Migrate RESOURCE_CONFIG_PATH env var to GIGL_RESOURCE_CONFIG_URI

Makefile

@@ -135,7 +135,7 @@ integration_test: build_cpp_extensions
 
 
 notebooks_test:
-	RESOURCE_CONFIG_PATH=${GIGL_TEST_DEFAULT_RESOURCE_CONFIG} python -m tests.config_tests.notebooks_test
+	GIGL_RESOURCE_CONFIG_URI=${GIGL_TEST_DEFAULT_RESOURCE_CONFIG} python -m tests.config_tests.notebooks_test
 
 mock_assets:
 	uv run python -m gigl.src.mocking.dataset_asset_mocking_suite --resource_config_uri="deployment/configs/e2e_cicd_resource_config.yaml" --env test

gigl/env/constants.py

@@ -1,15 +1,54 @@
-"""Environment-variable keys exported by ``launch_custom``.
+"""Environment-variable keys used across GiGL.
 
-These keys are set on the subprocess env (never on the parent
-``os.environ``) by ``gigl.src.common.custom_launcher.launch_custom`` so
-that receiving CLIs can ``os.environ.get(...)`` their runtime context.
+Most of these keys are set on subprocess env (never on the parent
+``os.environ``) by ``gigl.src.common.custom_launcher.launch_custom`` so that
+receiving CLIs can ``os.environ.get(...)`` their runtime context.
+
+``GIGL_RESOURCE_CONFIG_URI`` is also written to the parent ``os.environ`` by
+``gigl.env.pipelines_config.get_resource_config`` so that downstream readers
+(e.g. ``GiglResourceConfigWrapper.get_resource_config_uri``) can recover the
+value within the same process. Use :func:`read_resource_config_uri_from_env`
+to read it; that helper also falls back to the deprecated
+``RESOURCE_CONFIG_PATH`` with a one-time warning.
 """
 
-from typing import Final
+import os
+from typing import Final, Optional
+
+from gigl.common.logger import Logger
 
 GIGL_APPLIED_TASK_IDENTIFIER_ENV_KEY: Final[str] = "GIGL_APPLIED_TASK_IDENTIFIER"
 GIGL_TASK_CONFIG_URI_ENV_KEY: Final[str] = "GIGL_TASK_CONFIG_URI"
 GIGL_RESOURCE_CONFIG_URI_ENV_KEY: Final[str] = "GIGL_RESOURCE_CONFIG_URI"
 GIGL_CPU_DOCKER_URI_ENV_KEY: Final[str] = "GIGL_CPU_DOCKER_URI"
 GIGL_CUDA_DOCKER_URI_ENV_KEY: Final[str] = "GIGL_CUDA_DOCKER_URI"
 GIGL_COMPONENT_ENV_KEY: Final[str] = "GIGL_COMPONENT"
+
+_LEGACY_RESOURCE_CONFIG_ENV_KEY: Final[str] = "RESOURCE_CONFIG_PATH"
+_legacy_resource_config_env_warned: bool = False
+_logger = Logger()
+
+
+def read_resource_config_uri_from_env() -> Optional[str]:
+    """Read the resource-config URI from the environment.
+
+    Prefers ``GIGL_RESOURCE_CONFIG_URI``. Falls back to the deprecated
+    ``RESOURCE_CONFIG_PATH`` and emits a one-time warning if that path is taken.
+
+    Returns:
+        The URI string if set under either name, else ``None``.
+    """
+    global _legacy_resource_config_env_warned
+    value = os.environ.get(GIGL_RESOURCE_CONFIG_URI_ENV_KEY)
+    if value is not None:
+        return value
+
+    legacy_value = os.environ.get(_LEGACY_RESOURCE_CONFIG_ENV_KEY)
+    if legacy_value is not None and not _legacy_resource_config_env_warned:
+        _logger.warning(
+            f"Environment variable {_LEGACY_RESOURCE_CONFIG_ENV_KEY!r} is deprecated; "
+            f"use {GIGL_RESOURCE_CONFIG_URI_ENV_KEY!r} instead. "
+            "Support for the legacy name will be removed in a future release."
+        )
+        _legacy_resource_config_env_warned = True
+    return legacy_value

