def set_training_data_input()

in orbit/forecaster/forecaster.py [0:0]

• 49 lines of code
• 12 McCabe index (conditional complexity)

    def set_training_data_input(self):
        """Collects data attributes into a dict for sampling/optimization api"""
        # refresh a clean dict
        data_input_mapper = self._model.get_data_input_mapper()
        if not data_input_mapper:
            raise ForecasterException("Empty or invalid data_input_mapper")

        # always get standard input from training
        training_meta = self.get_training_meta()

        training_data_input = {
            TrainingMetaKeys.RESPONSE.value.upper(): training_meta[
                TrainingMetaKeys.RESPONSE.value
            ],
            TrainingMetaKeys.RESPONSE_SD.value.upper(): training_meta[
                TrainingMetaKeys.RESPONSE_SD.value
            ],
            TrainingMetaKeys.RESPONSE_MEAN.value.upper(): training_meta[
                TrainingMetaKeys.RESPONSE_MEAN.value
            ],
            TrainingMetaKeys.NUM_OF_OBS.value.upper(): training_meta[
                TrainingMetaKeys.NUM_OF_OBS.value
            ],
        }

        training_data_input = self.set_forecaster_training_meta(
            data_input=training_data_input
        )

        if isinstance(data_input_mapper, list):
            # if a list is provided, we assume an upper case in the mapper and reuse as the input value
            for key in data_input_mapper:
                key_lower = key.lower()
                input_value = getattr(self._model, key_lower, None)
                if input_value is None:
                    raise ForecasterException(
                        "{} is missing from data input".format(key_lower)
                    )
                # stan accepts bool as int only
                if isinstance(input_value, bool):
                    input_value = int(input_value)
                training_data_input[key] = input_value
        elif issubclass(data_input_mapper, Enum):
            # isinstance(data_input_mapper, object):
            for key in data_input_mapper:
                # mapper keys in upper case; attributes is defined in lower case; need a cae casting in conversion
                key_lower = key.name.lower()
                input_value = getattr(self._model, key_lower, None)
                if input_value is None:
                    raise ForecasterException(
                        "{} is missing from data input".format(key_lower)
                    )
                if isinstance(input_value, bool):
                    # stan accepts bool as int only
                    input_value = int(input_value)
                training_data_input[key.value] = input_value
        else:
            raise Exception(
                "Invalid type: data_input_mapper needs to be either an Enum or list."
            )

        self._training_data_input = training_data_input