[ENH] Improve Global Config Architecture (#1577)

#### Metadata
* Reference Issue: fixes #1564 
* New Tests Added: No
* Documentation Updated: No
* Change Log Entry: Replaces global variables with an instance of a
dataclass `OpenMLConfig`.


#### Details 
* What does this PR implement/fix? Explain your changes.
* Mainly Introduces an `OpenMLConfig` dataclass and an
`OpenMLConfigManager` class to accumulate all config functions and
varialbles into one class, and a single module-level instance `_config`
in `openml.config`.
  * Reads/parses configuration defaults from `OpenMLConfig`.
* Adds `__getattr__` so attribute reads on openml.config forward to the
dataclass (preserving read compatibility).
* Changes `openml/__init__.py` to route to the instance of
`OpenMLConfigManager` when `openml.config` is called.
* Moved `ConfigurationForExamples` inside `OpenMLConfigManager` to
preserve API.
  
* Why is this change necessary? What is the problem it solves?
  * consolidates defaults and runtime state into one place,
  * simplifies consistent parsing and writing of the config file,
  * removes the use of global variables.

---------

Co-authored-by: Franz Király <fkiraly@gcos.ai>
Co-authored-by: Armaghan Shakir <raoarmaghanshakir040@gmail.com>

Author: 3 people

openml/__init__.py

@@ -18,9 +18,11 @@
 # License: BSD 3-Clause
 from __future__ import annotations
 
+from typing import TYPE_CHECKING
+
 from . import (
     _api_calls,
-    config,
+    _config as _config_module,
     datasets,
     evaluations,
     exceptions,
@@ -49,6 +51,11 @@
     OpenMLTask,
 )
 
+if TYPE_CHECKING:
+    from ._config import OpenMLConfigManager
+
+config: OpenMLConfigManager = _config_module.__config
+
 
 def populate_cache(
     task_ids: list[int] | None = None,

openml/_api_calls.py

@@ -12,14 +12,16 @@
 import xml
 import zipfile
 from pathlib import Path
+from typing import cast
 
 import minio
 import requests
 import requests.utils
 import xmltodict
 from urllib3 import ProxyManager
 
-from . import config
+import openml
+
 from .__version__ import __version__
 from .exceptions import (
     OpenMLAuthenticationError,
@@ -70,7 +72,7 @@ def resolve_env_proxies(url: str) -> str | None:
 
 
 def _create_url_from_endpoint(endpoint: str) -> str:
-    url = config.server
+    url = cast("str", openml.config.server)
     if not url.endswith("/"):
         url += "/"
     url += endpoint
@@ -171,7 +173,7 @@ def _download_minio_file(
             bucket_name=bucket,
             object_name=object_name,
             file_path=str(destination),
-            progress=ProgressBar() if config.show_progress else None,
+            progress=ProgressBar() if openml.config.show_progress else None,
             request_headers=_HEADERS,
         )
         if destination.is_file() and destination.suffix == ".zip":
@@ -300,7 +302,8 @@ def _file_id_to_url(file_id: int, filename: str | None = None) -> str:
     Presents the URL how to download a given file id
     filename is optional
     """
-    openml_url = config.server.split("/api/")
+    openml_server = cast("str", openml.config.server)
+    openml_url = openml_server.split("/api/")
     url = openml_url[0] + f"/data/download/{file_id!s}"
     if filename is not None:
         url += "/" + filename
@@ -316,7 +319,7 @@ def _read_url_files(
     and sending file_elements as files
     """
     data = {} if data is None else data
-    data["api_key"] = config.apikey
+    data["api_key"] = openml.config.apikey
     if file_elements is None:
         file_elements = {}
     # Using requests.post sets header 'Accept-encoding' automatically to
@@ -336,8 +339,8 @@ def __read_url(
     md5_checksum: str | None = None,
 ) -> requests.Response:
     data = {} if data is None else data
-    if config.apikey:
-        data["api_key"] = config.apikey
+    if openml.config.apikey:
+        data["api_key"] = openml.config.apikey
     return _send_request(
         request_method=request_method,
         url=url,
@@ -362,10 +365,10 @@ def _send_request(  # noqa: C901, PLR0912
     files: FILE_ELEMENTS_TYPE | None = None,
     md5_checksum: str | None = None,
 ) -> requests.Response:
-    n_retries = max(1, config.connection_n_retries)
+    n_retries = max(1, openml.config.connection_n_retries)
 
     response: requests.Response | None = None
-    delay_method = _human_delay if config.retry_policy == "human" else _robot_delay
+    delay_method = _human_delay if openml.config.retry_policy == "human" else _robot_delay
 
     # Error to raise in case of retrying too often. Will be set to the last observed exception.
     retry_raise_e: Exception | None = None