def input_handler()

in code/inference.py [0:0]

• 35 lines of code
• 2 McCabe index (conditional complexity)

def input_handler(data, context):
    """ Pre-process request input before it is sent to TensorFlow Serving REST API
    Args:
        data (obj): the request data, in format of dict or string
        context (Context): an object containing request and configuration details
    Returns:
        (dict): a JSON-serializable dict that contains request body and headers
    """
    logging.info("Starting input handler")
    
    if context.request_content_type == 'application/json':
        # pass through json (assumes it's correctly formed)
        d = json.load(data)
    
    # Splitting input into words
    logging.info("splitting input into words")
    global ids
    global sentences
    ids = [line.get('id', '') for line in d]
    sentences = [line.get('sentence', '').split() for line in d]
    
    logging.info("sentence preprocessing (split with max sequence length)")
    global split_sentences
    global split_idx
    split_sentences, split_idx = senpre.split_and_duplicate_index(sentences, MAX_SEQUENCE_LENGTH)

    logging.info("creating tags placement (-PAD- for unlabelled data)")
    tags_placement = []
    for sentence in split_sentences:
        tags_placement.append(['-PAD-']*len(sentence))
    
    logging.info("calling convert text to examples")
    bert_example = berpre.convert_text_to_examples(split_sentences, tags_placement)
    
    logging.info("convert examples to bert features")
    (input_ids, input_masks, segment_ids, _
    ) = berpre.convert_examples_to_features(tokenizer, bert_example, tag2int, max_seq_length=MAX_SEQUENCE_LENGTH+2)
    
    logging.info("convert bert features to necessary format for tensorflow serving")
    input_ids = input_ids.tolist()
    input_masks = input_masks.tolist()
    segment_ids = segment_ids.tolist()

    result = {'inputs':
                  {"input_word_ids": input_ids,
                   "input_mask": input_masks,
                   "input_type_ids": segment_ids
                   }
              }

    result = json.dumps(result)
    logging.info("returning right input for model in TFS format")
    return result