Skip to content

Added function for deterministic ID for class definitions by hashing. #36793

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
tvalentyn merged 3 commits into from
Nov 13, 2025
Merged

Added function for deterministic ID for class definitions by hashing. #36793

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
6bcd646
Added function for deterministic ID for class definitions by hashing.
praneetnadella Nov 11, 2025
9f7e521
Trigger CI: Rerun checks
praneetnadella Nov 11, 2025
a543b6d
addresrsing reviwer comments
praneetnadella Nov 12, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
28 changes: 24 additions & 4 deletions sdks/python/apache_beam/internal/cloudpickle/cloudpickle.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,6 +62,7 @@
import dis
from enum import Enum
import functools
import hashlib
import io
import itertools
import logging
Expand Down Expand Up @@ -98,7 +99,7 @@
_DYNAMIC_CLASS_TRACKER_BY_CLASS = weakref.WeakKeyDictionary()
_DYNAMIC_CLASS_TRACKER_BY_ID = weakref.WeakValueDictionary()
_DYNAMIC_CLASS_STATE_TRACKER_BY_CLASS = weakref.WeakKeyDictionary()
_DYNAMIC_CLASS_TRACKER_LOCK = threading.Lock()
_DYNAMIC_CLASS_TRACKER_LOCK = threading.RLock()

PYPY = platform.python_implementation() == "PyPy"

Expand Down Expand Up @@ -168,6 +169,7 @@ class CloudPickleConfig:


DEFAULT_CONFIG = CloudPickleConfig()
_GENERATING_SENTINEL = object()
builtin_code_type = None
if PYPY:
# builtin-code objects only exist in pypy
Expand All @@ -179,10 +181,21 @@ class CloudPickleConfig:
def _get_or_create_tracker_id(class_def, id_generator):
with _DYNAMIC_CLASS_TRACKER_LOCK:
class_tracker_id = _DYNAMIC_CLASS_TRACKER_BY_CLASS.get(class_def)
if class_tracker_id is _GENERATING_SENTINEL and id_generator:
raise RuntimeError(
f"Recursive ID generation detected for {class_def}. "
f"The id_generator cannot recursively request an ID for the same class."
)

if class_tracker_id is None and id_generator is not None:
class_tracker_id = id_generator(class_def)
_DYNAMIC_CLASS_TRACKER_BY_CLASS[class_def] = class_tracker_id
_DYNAMIC_CLASS_TRACKER_BY_ID[class_tracker_id] = class_def
_DYNAMIC_CLASS_TRACKER_BY_CLASS[class_def] = _GENERATING_SENTINEL
try:
class_tracker_id = id_generator(class_def)
_DYNAMIC_CLASS_TRACKER_BY_CLASS[class_def] = class_tracker_id
_DYNAMIC_CLASS_TRACKER_BY_ID[class_tracker_id] = class_def
except Exception:
_DYNAMIC_CLASS_TRACKER_BY_CLASS.pop(class_def, None)
raise
return class_tracker_id


Expand Down Expand Up @@ -1720,3 +1733,10 @@ def dumps(

# Backward compat alias.
CloudPickler = Pickler


def hash_dynamic_classdef(classdef):
Copy link
Contributor

@tvalentyn tvalentyn Nov 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  1. This is seemingly unused?

  2. Do we have to worry about (unlikely) hash collisions?

Copy link
Collaborator

@claudevdm claudevdm Nov 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It will be used by flume runner. It is tested in the cloud pickle repo.

Conflicts would be bad, should we try to detect?

Copy link
Collaborator

@claudevdm claudevdm Nov 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Although conflicts should result in unpickling ereor

Copy link
Collaborator

@claudevdm claudevdm Nov 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Although the original uuid approach can also cause collision?

Copy link
Contributor

@tvalentyn tvalentyn Nov 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

i'd worry if the errors are silenced.
Is detecting cheap? If so, that would help with a cleaner error message. Or if we believe this is really rare, we could add some comment in the codepath that will error-out.

Copy link
Contributor Author

@praneetnadella praneetnadella Nov 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

  1. The goal is to make this available for use and not called yet as an alternative to id_generator within CPConfig.

  2. I feel like since we are using SHA-256 which is pretty collision-resistant, and because a single Beam pipeline will likely have a wide array of variability/complexity in class definitions, the odds are extremely low.

"""Generates a deterministic ID by hashing the pickled class definition."""
hexdigest = hashlib.sha256(
dumps(classdef, config=CloudPickleConfig(id_generator=None))).hexdigest()
return hexdigest
Loading