[DON'T MERGE] Go-SDK with Coordinator interface

.github/ISSUE_TEMPLATE/1-airflow_bug_report.yml

@@ -192,6 +192,7 @@ body:
         - redis
         - salesforce
         - samba
+        - sdk-java
         - segment
         - sendgrid
         - sftp

.github/boring-cyborg.yml

@@ -189,6 +189,9 @@ labelPRBasedOnFilePath:
   provider:keycloak:
     - providers/keycloak/**
 
+  provider:sdk-java:
+    - providers/sdk/java/**
+
   provider:microsoft-azure:
     - providers/microsoft/azure/**
 
@@ -261,6 +264,9 @@ labelPRBasedOnFilePath:
   provider:samba:
     - providers/samba/**
 
+  provider:sdk-executable:
+    - providers/sdk/executable/**
+
   provider:segment:
     - providers/segment/**
 

airflow-core/docs/extra-packages-ref.rst

@@ -178,6 +178,17 @@ all the ``airflow`` packages together - similarly to what happened in Airflow 2.
 ``airflow-task-sdk`` separately, if you want to install providers, you need to install them separately as
 ``apache-airflow-providers-*`` distribution packages.
 
+Multi-Language extras
+=====================
+
+These are extras that add dependencies needed for integration with other languages runtimes. Currently we have only Java SDK related extra, but in the future we might add more extras related to other languages runtimes.
+
++----------+------------------------------------------+------------------------------------------------------------------+
+| extra    | install command                          | enables                                                          |
++==========+==========================================+==================================================================+
+| sdk.java | ``pip install apache-airflow[sdk.java]`` | JavaCoordinator for both dag processing and workload execution.  |
++----------+------------------------------------------+------------------------------------------------------------------+
+
 Apache Software extras
 ======================
 

airflow-core/pyproject.toml

@@ -249,6 +249,7 @@ exclude = [
 "../shared/serialization/src/airflow_shared/serialization" = "src/airflow/_shared/serialization"
 "../shared/state/src/airflow_shared/state" = "src/airflow/_shared/state"
 "../shared/timezones/src/airflow_shared/timezones" = "src/airflow/_shared/timezones"
+"../shared/workloads/src/airflow_shared/workloads" = "src/airflow/_shared/workloads"
 "../shared/listeners/src/airflow_shared/listeners" = "src/airflow/_shared/listeners"
 "../shared/plugins_manager/src/airflow_shared/plugins_manager" = "src/airflow/_shared/plugins_manager"
 "../shared/providers_discovery/src/airflow_shared/providers_discovery" = "src/airflow/_shared/providers_discovery"
@@ -337,6 +338,7 @@ shared_distributions = [
     "apache-airflow-shared-serialization",
     "apache-airflow-shared-state",
     "apache-airflow-shared-timezones",
+    "apache-airflow-shared-workloads",
     "apache-airflow-shared-plugins-manager",
     "apache-airflow-shared-providers-discovery",
 ]

airflow-core/src/airflow/_shared/workloads

@@ -0,0 +1 @@
+../../../../shared/workloads/src/airflow_shared/workloads

airflow-core/src/airflow/config_templates/config.yml

@@ -1967,6 +1967,21 @@ workers:
       type: integer
       example: ~
       default: "60"
+sdk:
+  description: Settings for non-Python SDK runtime coordination
+  options:
+    queue_to_sdk:
+      description: |
+        JSON mapping of queue names to SDK runtime coordinator names.
+
+        When a task's ``language`` field is not set, this mapping is checked
+        to route the task to a non-Python runtime coordinator based on its
+        queue.  This is useful when queues are used as environment or
+        isolation identifiers (e.g. ``foo``, ``bar``).
+      version_added: 3.1.7
+      type: string
+      example: '{"foo": "java", "bar": "java", "go-queue": "go"}'
+      default: "{{}}"
 api_auth:
   description: Settings relating to authentication on the Airflow APIs
   options:

airflow-core/src/airflow/dag_processing/manager.py

@@ -66,7 +66,7 @@
 from airflow.sdk import SecretCache
 from airflow.sdk.log import init_log_file, logging_processors
 from airflow.typing_compat import assert_never
-from airflow.utils.file import list_py_file_paths, might_contain_dag
+from airflow.utils.file import might_contain_dag
 from airflow.utils.log.logging_mixin import LoggingMixin
 from airflow.utils.net import get_hostname
 from airflow.utils.process_utils import (
@@ -88,6 +88,9 @@
     from airflow.sdk.api.client import Client
 
 
+log = logging.getLogger(__name__)
+
+
 class DagParsingStat(NamedTuple):
     """Information on processing progress."""
 
