183 lines
6.1 KiB
Cython
183 lines
6.1 KiB
Cython
from ray.includes.unique_ids cimport CObjectID
|
|
from ray.includes.optional cimport (
|
|
optional,
|
|
nullopt,
|
|
)
|
|
|
|
import asyncio
|
|
import concurrent.futures
|
|
import functools
|
|
import logging
|
|
import threading
|
|
from typing import Callable, Any, Union, Optional
|
|
from _collections_abc import GenericAlias
|
|
from builtins import StopAsyncIteration
|
|
|
|
import ray
|
|
import cython
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
def _set_future_helper(
|
|
result: Any,
|
|
*,
|
|
py_future: Union[asyncio.Future, concurrent.futures.Future],
|
|
):
|
|
# Issue #11030, #8841
|
|
# If this future has result set already, we just need to
|
|
# skip the set result/exception procedure.
|
|
if py_future.done():
|
|
return
|
|
|
|
if isinstance(result, RayTaskError):
|
|
exc = result.as_instanceof_cause()
|
|
# Convert StopIteration to RuntimeError to prevent segfaults due to
|
|
# Cpython's behavior w.r.t. PEP479
|
|
if isinstance(exc, StopIteration) or isinstance(exc, StopAsyncIteration):
|
|
runtime_error = RuntimeError(f"generator raised {type(exc).__name__}")
|
|
runtime_error.__cause__ = exc
|
|
py_future.set_exception(runtime_error)
|
|
else:
|
|
py_future.set_exception(exc)
|
|
elif isinstance(result, RayError):
|
|
# Directly raise exception for RayActorError
|
|
py_future.set_exception(result)
|
|
else:
|
|
py_future.set_result(result)
|
|
|
|
|
|
cdef class ObjectRef(BaseID):
|
|
__class_getitem__ = classmethod(GenericAlias) # should match how typing.Generic works
|
|
|
|
def __cinit__(self):
|
|
self.in_core_worker = False
|
|
|
|
def __init__(
|
|
self, id, owner_addr="", call_site_data="",
|
|
skip_adding_local_ref=False, tensor_transport: Optional[str] = None):
|
|
self._set_id(id)
|
|
self.owner_addr = owner_addr
|
|
self.in_core_worker = False
|
|
self.call_site_data = call_site_data
|
|
self._tensor_transport = tensor_transport
|
|
worker = ray._private.worker.global_worker
|
|
# TODO(edoakes): We should be able to remove the in_core_worker flag.
|
|
# But there are still some dummy object refs being created outside the
|
|
# context of a core worker.
|
|
if hasattr(worker, "core_worker"):
|
|
if not skip_adding_local_ref:
|
|
worker.core_worker.add_object_ref_reference(self)
|
|
self.in_core_worker = True
|
|
|
|
def __dealloc__(self):
|
|
if self.in_core_worker:
|
|
try:
|
|
worker = ray._private.worker.global_worker
|
|
worker.core_worker.remove_object_ref_reference(self)
|
|
except Exception as e:
|
|
# There is a strange error in rllib that causes the above to
|
|
# fail. Somehow the global 'ray' variable corresponding to the
|
|
# imported package is None when this gets called. Unfortunately
|
|
# this is hard to debug because __dealloc__ is called during
|
|
# garbage collection so we can't get a good stack trace. In any
|
|
# case, there's not much we can do besides ignore it
|
|
# (re-importing ray won't help).
|
|
pass
|
|
|
|
cdef CObjectID native(self):
|
|
return <CObjectID>self.data
|
|
|
|
def binary(self):
|
|
return self.data.Binary()
|
|
|
|
def hex(self):
|
|
return decode(self.data.Hex())
|
|
|
|
def is_nil(self):
|
|
return self.data.IsNil()
|
|
|
|
cdef size_t hash(self):
|
|
return self.data.Hash()
|
|
|
|
def task_id(self):
|
|
return TaskID(self.data.TaskId().Binary())
|
|
|
|
def job_id(self):
|
|
return self.task_id().job_id()
|
|
|
|
def owner_address(self):
|
|
return self.owner_addr
|
|
|
|
def call_site(self):
|
|
return decode(self.call_site_data)
|
|
|
|
@classmethod
|
|
def size(cls):
|
|
return CObjectID.Size()
|
|
|
|
def _set_id(self, id):
|
|
check_id(id)
|
|
self.data = CObjectID.FromBinary(<c_string>id)
|
|
|
|
@classmethod
|
|
def nil(cls):
|
|
return cls(CObjectID.Nil().Binary())
|
|
|
|
@classmethod
|
|
def from_random(cls):
|
|
return cls(CObjectID.FromRandom().Binary())
|
|
|
|
def future(self) -> concurrent.futures.Future:
|
|
"""Wrap ObjectRef with a concurrent.futures.Future
|
|
|
|
Note that the future cancellation will not cancel the correspoding
|
|
task when the ObjectRef representing return object of a task.
|
|
Additionally, future.running() will always be ``False`` even if the
|
|
underlying task is running.
|
|
"""
|
|
py_future = concurrent.futures.Future()
|
|
|
|
self._on_completed(
|
|
functools.partial(_set_future_helper, py_future=py_future))
|
|
|
|
# A hack to keep a reference to the object ref for ref counting.
|
|
py_future.object_ref = self
|
|
return py_future
|
|
|
|
def __await__(self):
|
|
return self.as_future(_internal=True).__await__()
|
|
|
|
def as_future(self, _internal=False) -> asyncio.Future:
|
|
"""Wrap ObjectRef with an asyncio.Future.
|
|
|
|
Note that the future cancellation will not cancel the correspoding
|
|
task when the ObjectRef representing return object of a task.
|
|
"""
|
|
if not _internal:
|
|
logger.warning("ref.as_future() is deprecated in favor of "
|
|
"asyncio.wrap_future(ref.future()).")
|
|
return asyncio.wrap_future(self.future())
|
|
|
|
def _on_completed(self, py_callback: Callable[[Any], None]):
|
|
"""Register a callback that will be called after Object is ready.
|
|
If the ObjectRef is already ready, the callback will be called soon.
|
|
The callback should take the result as the only argument. The result
|
|
can be an exception object in case of task error.
|
|
"""
|
|
core_worker = ray._private.worker.global_worker.core_worker
|
|
core_worker.set_get_async_callback(self, py_callback)
|
|
return self
|
|
|
|
def tensor_transport(self):
|
|
return self._tensor_transport
|
|
|
|
cdef optional[c_string] c_tensor_transport(self):
|
|
cdef:
|
|
optional[c_string] c_tensor_transport = nullopt
|
|
c_string c_tensor_transport_str
|
|
if self._tensor_transport is not None:
|
|
c_tensor_transport_str = self._tensor_transport.encode("utf-8")
|
|
c_tensor_transport.emplace(move(c_tensor_transport_str))
|
|
return c_tensor_transport
|