apache/beam · error · AttributeError

Could not find code object with path: {path}

Error message

Could not find code object with path: {path}

What it means

apache_beam's code_object_pickler serializes functions by identifier; _get_code_object_from_single_name_pattern walks an object's nested code constants (co_consts) looking for a nested function/code object whose co_name matches the final path component. When no matching nested code object exists it raises AttributeError with the full path. This means the dotted identifier does not resolve to a real code object in the module.

Source

Thrown at sdks/python/apache_beam/internal/code_object_pickler.py:357

    name_result: The result of the name pattern search.
    path: The path to the code object.

  Returns:
    The code object.

  Raises:
    ValueError: If the pattern is invalid.
    AttributeError: If the code object is not found.
  """
  if len(name_result.groups()) > 1:
    raise ValueError(f'Invalid pattern for single name: {name_result.group(0)}')
  # Groups are indexed starting at 1, group(0) is the entire match.
  name = name_result.group(1)
  if hasattr(obj, 'co_consts'):
    for co_const in obj.co_consts:
      if inspect.iscode(co_const) and co_const.co_name == name:
        return co_const
  raise AttributeError(f'Could not find code object with path: {path}')


def _get_code_object_from_lambda_with_args_pattern(
    obj: types.ModuleType, lambda_with_args_result: re.Match[str], path: str):
  """Returns the code object from a lambda with args pattern.

  Args:
    obj: The object to search within.
    lambda_with_args_result: The result of the lambda with args pattern search.
    path: The path to the code object.

  Returns:
    The code object.

  Raises:
    AttributeError: If the code object is not found.
  """
  name = lambda_with_args_result.group(1)

View on GitHub (pinned to 12126d8942)

Solutions

1. Verify the module imported on the runner actually contains the nested function named in the path (inspect the co_consts of the parent object).
2. Re-pin the Beam/dependency version so submission and worker use identical module code.
3. Regenerate the identifier from the current source instead of a stale one (re-run the pickling pipeline step).
4. Refactor the nested function to a module-level function so it is referenced by name directly rather than through co_consts traversal.

Example fix

// before
get_code_from_identifier('mymodule.outer.<lambda>')  # no nested code object named in co_consts
// after
# reference a real resolvable name instead
get_code_from_identifier('mymodule.outer')  # or refactor helper to top-level: 'mymodule.helper'

import sys, inspect, types
def identifier_resolves(path):
    parts = path.split('.')
    obj = sys.modules.get(parts[0])
    if obj is None: return False
    for part in parts[1:]:
        consts = getattr(obj, 'co_consts', None)
        if consts is None or not any(inspect.iscode(c) and c.co_name == part for c in consts):
            return False
    return True

def is_resolvable_identifier(path) -> bool:
    import sys
    parts = path.split('.')
    return bool(parts[0]) and parts[0] in sys.modules

try:
    code = get_code_from_identifier(path)
except AttributeError as e:
    logger.error('Identifier %s does not resolve (module/worker version skew?): %s', path, e)
    raise

When it happens

Trigger: Calling get_code_from_identifier (via _make_function_from_identifier during Beam pickling) with an identifier like 'mymodule.outer.<name>' where <name> is not a nested function/class of 'outer' at the version of the code actually loaded — e.g. the module on the worker differs from the one that created the identifier, or the name was renamed.

Common situations: Code version skew between job-submission environment and worker (stale .pyc or different package version on the cluster); renaming a nested helper without re-generating the identifier; hand-constructed identifiers for testing; __main__ vs installed-module mismatch in interactive/notebook runs.

Understand the failure class

Background: "Not found" and "does not exist" errors: why "Task not found", "No such folder", and "Can't find" fire when a lookup comes back empty — this error's family across 14 libraries.

Related errors

• cannot check importability of {} instances
• Qual name parts too long
• Invalid pattern for single name: {name_result.group(0)}
• Path must not be empty.
• Module {parts[0]} not found in sys.modules

AI-assisted analysis of apache/beam@12126d8942 (2026-09-13). Data as JSON: /api/errors/220e2e0da78c2f5d. Report an issue: GitHub.