Source code for rectorch.configuration

r"""The module contains useful classes to manage the configuration files.

Configuration files are useful to correctly initialize the data processing and the
recommendation engines. Configuration files must be `.json <https://www.json.org/json-en.html>`_
files with a specific format. Details about the file formats are described in :ref:`config-format`.
"""
import json
from os.path import exists
from munch import DefaultMunch

__all__ = ['DataConfig', 'ModelConfig', 'ConfigManager']

class Singleton(type):
    r"""Define an Instance operation that lets clients access its unique instance.
    """
    def __init__(cls, name, bases, attrs):
        super().__init__(name, bases, attrs)
        cls._instance = None

    def __call__(cls, *args, **kwargs):
        if cls._instance is None:
            cls._instance = super().__call__(*args, **kwargs)
        return cls._instance


[docs]class DataConfig(DefaultMunch): r"""Class containing the configurations for reading/writing the data set. Parameters ---------- file_path : :obj:`str` The path to the data configuration `.json <https://www.json.org/json-en.html>`_ file. Notes ----- The data configuration file **must** have the structure described in :ref:`config-format`. """ def __init__(self, file_path): super(DataConfig, self).__init__(None, json.load(open(file_path, "r"))) def __str__(self): return "DataConfig(" + ", ".join(["%s=%s" %(k, self[k]) for k in self]) + ")" def __repr__(self): return str(self)
[docs]class ModelConfig(): r"""Class containing the configurations for creating, training and testing the model. Parameters ---------- file_path : :obj:`str` The path to the model configuration `.json <https://www.json.org/json-en.html>`_ file. Attributes ---------- model : `DefaultMunch <https://github.com/Infinidat/munch>`_ Munch object containing the model's configurations according to the ``model`` key in the ``file_path`` json file. train : `DefaultMunch <https://github.com/Infinidat/munch>`_ Munch object containing the model training's configurations according to the ``train`` key in the ``file_path`` json file. test : `DefaultMunch <https://github.com/Infinidat/munch>`_ Munch object containing the test configurations according to the ``test`` key in the ``file_path`` json file. sampler : `DefaultMunch <https://github.com/Infinidat/munch>`_ Munch object containing the sampler configurations according to the ``model`` key in the ``file_path`` json file. Notes ----- The data configuration file **must** have the structure described in :ref:`config-format`. """ def __init__(self, file_path): json_cfg = json.load(open(file_path, "r")) self.model = DefaultMunch(None, json_cfg["model"]) self.train = DefaultMunch(None, json_cfg["train"]) self.test = DefaultMunch(None, json_cfg["test"]) self.sampler = DefaultMunch(None, json_cfg["sampler"]) def __str__(self): return "ModelConfig(model={}, train={}, test={}, sampler={}".format( self.model, self.train, self.test, self.sampler) def __repr__(self): return str(self)
[docs]class ConfigManager(metaclass=Singleton): r"""Wrapper class for both the data and model configurations. Parameters ---------- data_config_path : :obj:`str` The path to the data configuration `.json <https://www.json.org/json-en.html>`_ file. model_config_path : :obj:`str` The path to the model configuration `.json <https://www.json.org/json-en.html>`_ file. Attributes ---------- data_config : :class:`DataConfig` Object containing the configurations for reading/writing the data set. model_config : :class:`ModelConfig` Object containing the configurations for creating, training and testing the model. Examples -------- Initializing the :class:`ConfigManager` singleton: >>> from rectorch.configuration import ConfigManager >>> ConfigManager("path/to/the/dataconfig/file", "path/to/the/modelconfig/file") ConfigManager(data_config=DataConfig(...), model_config=ModelConfig(...)) """
[docs] @classmethod def get(cls): r"""Return the singleton ConfigManager instance. Returns ------- :class:`ConfigManager` The singletion :class:`ConfigManager` instance. Raises ------ :class:`Exception` Raised when the singleton :class:`ConfigManager` object has not been previously created. To initialize the :class:`ConfigManager` simply call its constructor. Please, see **Examples**. Examples -------- >>> from rectorch.configuration import ConfigManager >>> man = ConfigManager.get() Exception: Singleton object not instantiated! The :class:`ConfigManager` singleton object must be initialized to get it. >>> ConfigManager("path/to/the/dataconfig/file", "path/to/the/modelconfig/file") ConfigManager(data_config=DataConfig(...), model_config=ModelConfig(...)) >>> man = ConfigManager.get() >>> man ConfigManager(data_config=DataConfig(...), model_config=ModelConfig(...)) Getting the configuration objects is as easy as getting an attribute >>> man.data_config DataConfig(...) >>> man.model_config ModelConfig(...) """ if cls._instance: return cls._instance else: raise Exception("Singleton object not instantiated!")
def __init__(self, data_config_path, model_config_path): assert exists(data_config_path), "Data config file does not exist." assert exists(model_config_path), "Model config file does not exist." self.data_config = DataConfig(data_config_path) self.model_config = ModelConfig(model_config_path) def __str__(self): return "ConfigManager(data_config=%s, model_config=%s"%(self.data_config, self.model_config) def __repr__(self): return str(self)