@@ -158,6 +161,62 @@ def utc_epoch() -> datetime:
     return result
 
 
+def discover_dag_file_paths(
+    directory: str | os.PathLike[str] | None,
+    bundle_name: str = "",
+    safe_mode: bool = conf.getboolean("core", "DAG_DISCOVERY_SAFE_MODE", fallback=True),
+) -> list[str]:
+    """
+    Discover paths of DAG files within a directory.
+
+    Walks ``directory`` (honouring ``.airflowignore``) and returns each file that is
+    either a Python DAG candidate (``.py`` source or ZIP archive that passes
+    :func:`~airflow.utils.file.might_contain_dag`) or accepted by a registered coordinator's
+    :meth:`~airflow.sdk.execution_time.coordinator.BaseCoordinator.can_handle_dag_file`
+    (e.g. a ``.jar`` for the Java SDK, a self-contained executable for the Go SDK).
+
+    Coordinator handling takes precedence over the generic ZIP heuristic so that,
+    for example, a ``.jar`` is delegated to its coordinator rather than being
+    scanned for embedded ``.py`` modules.
+
+    :param directory: Directory to scan, or a single file path. ``None`` returns
+        an empty list. A single file is returned as-is without filtering.
+    :param bundle_name: Bundle name forwarded to ``can_handle_dag_file``.
+    :param safe_mode: Whether to apply the Python DAG heuristic; see
+        :func:`~airflow.utils.file.might_contain_dag`.
+    :return: Absolute paths discovered as DAG sources.
+    """
+    if directory is None:
+        return []
+    if os.path.isfile(directory):
+        return [str(directory)]
+    if not os.path.isdir(directory):
+        return []
+
+    from airflow._shared.module_loading.file_discovery import find_path_from_directory
+    from airflow.providers_manager import ProvidersManager
+
+    coordinators = ProvidersManager().coordinators
+    ignore_file_syntax = conf.get_mandatory_value("core", "DAG_IGNORE_FILE_SYNTAX", fallback="glob")
+
+    file_paths: list[str] = []
+    for file_path in find_path_from_directory(directory, ".airflowignore", ignore_file_syntax):
+        path = Path(file_path)
+        try:
+            if not path.is_file():
+                continue
+            if path.suffix == ".py":
+                if might_contain_dag(file_path, safe_mode):
+                    file_paths.append(file_path)
+            elif any(c.can_handle_dag_file(bundle_name, file_path) for c in coordinators):
+                file_paths.append(file_path)
+            elif zipfile.is_zipfile(path) and might_contain_dag(file_path, safe_mode):
+                file_paths.append(file_path)
+        except Exception:
+            log.exception("Error while examining %s", file_path)
+    return file_paths
+
+
 class _StubSelector(selectors.BaseSelector):
     """
     Stub to stand in until the real selector is created.
@@ -808,9 +867,11 @@ def _refresh_dag_bundles(self, known_files: dict[str, set[DagFileInfo]]):
 
     def _find_files_in_bundle(self, bundle: BaseDagBundle) -> list[Path]:
         """Get relative paths for dag files from bundle dir."""
-        # Build up a list of Python files that could contain DAGs
         self.log.info("Searching for files in %s at %s", bundle.name, bundle.path)
-        rel_paths = [Path(x).relative_to(bundle.path) for x in list_py_file_paths(bundle.path)]
+        rel_paths = [
+            Path(x).relative_to(bundle.path)
+            for x in discover_dag_file_paths(bundle.path, bundle_name=bundle.name)
+        ]
         self.log.info("Found %s files for bundle %s", len(rel_paths), bundle.name)
 
         return rel_paths
@@ -822,7 +883,13 @@ def _get_observed_filelocs(self, present: set[DagFileInfo]) -> set[str]:
         For regular files this includes the relative file path.
         For ZIP archives this includes DAG-like inner paths such as
         ``archive.zip/dag.py``.
+
+        Files claimed by a registered runtime coordinator (e.g. ``.jar``)
+        are treated as opaque files rather than ZIP archives.
         """
+        from airflow.providers_manager import ProvidersManager
+
+        coordinators = ProvidersManager().coordinators
 
