"""Core functions of the cfg_load."""
# Core Library
import collections
import importlib.util
import json
import logging
import os
import pprint
import sys
from copy import deepcopy
from datetime import datetime
from typing import Any, Dict, List, Union
# Third party
import mpu
import pytz
import yaml
from mpu.datastructures import dict_merge, set_dict_value
from six.moves import configparser
# First party
import cfg_load.paths
import cfg_load.remote
from cfg_load._version import __version__ # noqa
[docs]def load(
filepath: str, load_raw: bool = False, load_remote: bool = True, **kwargs: Any
) -> Union["Configuration", Dict]:
"""
Load a configuration file.
Parameters
----------
filepath : str
Path to the configuration file.
load_raw : bool, optional (default: False)
Load only the raw configuration file as a dict,
without applying any logic to it.
load_remote : bool, optional (default: True)
Load files stored remotely, e.g. from a webserver or S3
**kwargs
Arbitrary keyword arguments which get passed to the loader functions.
Returns
-------
config : Configuration
"""
if filepath.lower().endswith(".yaml") or filepath.lower().endswith(".yml"):
config_dict = load_yaml(filepath, **kwargs)
elif filepath.lower().endswith(".json"):
config_dict = load_json(filepath, **kwargs)
elif filepath.lower().endswith(".ini"):
config_dict = load_ini(filepath, **kwargs)
else:
raise NotImplementedError(
f"Extension of the file '{filepath}' was not recognized."
)
if not load_raw:
reference_dir = os.path.dirname(filepath)
config_dict = cfg_load.paths.make_paths_absolute(reference_dir, config_dict)
config_dict = load_env(config_dict)
meta = mpu.io.get_file_meta(filepath)
meta["parse_datetime"] = datetime.now(pytz.utc)
config = Configuration(config_dict, meta=meta, load_remote=load_remote)
return config
def load_yaml(yaml_filepath: str, safe_load: bool = True, **kwargs: Any) -> Dict:
"""
Load a YAML file.
Parameters
----------
yaml_filepath : str
safe_load : bool, optional (default: True)
This triggers the usage of yaml.safe_load.
yaml.load can call any Python function and should only be used if the
source of the configuration file is trusted.
**kwargs : Any
Arbitrary keyword arguments which get passed to the loader functions.
Returns
-------
config : Dict
"""
with open(yaml_filepath) as stream:
if safe_load:
config = yaml.safe_load(stream)
else:
config = yaml.load(stream, **kwargs) # noqa
return config
def load_json(json_filepath: str, **kwargs: Any) -> Dict:
"""
Load a JSON file.
Parameters
----------
json_filepath : str
**kwargs : Any
Arbitrary keyword arguments which get passed to the loader functions.
Returns
-------
config : Dict
"""
with open(json_filepath) as stream:
config = json.load(stream, **kwargs)
return config
def load_ini(ini_filepath: str, **kwargs: Any) -> collections.OrderedDict:
"""
Load a ini file.
Parameters
----------
ini_filepath : str
**kwargs : Any
Arbitrary keyword arguments which get passed to the loader functions.
Returns
-------
config : OrderedDict
"""
config = configparser.ConfigParser(**kwargs)
config.read(ini_filepath)
# This is not so nice as it accesses a private property of the INI parser
return config._sections # type: ignore
def load_env(config: Dict) -> Dict:
"""
Load environment variables in config.
Parameters
----------
config : Dict
Returns
-------
config : Dict
"""
logger = logging.getLogger(__name__)
for env_name in os.environ:
if env_name.startswith("_"):
continue
if env_name in config:
if isinstance(config[env_name], str):
config[env_name] = os.environ[env_name]
elif isinstance(config[env_name], (list, dict, float, int, bool)):
config[env_name] = json.loads(os.environ[env_name])
else:
logger.warning(
f"Configuration value of {env_name} was "
f"{config[env_name]} of type {type(config[env_name])}, "
"but is overwritten with a string from the environment"
)
config[env_name] = os.environ[env_name]
return config
[docs]class Configuration(collections.abc.Mapping):
"""
Configuration class.
Essentially, this is an immutable dictionary.
Parameters
----------
cfg_dict : Dict
meta : Dict
load_remote : bool
"""
def __init__(self, cfg_dict: Dict, meta: Dict, load_remote: bool = True):
self._dict = deepcopy(cfg_dict) # make a copy
self._hash = None
meta["load_remote"] = load_remote
self._add_meta(meta)
self.modules: Dict = {}
self._load_modules(self._dict)
if load_remote:
self._load_remote(self._dict)
def __getitem__(self, key: Any) -> Any:
return self._dict[key]
def __len__(self) -> int:
return len(self._dict)
def __iter__(self) -> Any:
return iter(self._dict)
def __eq__(self, other: Any) -> bool:
return self._dict == other._dict
[docs] def set(self, key: str, value: Any) -> "Configuration": # noqa
"""
Set a value in the configuration.
Although it is discouraged to do so, it might be necessary in some
cases.