gigl/env/pipelines_config.py

@@ -5,6 +5,10 @@
 
 from gigl.common import Uri, UriFactory
 from gigl.common.logger import Logger
+from gigl.env.constants import (
+    GIGL_RESOURCE_CONFIG_URI_ENV_KEY,
+    read_resource_config_uri_from_env,
+)
 from gigl.src.common.types.pb_wrappers.gigl_resource_config import (
     GiglResourceConfigWrapper,
 )
@@ -55,8 +59,9 @@ def get_resource_config(
     Args:
         resource_config_uri: Optional[Uri] = None
             The URI of the resource config file. If None, the function will try to load the resource config from the
-            command-line argument --resource_config_uri or the environment variable RESOURCE_CONFIG_PATH. If these are
-            not set, the function will try to load the resource config from the pipeline options.
+            command-line argument --resource_config_uri or the environment variable GIGL_RESOURCE_CONFIG_URI (the
+            deprecated RESOURCE_CONFIG_PATH is still read as a fallback, with a one-time warning). If these are not
+            set, the function will try to load the resource config from the pipeline options.
 
     Returns:
         resource_config: GiglResourceConfigWrapper
@@ -77,8 +82,8 @@ def get_resource_config(
             required=False,
         )
         args, _ = parser.parse_known_args()
-        resource_config_str = args.resource_config_uri or os.getenv(
-            "RESOURCE_CONFIG_PATH"
+        resource_config_str = (
+            args.resource_config_uri or read_resource_config_uri_from_env()
         )
 
         if resource_config_str is None:
@@ -90,7 +95,7 @@ def get_resource_config(
                     "No resource config provided, either via command-line argument or environment variable."
                 )
 
-    os.environ["RESOURCE_CONFIG_PATH"] = resource_config_str
+    os.environ[GIGL_RESOURCE_CONFIG_URI_ENV_KEY] = resource_config_str
     resource_config_path = UriFactory.create_uri(uri=resource_config_str)
 
     from gigl.common.utils.proto_utils import ProtoUtils

gigl/src/common/constants/resource_config.py

@@ -1,3 +1,2 @@
 SPARK_SPLIT_GENERATOR_CONFIG = "spark_split_generator_config"
 DATAFLOW_SPLIT_GENERATOR_CONFIG = "dataflow_split_generator_config"
-RESOURCE_CONFIG_OS_ENV = "RESOURCE_CONFIG_PATH"

gigl/src/common/types/pb_wrappers/gigl_resource_config.py

@@ -1,11 +1,10 @@
 import argparse
-import os
 from dataclasses import dataclass
 from typing import Optional, Union
 
-import gigl.src.common.constants.resource_config as resource_config_constants
 from gigl.common import GcsUri, UriFactory
 from gigl.common.logger import Logger
+from gigl.env.constants import read_resource_config_uri_from_env
 from gigl.src.common.constants.components import GiGLComponents
 from snapchat.research.gbml.gigl_resource_config_pb2 import (
     CustomLauncherConfig,
@@ -108,8 +107,8 @@ def get_resource_config_uri(self) -> str:
         )
         args, _ = parser.parse_known_args()
 
-        resource_config_path = args.resource_config_uri or os.getenv(
-            resource_config_constants.RESOURCE_CONFIG_OS_ENV
+        resource_config_path = (
+            args.resource_config_uri or read_resource_config_uri_from_env()
         )
 
         return str(resource_config_path)

scripts/launch_graph_store_local.py

@@ -67,6 +67,7 @@
 import yaml
 
 from gigl.distributed.utils import get_free_ports
+from gigl.env.constants import GIGL_RESOURCE_CONFIG_URI_ENV_KEY
 
 DEFAULT_JOB_ROOT = Path("/tmp/gigl")
 JOB_ID_SUFFIX_RE = re.compile(r"^(.*?)(\d+)$")
@@ -735,7 +736,7 @@ def main(argv: Optional[list[str]] = None) -> int:
             "MASTER_ADDR": args.host,
             "MASTER_PORT": str(master_port),
             "WORLD_SIZE": str(world_size),
-            "RESOURCE_CONFIG_PATH": resource_config_uri,
+            GIGL_RESOURCE_CONFIG_URI_ENV_KEY: resource_config_uri,
             "COMPUTE_CLUSTER_LOCAL_WORLD_SIZE": str(args.compute_procs_per_node),
             "PYTHONUNBUFFERED": "1",
         }