         def find_zipped_dags(abs_path: os.PathLike) -> Iterator[str]:
             """Yield absolute paths for DAG-like files inside a ZIP archive."""
@@ -837,7 +904,10 @@ def find_zipped_dags(abs_path: os.PathLike) -> Iterator[str]:
         observed_filelocs: set[str] = set()
         for info in present:
             abs_path = str(info.absolute_path)
-            if abs_path.endswith(".py") or not zipfile.is_zipfile(abs_path):
+            handled_by_coordinator = any(
+                c.can_handle_dag_file(info.bundle_name, abs_path) for c in coordinators
+            )
+            if abs_path.endswith(".py") or handled_by_coordinator or not zipfile.is_zipfile(abs_path):
                 observed_filelocs.add(str(info.rel_path))
             else:
                 if TYPE_CHECKING:

airflow-core/src/airflow/dag_processing/processor.py

@@ -17,6 +17,7 @@
 from __future__ import annotations
 
 import contextlib
+import functools
 import importlib
 import logging
 import os
@@ -76,8 +77,6 @@
 from airflow.utils.state import TaskInstanceState
 
 if TYPE_CHECKING:
-    from socket import socket
-
     from structlog.typing import FilteringBoundLogger
 
     from airflow.api_fastapi.execution_api.app import InProcessExecutionAPI
@@ -86,6 +85,7 @@
     from airflow.sdk.definitions.context import Context
     from airflow.sdk.definitions.dag import DAG
     from airflow.sdk.definitions.mappedoperator import MappedOperator
+    from airflow.sdk.execution_time.supervisor import SelectorCallback
     from airflow.typing_compat import Self
 
 
@@ -553,7 +553,14 @@ def start(  # type: ignore[override]
     ) -> Self:
         logger = kwargs["logger"]
 
-        _pre_import_airflow_modules(os.fspath(path), logger)
+        # Check if a provider-registered runtime coordinator should handle this file
+        logger.debug("Checking for provider-registered runtime coordinator entrypoint for file", path=path)
+        resolved_target = cls._resolve_processor_target(path, bundle_name, bundle_path, logger)
+        if resolved_target is not None:
+            target = resolved_target
+            logger.debug("Resolved provider-registered runtime coordinator entrypoint for file", path=path)
+        else:
+            _pre_import_airflow_modules(os.fspath(path), logger)
 
         proc: Self = super().start(
             target=target,
@@ -566,6 +573,53 @@ def start(  # type: ignore[override]
         proc._on_child_started(callbacks, path, bundle_path, bundle_name)
         return proc
 
+    @staticmethod
+    def _resolve_processor_target(
+        path: str | os.PathLike[str],
+        bundle_name: str,
+        bundle_path: Path,
+        log: FilteringBoundLogger,
+    ) -> Callable[[], None] | None:
+        """
+        Return the entrypoint of the first provider runtime coordinator that can handle *path*.
+
+        The returned callable is a ``functools.partial`` that binds *path*, *bundle_name*
+        and *bundle_path* so the supervisor can pass it as a no-arg ``target`` to
+        ``WatchedSubprocess.start``.
+        """
+        from airflow.providers_manager import ProvidersManager
+
+        for coordinator_cls in ProvidersManager().coordinators:
+            try:
+                log.debug(
+                    "Checking runtime coordinator %s for file %s",
+                    coordinator_cls,
+                    path,
+                )
+                if coordinator_cls.can_handle_dag_file(bundle_name, path):
+                    log.debug(
+                        "Using runtime coordinator %s for file %s",
+                        coordinator_cls,
+                        path,
+                    )
+                    return functools.partial(
+                        coordinator_cls.run_dag_parsing,
+                        path=os.fspath(path),
+                        bundle_name=bundle_name,
+                        bundle_path=os.fspath(bundle_path),
+                    )
+                log.debug(
+                    "Runtime coordinator %s cannot handle file %s with bundle name %s",
+                    coordinator_cls,
+                    path,
+                    bundle_name,
+                )
+            except Exception:
+                log.warning("Failed to check runtime coordinator %s", coordinator_cls, exc_info=True)
+
+        log.debug("No runtime coordinator found for file %s, using default processor", path)
+        return None
+
     def _on_child_started(
         self,
         callbacks: list[CallbackRequest],
@@ -591,7 +645,7 @@ def _get_target_loggers(self) -> tuple[FilteringBoundLogger, ...]:
 