If you need to overwrite a dictionary, then you should do:
>> inner_dict = cfg['key']
>> inner_dict['inner_key'] = 'new_value'
>> cfg.set('key', inner_dict)
"""
self._dict[key] = value
return self
def __str__(self) -> str:
class_name = self.__class__.__name__
return "{class_name}({cfg_filepath})".format(
class_name=class_name, cfg_filepath=self.meta["filepath"]
)
def __repr__(self) -> str:
class_name = self.__class__.__name__
return (
"{class_name}(cfg_dict={cfg_dict}, meta={meta}, "
"load_remote={load_remote})".format(
class_name=class_name,
cfg_dict=self._dict,
meta=self.meta,
load_remote=self.meta["load_remote"],
)
)
def _add_meta(self, meta: Dict) -> "Configuration":
"""
Add meta data to configuration.
Parameters
----------
meta : Dict
Returns
-------
config : Configuration
"""
assert isinstance(meta, dict), f"type(meta)={type(meta)}, meta={meta}"
assert "parse_datetime" in meta, "meta does not contain parse_datetime"
self.meta = meta
self.meta["filepath"] = os.path.abspath(meta["filepath"])
return self
def _load_modules(self, config: Dict) -> Dict:
"""
Every key [SOMETHING]_module_path is loaded as a module.
The module is accessible at config.modules['SOMETHING'].
Parameters
----------
config : Dict
Returns
-------
config : Dict
"""
keyword = "_module_path"
if isinstance(config, list):
for i, el in enumerate(config):
config[i] = self._load_modules(el)
elif isinstance(config, dict):
for key in list(config.keys()):
if hasattr(key, "endswith"):
if key.startswith("_"):
continue
if key.endswith(keyword):
# Handler
sys.path.insert(1, os.path.dirname(config[key]))
spec = importlib.util.spec_from_file_location(
"foobar", config[key]
)
loaded_module = importlib.util.module_from_spec(spec)
target_key = key[: -len(keyword)]
self.modules[target_key] = loaded_module
if type(config[key]) is dict:
config[key] = self._load_modules(config[key])
return config
def _load_remote(self, config: Dict) -> Dict:
"""
Load remote paths.
Every key ending with `_load_url` has to have `source_url` and
`sink_path`. Sources which are AWS S3 URLs and URLs starting with
http(s) will be loaded automatically and stored in the sink. A `policy`
parameter can specify if it should be `load_always` or
`load_if_missing`.
Parameters
----------
config : Dict
Returns
-------
config : Dict
"""
keyword = "_load_url"
if isinstance(config, list):
for i, el in enumerate(config):
config[i] = self._load_remote(el)
elif isinstance(config, dict):
for key in list(config.keys()):
if hasattr(key, "endswith"):
if key.startswith("_"):
continue
if key.endswith(keyword):
# Handler
has_dl_info = (
"source_url" in config[key] and "sink_path" in config[key]
)
if not has_dl_info:
logging.warning(
f"The key '{key}' has not both keys "
"'source_url' and 'sink_path' "
)
else:
source = config[key]["source_url"]
sink = config[key]["sink_path"]
cfg_load.remote.load(source, sink)
if type(config[key]) is dict:
config[key] = self._load_remote(config[key])
return config
[docs] def update(self, other: "Configuration") -> "Configuration":
"""
Update this configuration with values of the other configuration.
Paramters
---------
other : Configuration
Returns
-------
updated_config : Configuration
"""
this_dict = deepcopy(self._dict)
other_dict = deepcopy(other._dict)
merged_dict = dict_merge(this_dict, other_dict, merge_method="take_right_deep")
cfg = Configuration(merged_dict, other.meta)
return cfg
[docs] def apply_env(self, env_mapping: List[Dict[str, Any]]) -> "Configuration":
"""
Apply environment variables to overwrite the current Configuration.
The env_mapping has the following structure (in YAML):
```
- env_name: "AWS_REGION"
keys: ["AWS", "REGION"]
converter: str
- env_name: "AWS_IS_ENABLED"
keys: ["AWS", "IS_ENABLED"]
converter: bool
```
If the env_name is not an ENVIRONMENT variable, then nothing is done.
If an ENVIRONMENT variable is not defined in env_mapping, nothing is
done.
Known converters:
* bool
* str
* str2str_or_none
* int
* float
* json
Parameters
----------
env_mapping : Configuration
Returns
-------
update_config : Configuration
"""
converters = {
"str": str,
"str2str_or_none": mpu.string.str2str_or_none,
"bool": mpu.string.str2bool,
"int": int,
"float": float,
"json": json.loads,
}
new_dict = deepcopy(self._dict)
for el in env_mapping:
env_name = el["env_name"]
if env_name not in os.environ:
continue
convert = converters[el["converter"]]
value = convert(os.environ[env_name])
set_dict_value(new_dict, el["keys"], value)
return Configuration(new_dict, self.meta)
[docs] def to_dict(self) -> Dict:
"""
Return a dictionary representation of the configuration.
This does NOT contain the metadata connected with the configuration.
It is discuraged to use this in production as it loses the metadata
and guarantees connected with the configuraiton object.
Returns
-------
config : dict
"""
return self._dict