ module Google
         end
 
         def get_op name
-          execute { ops.get_operation name }
+          execute do
+            Google::Gax::Operation.new \
+              ops.get_operation(name), ops,
+              V1::LongRunningRecognizeResponse,
+              V1::LongRunningRecognizeMetadata
+          end
         end
 
         def inspect

data/lib/google/cloud/speech/stream.rb

@@ -13,7 +13,7 @@
 # limitations under the License.
 
 
-require "google/cloud/speech/v1beta1"
+require "google/cloud/speech/v1"
 require "google/cloud/speech/result"
 require "monitor"
 require "forwardable"
@@ -31,14 +31,9 @@ module Google
       #
       #   speech = Google::Cloud::Speech.new
       #
-      #   stream = speech.stream encoding: :raw, sample_rate: 16000
-      #
-      #   # register callback for when a result is returned
-      #   stream.on_result do |results|
-      #     result = results.first
-      #     puts result.transcript # "how old is the Brooklyn Bridge"
-      #     puts result.confidence # 0.9826789498329163
-      #   end
+      #   stream = speech.stream encoding: :raw,
+      #                          language: "en-US",
+      #                          sample_rate: 16000
       #
       #   # Stream 5 seconds of audio from the microphone
       #   # Actual implementation of microphone input varies by platform
@@ -47,6 +42,12 @@ module Google
       #   end
       #
       #   stream.stop
+      #   stream.wait_until_complete!
+      #
+      #   results = stream.results
+      #   result = results.first
+      #   result.transcript #=> "how old is the Brooklyn Bridge"
+      #   result.confidence #=> 0.9826789498329163
       #
       class Stream
         include MonitorMixin
@@ -95,14 +96,9 @@ module Google
         #
         #   audio = speech.audio "path/to/audio.raw"
         #
-        #   stream = speech.stream encoding: :raw, sample_rate: 16000
-        #
-        #   # register callback for when a result is returned
-        #   stream.on_result do |results|
-        #     result = results.first
-        #     puts result.transcript # "how old is the Brooklyn Bridge"
-        #     puts result.confidence # 0.9826789498329163
-        #   end
+        #   stream = speech.stream encoding: :raw,
+        #                          language: "en-US",
+        #                          sample_rate: 16000
         #
         #   # Stream 5 seconds of audio from the microphone
         #   # Actual implementation of microphone input varies by platform
@@ -111,12 +107,18 @@ module Google
         #   end
         #
         #   stream.stop
+        #   stream.wait_until_complete!
+        #
+        #   results = stream.results
+        #   result = results.first
+        #   result.transcript #=> "how old is the Brooklyn Bridge"
+        #   result.confidence #=> 0.9826789498329163
         #
         def send bytes
           start # lazily call start if the stream wasn't started yet
           # TODO: do not send if stopped?
           synchronize do
-            req = V1beta1::StreamingRecognizeRequest.new(
+            req = V1::StreamingRecognizeRequest.new(
               audio_content: bytes.encode("ASCII-8BIT"))
             @request_queue.push req
           end
@@ -153,7 +155,9 @@ module Google
         #
         #   speech = Google::Cloud::Speech.new
         #
-        #   stream = speech.stream encoding: :raw, sample_rate: 16000
+        #   stream = speech.stream encoding: :raw,
+        #                          language: "en-US",
+        #                          sample_rate: 16000
         #
         #   # Stream 5 seconds of audio from the microphone
         #   # Actual implementation of microphone input varies by platform
@@ -176,26 +180,18 @@ module Google
         end
 
         ##
-        # Register to be notified on the reception of an interim result.
+        # Whether all speech recognition results have been returned.
         #
-        # @yield [callback] The block for accessing final and interim results.
-        # @yieldparam [Array<Result>] final_results The final results.
-        # @yieldparam [Array<Result>] interim_results The interim results.
+        # @return [Boolean] All speech recognition results have been returned.
         #
         # @example
         #   require "google/cloud/speech"
         #
         #   speech = Google::Cloud::Speech.new
         #
-        #   stream = speech.stream encoding: :raw, sample_rate: 16000
-        #
-        #   # register callback for when an interim result is returned
-        #   stream.on_interim do |final_results, interim_results|
-        #     interim_result = interim_results.first
-        #     puts interim_result.transcript # "how old is the Brooklyn Bridge"
-        #     puts interim_result.confidence # 0.9826789498329163
-        #     puts interim_result.stability # 0.8999
-        #   end
+        #   stream = speech.stream encoding: :raw,
+        #                          language: "en-US",
+        #                          sample_rate: 16000
         #
         #   # Stream 5 seconds of audio from the microphone
         #   # Actual implementation of microphone input varies by platform
@@ -205,39 +201,32 @@ module Google
         #
         #   stream.stop
         #
-        def on_interim &block
-          synchronize do
-            @callbacks[:interim] << block
-          end
-        end
-
-        # @private yields two arguments, all final results and the
-        # non-final/incomplete result
-        def interim! interim_results
+        #   stream.wait_until_complete!
+        #   stream.complete? #=> true
+        #
+        #   results = stream.results
+        #   results.each do |result|
+        #     puts result.transcript
+        #     puts result.confidence
+        #   end
+        #
+        def complete?
           synchronize do
-            @callbacks[:interim].each { |c| c.call results, interim_results }
+            @complete
           end
         end
 