     def _create_log_forwarder(
         self, loggers: tuple[FilteringBoundLogger, ...], name: str, log_level: int = logging.INFO
-    ) -> Callable[[socket], bool]:
+    ) -> SelectorCallback:
         return super()._create_log_forwarder(loggers, name.replace("task.", "dag_processor.", 1), log_level)
 
     def _handle_request(self, msg: ToManager, log: FilteringBoundLogger, req_id: int) -> None:

airflow-core/src/airflow/executors/base_executor.py

@@ -651,10 +651,8 @@ def run_workload(
         if isinstance(workload, ExecuteTask):
             from airflow.sdk.execution_time.supervisor import supervise_task
 
-            # workload.ti is a TaskInstanceDTO which duck-types as TaskInstance.
-            # TODO: Create a protocol for this.
             return supervise_task(
-                ti=workload.ti,  # type: ignore[arg-type]
+                ti=workload.ti,
                 bundle_info=workload.bundle_info,
                 dag_rel_path=workload.dag_rel_path,
                 token=workload.token,

airflow-core/src/airflow/executors/workloads/task.py

@@ -22,8 +22,9 @@
 from pathlib import Path
 from typing import TYPE_CHECKING, Literal
 
-from pydantic import BaseModel, Field
+from pydantic import Field
 
+from airflow._shared.workloads import TaskInstanceDTO as _BaseTaskInstanceDTO
 from airflow.executors.workloads.base import BaseDagBundleWorkload, BundleInfo
 from airflow.utils.state import TaskInstanceState
 
@@ -33,8 +34,13 @@
     from airflow.models.taskinstancekey import TaskInstanceKey
 
 
-class TaskInstanceDTO(BaseModel):
-    """Schema for TaskInstance with minimal required fields needed for Executors and Task SDK."""
+class TaskInstanceDTO(_BaseTaskInstanceDTO):
+    """
+    TaskInstanceDTO with executor-specific ``key`` property.
+
+    Extends the shared :class:`~airflow._shared.workloads.TaskInstanceDTO`
+    to add the :attr:`key` property used by executors for workload tracking.
+    """
 
     id: uuid.UUID
     dag_version_id: uuid.UUID

airflow-core/src/airflow/models/dagcode.py

@@ -119,6 +119,16 @@ def code(cls, dag_id, session: Session = NEW_SESSION) -> str:
 
     @staticmethod
     def get_code_from_file(fileloc):
+        # Try from runtime coordinator first (classes are pre-loaded by ProvidersManager)
+        from airflow.providers_manager import ProvidersManager
+
+        for coordinator_cls in ProvidersManager().coordinators:
+            # TODO: Perhaps the `can_handle_dag_file` interface should just accept `path` only?
+            # Or maybe we can have different granularity for this. that 1 with bundle + path, another with just path
+            if coordinator_cls.can_handle_dag_file("", fileloc):
+                return coordinator_cls.get_code_from_file(fileloc)
+
+        # Then fallback to python native
         try:
             with open_maybe_zipped(fileloc, "r") as f:
                 code = f.read()

airflow-core/src/airflow/provider.yaml.schema.json

@@ -624,6 +624,13 @@
                 }
             }
         },
+        "coordinators": {
+            "type": "array",
+            "description": "Runtime Coordinator class names (BaseCoordinator subclasses)",
+            "items": {
+                "type": "string"
+            }
+        },
         "source-date-epoch": {
             "type": "integer",
             "description": "Source date epoch - seconds since epoch (gmtime) when the release documentation was prepared. Used to generate reproducible package builds with flint.",

airflow-core/src/airflow/provider_info.schema.json

@@ -446,6 +446,13 @@
                     "type": "string"
                 }
             }
+        },
+        "coordinators": {
+            "type": "array",
+            "description": "Runtime Coordinator class names (BaseCoordinator subclasses)",
+            "items": {
+                "type": "string"
+            }
         }
     },
     "definitions": {