serializer.py
models.py
view.py
serializer.py
class AnswerListSerializer(serializers.ModelSerializer): question_id = serializers.IntegerField(source='question.id', required=True) video_url = serializers.CharField(source='get_video_url', required=False) class Meta: model = Answer fields = ( 'id', 'question_id', 'body', 'video_url', 'rtmp_key', ) extra_kwargs = { 'id': { 'read_only': True, }, 'body': { 'required': False, }, 'rtmp_key': { 'required': False, }, } def validate(self, data): rtmp_key = data.get('rtmp_key', None) video_url = data.pop('get_video_url', None) content = data.get('content',None) check_type = data.get('check_type',None) check_data = data.get('check_data',None) if not all([content,check_data,check_type]): raise BadRequestException(msg="Missed content or check_type or check_data") if rtmp_key is None and video_url is None: raise BadRequestException(msg="Missed rtmp_key or video_url") if rtmp_key is not None and video_url is not None: raise BadRequestException(msg="Can not set rtmp_key and video_url at same time") question_id = int(data.pop('question')['id']) try: question = Question.objects.get( id=question_id, ) except Exception: raise BadRequestException(msg="Question object with %d not exists" % question_id) # raise serializers.ValidationError("question object with %d not exists" % question_id) # noqa data['question'] = question data['video_url'] = video_url return data def create(self, validated_data): create_data = { 'application': validated_data.get('application', None), 'user': validated_data.get('user', None), 'question': validated_data.get('question', None) } try: answer, created = Answer.objects.get_or_create(**create_data) except Exception as e: print str(e) raise DatabaseErrorException('Database error when create answer') video_url = validated_data.get('video_url', None) if video_url: answer.video_url = video_url body = validated_data.get('body', None) rtmp_key = validated_data.get('rtmp_key', None) content = validated_data.get('content', None), check_type = validated_data.get('check_type', None), check_data = validated_data.get('check_data', None) if rtmp_key: answer.rtmp_key = rtmp_key if body: answer.body = body if content: answer.content = content if check_type: answer.check_type = check_type if check_data: answer.check_data = check_data answer.save() return answer
models.py
class Answer(models.Model): user = models.ForeignKey( User, related_name='answers', null=True, ) application = models.ForeignKey( Application, related_name='answers', null=True, ) question = models.ForeignKey( Question, related_name='answers', ) created_date = models.DateTimeField( auto_now_add=True, ) update_date = models.DateTimeField( auto_now=True, ) body = models.CharField( max_length=10240, null=True, blank=True, ) video_url = models.CharField( max_length=1024, null=True, blank=True, ) unified_video_url = models.CharField( max_length=1024, null=True, blank=True, ) rtmp_key = models.CharField( max_length=256, null=True, blank=True, ) audio_url = models.CharField( max_length=1024, null=True, blank=True, ) ANSWER_RATING_SCORE_DEFAULT = -1 ANSWER_RATING_SCORE_ACCEPT_MIN = 0 ANSWER_RATING_SCORE_ACCEPT_MAX = 5 rating_score = models.IntegerField( default=ANSWER_RATING_SCORE_DEFAULT, validators=[MinValueValidator(ANSWER_RATING_SCORE_DEFAULT), MaxValueValidator(ANSWER_RATING_SCORE_ACCEPT_MAX)], ) rater = models.ForeignKey( User, related_name='rating_applications', null=True, blank=True, ) ANSWER_FRAME_STATUS_NOT_YET = 0 ANSWER_FRAME_STATUS_FAILED = 1 ANSWER_FRAME_STATUS_SUCCESS = 2 ANSWER_FRAME_STATUS_CHOICES = ( (ANSWER_FRAME_STATUS_NOT_YET, 'Not Yet'), (ANSWER_FRAME_STATUS_FAILED, 'Failed'), (ANSWER_FRAME_STATUS_SUCCESS, 'Success'), ) answer_frame_status = models.PositiveSmallIntegerField( default=ANSWER_FRAME_STATUS_NOT_YET, choices=ANSWER_FRAME_STATUS_CHOICES, ) content = models.TextField( verbose_name="test content", default="" ) APPLICATION_TYPE_SIMPLE = 0 APPLICATION_TYPE_MANY = 1 APPLICATION_TYPE_CHOICES = ( (APPLICATION_TYPE_SIMPLE, 'simple select'), (APPLICATION_TYPE_MANY, 'many select'), ) check_type = models.PositiveSmallIntegerField( choices=APPLICATION_TYPE_CHOICES, default=APPLICATION_TYPE_SIMPLE, ) check_data = models.CharField( max_length=10, verbose_name="选择结果", default="", blank=True ) def __unicode__(self): return '%s-%s-%s' % (self.user.username, self.question.title, self.application.position.title) @classmethod def is_a_valid_rating_score(_cls, rating_score): return rating_score in range(_cls.ANSWER_RATING_SCORE_ACCEPT_MIN, _cls.ANSWER_RATING_SCORE_ACCEPT_MAX + 1) class Meta: unique_together = ('user', 'application', 'question', ) def get_video_url(self): # use video_url as first priority # if user upload video via rtmp, rtmp playlist key will first store in "rtmp_key" field # then, rtmp playlist will be transcoded into mp4 format and store in "video_url" field if self.video_url is not None and self.video_url != '': return self.video_url if self.rtmp_key is not None and self.rtmp_key != '': url = '' try: channel, key = tuple(self.rtmp_key.split('/')) url = get_signed_playlist_url(channel, key) except Exception: pass return url return '' def get_mp4_video_url(self): if self.video_url is None or self.video_url == '': return None return self.video_url def get_rtmp_key(self): if self.rtmp_key is None or self.rtmp_key == '': return None return self.rtmp_key def get_author_obj(self): return self.application.user def get_question_id(self): return self.question.id def get_question_title(self): return self.question.title def get_question_competency_name(self): return self.question.get_competency_name() def get_latest_emotion_score(self): if self.emotion_scores.count() > 0: return self.emotion_scores.all().order_by('-created_date')[0] return None def get_emotion_score_overall(self): emotion_score = self.get_latest_emotion_score() if emotion_score: return emotion_score.overall return None def get_frame_score(self): emotion_score = self.get_latest_emotion_score() if emotion_score: return emotion_score.frame_score return {} def is_answer_frame_status_success(self): return self.answer_frame_status == Answer.ANSWER_FRAME_STATUS_SUCCESS def set_frame_status(self, frame_status): self.answer_frame_status = frame_status self.save() def set_frame_status_fail(self): return self.set_frame_status(Answer.ANSWER_FRAME_STATUS_FAILED) def set_frame_status_success(self): return self.set_frame_status(Answer.ANSWER_FRAME_STATUS_SUCCESS) def get_no_emotion_data_frames(self): if getattr(self, 'video_frames', None) is None: return None return self.video_frames.filter(azure_fer=None) def get_lang(self): return self.get_question().get_answer_lang() def get_question(self): return self.question # def can_emotion_score_be_calculated(self): # frames = self.video_frames.all() # for frame in frames: # if frame.azure_emotion_result is None: # return False # # if self.get_question_template_competency_emotion_factor is None: # return False # # return True def get_frames(self): return self.video_frames.all().order_by('ts_index') def get_emotion_raw(self): frames = self.get_frames() emotion_raw = [] last_non_empty_emotion_dict = {} for frame in frames: current_frame_emotion_dict = frame.azure_emotion_result if current_frame_emotion_dict is None: return [] if not current_frame_emotion_dict: if not last_non_empty_emotion_dict: current_frame_emotion_dict = {u'neutral': 0.0, u'sadness': 0.00, u'happiness': 0.0, u'disgust': 0.0, u'anger': 0.0, u'surprise': 0.0, u'fear': 0.0, u'contempt': 0.0} else: current_frame_emotion_dict = last_non_empty_emotion_dict else: last_non_empty_emotion_dict = current_frame_emotion_dict emotion_raw.append(current_frame_emotion_dict) return emotion_raw def get_question_template_competency_emotion_factor(self): return self.question.get_competency_emotion_factor() def calculate_emotion_score(self): frames = self.get_frames() for frame in frames: if frame.azure_emotion_result is None: return {} worker_logger.debug('frame is ok') q_c_e_factor = self.get_question_template_competency_emotion_factor() if q_c_e_factor is None: return {} worker_logger.debug('q_c_e_factor is ok') positive_score_list = [] negative_score_list = [] positive_factor = dict_to_list_by_sorted_key(q_c_e_factor.positive) positive_factor_np = np.array(positive_factor) negative_factor = dict_to_list_by_sorted_key(q_c_e_factor.negative) negative_factor_np = np.array(negative_factor) last_non_empty_emotion_sorted = [] for frame in frames: frame_emotion_data = dict_to_list_by_sorted_key(frame.azure_emotion_result) if len(frame_emotion_data) != settings.AZURE_EMOTION_DATA_DIM: if not last_non_empty_emotion_sorted: frame_emotion_data = [0] * settings.AZURE_EMOTION_DATA_DIM else: frame_emotion_data = last_non_empty_emotion_sorted else: last_non_empty_emotion_sorted = frame_emotion_data e_data_np = np.array([frame_emotion_data]) positive_score_list.append(np.around(np.dot(e_data_np, positive_factor_np)[0], decimals=2)) negative_score_list.append(np.around(np.dot(e_data_np, negative_factor_np)[0], decimals=2)) return { 'positive': positive_score_list, 'negative': negative_score_list, } def get_question_template_emotion_competency_model(self): return self.question.get_emotion_competency_model() def get_question_template_nlp_competency_model(self): return self.question.get_nlp_competency_model() def get_question_competency(self): return self.question.get_competency() def get_reference_position(self): return self.question.get_refer_position() def get_position(self): return self.question.get_position() def get_question_competency_wmd_reference(self): position = self.get_position() if not position: return None competency = self.get_question_competency() if not competency: return None from leapin.nlp.models import CompetencyWmdReference com_wmd_ref = None try: com_wmd_ref = CompetencyWmdReference.objects.get( position=position, competency=competency ) except Exception: pass return com_wmd_ref def get_competency_score(self): scores = self.competency_scores.all() if scores.count() > 0: return scores[0] return None def score(self, model_type_relation): worker_logger.debug('Answer %d Competency Score: start' % self.id) competency = self.get_question_competency() if competency is None: worker_logger.debug('Answer %d Competency Score: no competency found' % self.id) worker_logger.debug('Answer %d Competency Score: end' % self.id) return {} from leapin.leapinbox.models import AnswerCompetencyScore competency_score = None try: competency_score = AnswerCompetencyScore.objects.get( answer=self, competency=competency ) except ObjectDoesNotExist: emotion_score = 0.0 nlp_score = 0.0 for model_type in model_type_relation.keys(): if model_type == 'emotion': worker_logger.debug('Answer %d Emotion Competency Score: start' % self.id) emotion_score = self.score_competency_emotion() if emotion_score is None: worker_logger.debug('Answer %d Emotion Competency Score: error' % self.id) return {} worker_logger.debug('Answer %d Emotion Competency Score: end' % self.id) elif model_type == 'nlp': worker_logger.debug('Answer %d NLP Competency Score: start' % self.id) nlp_score = self.score_competency_nlp() if nlp_score is None: worker_logger.debug('Answer %d NLP Competency Score: error' % self.id) return {} worker_logger.debug('Answer %d NLP Competency Score: end' % self.id) competency_score = AnswerCompetencyScore.objects.create( answer=self, competency=competency, emotion_score=emotion_score, nlp_score=nlp_score ) return competency_score def score_competency(self): worker_logger.debug('Answer %d Competency Score: start' % self.id) competency = self.get_question_competency() if competency is None: worker_logger.debug('Answer %d Competency Score: no competency found' % self.id) worker_logger.debug('Answer %d Competency Score: end' % self.id) return {} worker_logger.debug('Answer %d Emotion Competency Score: start' % self.id) emotion_score = self.score_competency_emotion() worker_logger.debug('Answer %d Emotion Competency Score: end' % self.id) worker_logger.debug('Answer %d NLP Competency Score: start' % self.id) nlp_score = self.score_competency_nlp() worker_logger.debug('Answer %d NLP Competency Score: end' % self.id) return { 'competency': competency, 'emotion_score': emotion_score, 'nlp_score': nlp_score } def score_competency_emotion(self): frames = self.get_frames() frames_emotion_data_list = [] for frame in frames: if frame.azure_emotion_result is None: return None if len(frame.azure_emotion_result) > 0: frames_emotion_data_list.append(frame.azure_emotion_result) worker_logger.debug('Answer %d Emotion Competency Score: answer frames is ok' % self.id) q_c_s_m = self.get_question_template_emotion_competency_model() if q_c_s_m is None: return None worker_logger.debug('Answer %d Emotion Competency Score: q_c_s_m is ok' % self.id) # e_data_list = np.array(frames_emotion_data_list) # input = np.mean(e_data_list, axis=0) score = q_c_s_m.score_competency(frames_emotion_data_list) return score def get_nlp_data(self): return getattr(self, 'nlp_data', None) def score_competency_nlp(self): answer_nlp_data_record = self.get_nlp_data() if not answer_nlp_data_record: return None answer_nlp_data = answer_nlp_data_record.get_data() if not answer_nlp_data: return None worker_logger.debug('Answer %d NLP Competency Score: answer nlp data is ok' % self.id) q_c_s_m = self.get_question_template_nlp_competency_model() if q_c_s_m is None: return None worker_logger.debug('Answer %d NLP Competency Score: q_c_s_m is ok' % self.id) score = q_c_s_m.score_competency(answer_nlp_data) return score def m3u8_to_mp4(self): if self.get_mp4_video_url() is not None: return self if not self.rtmp_key: raise RtmpKeyNotValid('rtmp_key is invalid') mp4_video_url = m3u8_to_mp4(self.rtmp_key) self.video_url = mp4_video_url self.save() return self def convert_video_audio_to_mp3(self): if oss_object_exists(self.unified_video_url): return self if not self.video_url: raise VideoUrlNotValid('video url is invalid') unified_video_url = unified_video(self.video_url) self.unified_video_url = unified_video_url self.save() return self def frame_video(self): if self.is_answer_frame_status_success(): return self if not self.unified_video_url: raise VideoUrlNotValid('unified video url is invalid') prefix = 'application-%d/answer-%d' % (self.application.id, self.id) frame_list = mts_frame_video(self.unified_video_url, prefix) if len(frame_list) == 0: self.set_frame_status_fail() raise FrameVideoRequestFail('frame video api fail') from leapin.emotion.models import ( AnswerEmotionFrame, ) ts_index = 1 for frame in frame_list: answer_emotion_frame, created = AnswerEmotionFrame.objects.get_or_create( answer=self, ts_index=ts_index, ) answer_emotion_frame.aliyun_oss_key = frame answer_emotion_frame.save() ts_index = ts_index + 1 self.set_frame_status_success() return self def gen_frames_data(self): if not self.is_answer_frame_status_success(): raise AnswerFrameNotValid('answer has invalid frames') answer_frames = self.get_no_emotion_data_frames() if answer_frames is None: return self for frame in answer_frames: if frame.is_emotion_data_got(): continue emotion_api_type = 'free' position_settings = self.get_question().get_position().get_position_settings() if position_settings: api_type = position_settings.emotion_api_type face_info = get_frame_face_info( get_oss_object_signed_url(frame.aliyun_oss_key), api_type=emotion_api_type ) face_attr = face_info.get('faceAttributes', {}) emotion = face_attr.get('emotion', {}) frame.azure_fer = face_info frame.azure_emotion_result = emotion frame.save() return self def video_to_audio(self): if not self.unified_video_url: raise VideoUrlNotValid('unified video url is invalid') if self.audio_url: return self audio_url = interview_video_to_audio(self.unified_video_url) self.audio_url = audio_url self.save() return self def audio_to_text(self): if not self.audio_url: raise VideoUrlNotValid('audio url is invalid') if self.body: return self self.body = speech2text(self.audio_url, self.get_lang()) self.save() return self def get_body_lines(self): spliter = '.' if self.get_lang() == 'zh': spliter = u'。' return self.body.split(spliter) def get_wmd_reference_score(self, body_lines): wmd_reference = self.get_question_competency_wmd_reference() if not wmd_reference: return None return wmd_reference.get_wmd_reference_score(body_lines) def gen_nlp_data(self): positon = self.get_position() if not positon.is_nlp_support(): return if not self.body: raise Exception('answer body is none') sentimental = sentence_setimental(self.body, self.get_lang()) s_length = sentence_length(self.body, self.get_lang()) s_ratio = 1.0 answer_video_length = self.get_frames().count() if answer_video_length > 0: s_ratio = float(s_length)/float(answer_video_length) body_lines = self.get_body_lines() wmd_reference_score = self.get_wmd_reference_score(body_lines) if not wmd_reference_score: raise Exception('wmd reference score error. Answer %d' % self.id) from leapin.nlp.models import AnswerNLP a_nlp, created = AnswerNLP.objects.get_or_create(answer=self) a_nlp.sentimental_score = sentimental a_nlp.words_length = s_length a_nlp.words_ratio = s_ratio a_nlp.wmd_instance = wmd_reference_score a_nlp.save() return self
views.py
class AnswerListView(generics.ListCreateAPIView): serializer_class = AnswerListSerializer permission_classes = (IsAuthenticated, IsCandidate,) def get_queryset(self): user = self.request.user application_id = self.kwargs['application_id'] return Answer.objects.filter(user=user, application_id=application_id) def perform_create(self, serializer): application_id = self.kwargs['application_id'] question_id = self.request.data.get('question_id') try: application = Application.objects.get(id=int(application_id)) except Exception: raise BadRequestException('Application %s not exists' % str(application_id)) if not application.is_position_contains_question_by_id(question_id): raise BadRequestException('Question %s is not belong to position of Application %s' % (str(question_id), str(application.id))) serializer.save(user=self.request.user, application=application) def post(self, request, *args, **kwargs): """ <h4>Introduction</h4> <p>接口用于应聘者按问题提交answer</p> <br> <h4>Request</h4> <ul> <li> question_id - answer对应的问题id </li> <li> video_url - 直接上传oss的video地址 </li> <li> rtmp_key - 推流的oss点播文件key（前端缓存的） </li> </ul> <p>其中，video_url和rtmp_key是互斥参数，两者同时只能有一个。</p> <br> <h4>Response</h4> <p>返回中的video_url后端已做处理：</p> <ul> <li> 如果是使用直接上传oss的，则该字段返回video的url </li> <li> 如果是使用推流上传的，则该字段返回点播文件的加密播放地址，有效期为12个小时 </li> </ul> """ video_url = request.data.get('video_url', None) rtmp_key = request.data.get('rtmp_key', None) rtmp_playlist = request.user.get_oss_rtmp_playlist_by_rtmp_key(rtmp_key) if rtmp_playlist is None and video_url is None: raise BadRequestException('invalid rtmp_key') return self.create(request, *args, **kwargs) def get(self, request, *args, **kwargs): return self.list(request, *args, **kwargs)
urls.py
url('^application/(?P<application_id>\d+)/answers/$', csrf_exempt(views.AnswerListView.as_view())), # noqa