         ##
-        # Register to be notified on the reception of a final result.
-        #
-        # @yield [callback] The block for accessing final results.
-        # @yieldparam [Array<Result>] results The final results.
+        # Blocks until all speech recognition results have been returned.
         #
         # @example
         #   require "google/cloud/speech"
         #
         #   speech = Google::Cloud::Speech.new
         #
-        #   stream = speech.stream encoding: :raw, sample_rate: 16000
-        #
-        #   # register callback for when an interim result is returned
-        #   stream.on_result do |results|
-        #     result = results.first
-        #     puts result.transcript # "how old is the Brooklyn Bridge"
-        #     puts result.confidence # 0.9826789498329163
-        #   end
+        #   stream = speech.stream encoding: :raw,
+        #                          language: "en-US",
+        #                          sample_rate: 16000
         #
         #   # Stream 5 seconds of audio from the microphone
         #   # Actual implementation of microphone input varies by platform
@@ -247,45 +236,46 @@ module Google
         #
         #   stream.stop
         #
-        def on_result &block
-          synchronize do
-            @callbacks[:result] << block
-          end
-        end
-
-        # @private add a result object, and call the callbacks
-        def add_result!result_index, result_grpc
-          synchronize do
-            @results[result_index] = Result.from_grpc result_grpc
-          end
-          # callback for final result received
-          result!
-        end
-
-        # @private yields each final results as they are received
-        def result!
-          synchronize do
-            @callbacks[:result].each { |c| c.call results }
+        #   stream.wait_until_complete!
+        #   stream.complete? #=> true
+        #
+        #   results = stream.results
+        #   results.each do |result|
+        #     puts result.transcript
+        #     puts result.confidence
+        #   end
+        #
+        def wait_until_complete!
+          complete_check = nil
+          synchronize { complete_check = @complete }
+          while complete_check.nil?
+            sleep 1
+            synchronize { complete_check = @complete }
           end
         end
 
         ##
-        # Register to be notified when speech has been detected in the audio
-        # stream.
+        # Register to be notified on the reception of an interim result.
         #
-        # @yield [callback] The block to be called when speech has been detected
-        #   in the audio stream.
+        # @yield [callback] The block for accessing final and interim results.
+        # @yieldparam [Array<Result>] final_results The final results.
+        # @yieldparam [Array<Result>] interim_results The interim results.
         #
         # @example
         #   require "google/cloud/speech"
         #
         #   speech = Google::Cloud::Speech.new
         #
-        #   stream = speech.stream encoding: :raw, sample_rate: 16000
+        #   stream = speech.stream encoding: :raw,
+        #                          language: "en-US",
+        #                          sample_rate: 16000
         #
-        #   # register callback for when speech has started.
-        #   stream.on_speech_start do
-        #     puts "Speech has started."
+        #   # register callback for when an interim result is returned
+        #   stream.on_interim do |final_results, interim_results|
+        #     interim_result = interim_results.first
+        #     puts interim_result.transcript # "how old is the Brooklyn Bridge"
+        #     puts interim_result.confidence # 0.9826789498329163
+        #     puts interim_result.stability # 0.8999
         #   end
         #
         #   # Stream 5 seconds of audio from the microphone
@@ -296,38 +286,35 @@ module Google
         #
         #   stream.stop
         #
-        def on_speech_start &block
+        def on_interim &block
           synchronize do
-            @callbacks[:speech_start] << block
+            @callbacks[:interim] << block
           end
         end
 
-        # @private returns single final result once :END_OF_UTTERANCE is
-        # received.
-        def speech_start!
+        ##
+        # @private yields two arguments, all final results and the
+        # non-final/incomplete result
+        def pass_interim! interim_results
           synchronize do
-            @callbacks[:speech_start].each(&:call)
+            @callbacks[:interim].each { |c| c.call results, interim_results }
           end
         end
 
         ##
-        # Register to be notified when speech has ceased to be detected in the
-        # audio stream.
+        # Register to be notified on the reception of a final result.
         #
-        # @yield [callback] The block to be called when speech has ceased to be
-        #   detected in the audio stream.
+        # @yield [callback] The block for accessing final results.
+        # @yieldparam [Array<Result>] results The final results.
         #
         # @example
         #   require "google/cloud/speech"
         #
         #   speech = Google::Cloud::Speech.new
         #
-        #   stream = speech.stream encoding: :raw, sample_rate: 16000
-        #
-        #   # register callback for when speech has ended.
-        #   stream.on_speech_end do
-        #     puts "Speech has ended."
-        #   end
+        #   stream = speech.stream encoding: :raw,
+        #                          language: "en-US",
+        #                          sample_rate: 16000
         #
         #   # Stream 5 seconds of audio from the microphone
         #   # Actual implementation of microphone input varies by platform
@@ -336,18 +323,25 @@ module Google
         #   end
         #
         #   stream.stop
+        #   stream.wait_until_complete!
         #
-        def on_speech_end &block
+        #   results = stream.results
+        #   result = results.first
+        #   result.transcript #=> "how old is the Brooklyn Bridge"
+        #   result.confidence #=> 0.9826789498329163
+        #
+        def on_result &block
           synchronize do
-            @callbacks[:speech_end] << block
+            @callbacks[:result] << block
           end
         end
 
-        # @private yields single final result once :END_OF_UTTERANCE is
-        # received.
-        def speech_end!
+        ##
+        # @private add a result object, and call the callbacks
+        def pass_result! result_grpc
           synchronize do
-            @callbacks[:speech_end].each(&:call)
+            @results << Result.from_grpc(result_grpc)
+            @callbacks[:result].each { |c| c.call @results }
           end
         end
 
@@ -363,11 +357,13 @@ module Google
         #
         #   speech = Google::Cloud::Speech.new
         #
-        #   stream = speech.stream encoding: :raw, sample_rate: 16000
+        #   stream = speech.stream encoding: :raw,
+        #                          language: "en-US",
+        #                          sample_rate: 16000
         #
-        #   # register callback for when audio has ended.
+        #   # register callback for when stream has ended.
         #   stream.on_complete do
-        #     puts "Audio has ended."
+        #     puts "Stream has ended."
         #   end
         #
         #   # Stream 5 seconds of audio from the microphone
@@ -384,11 +380,11 @@ module Google
           end
         end
 
-        # @private yields all final results once the recognition is completed
-        # depending on how the Stream is configured, this can be on the
-        # reception of :END_OF_AUDIO or :END_OF_UTTERANCE.
-        def complete!
+        ##
+        # @private yields when the end of the audio stream has been reached.
+        def pass_complete!
           synchronize do
+            @complete = true
             @callbacks[:complete].each(&:call)
           end
         end
@@ -409,6 +405,7 @@ module Google
         #   speech = Google::Cloud::Speech.new
         #
         #   stream = speech.stream encoding: :raw,
+        #                          language: "en-US",
         #                          sample_rate: 16000,
         #                          utterance: true
         #
@@ -432,9 +429,10 @@ module Google
           end
         end
 
-        # @private returns single final result once :END_OF_UTTERANCE is
+        ##
+        # @private returns single final result once :END_OF_SINGLE_UTTERANCE is
         # received.
-        def utterance!
+        def pass_utterance!
           synchronize do
             @callbacks[:utterance].each(&:call)
           end
@@ -451,7 +449,9 @@ module Google
         #
         #   speech = Google::Cloud::Speech.new
         #
-        #   stream = speech.stream encoding: :raw, sample_rate: 16000
+        #   stream = speech.stream encoding: :raw,
+        #                          language: "en-US",
+        #                          sample_rate: 16000
         #
         #   # register callback for when an error is returned
         #   stream.on_error do |error|
@@ -487,12 +487,16 @@ module Google
           response_enum.each do |response|
             begin
               background_results response
-              background_endpointer response.endpointer_type
+              background_event_type response.speech_event_type
               background_error response.error
             rescue => e
               error! Google::Cloud::Error.from_error(e)
             end
           end
+        rescue => e
+          error! Google::Cloud::Error.from_error(e)
+        ensure
+          pass_complete!
           Thread.pass
         end
 
@@ -501,10 +505,9 @@ module Google
           return unless response.results && response.results.any?
 
           final_grpc, interim_grpcs = *response.results
-          if final_grpc && final_grpc.is_final
-            add_result! response.result_index, final_grpc
-          else
+          unless final_grpc && final_grpc.is_final
             # all results are interim
+            final_grpc = nil
             interim_grpcs = response.results
           end
 
@@ -512,23 +515,17 @@ module Google
           interim_results = Array(interim_grpcs).map do |grpc|
             InterimResult.from_grpc grpc
           end
+
           # callback for interim results received
-          interim! interim_results if interim_results.any?
+          pass_interim! interim_results if interim_results.any?
+          # callback for final results received, if any
+          pass_result! final_grpc if final_grpc
         end
 
-        def background_endpointer endpointer
-          # Handle the endpointer by raising events
-          if endpointer == :START_OF_SPEECH
-            speech_start!
-          elsif endpointer == :END_OF_SPEECH
-            speech_end!
-          elsif endpointer == :END_OF_AUDIO
-            # TODO: do we automatically call stop here?
-            complete!
-          elsif endpointer == :END_OF_UTTERANCE
-            # TODO: do we automatically call stop here?
-            utterance!
-          end
+        def background_event_type event_type
+          # Handle the event_type by raising events
+          # TODO: do we automatically call stop here?
+          pass_utterance! if event_type == :END_OF_SINGLE_UTTERANCE
         end
 
         def background_error error