apache/beam · error · ValueError

Error parsing windowing config string at

Error message

Error parsing windowing config string at {identify_object(spec)}: {e}

What it means

When config.windowing is supplied as a string, preprocess_windowing parses it with yaml.safe_load; any failure — invalid YAML syntax or the parsed value not being a dict — is wrapped in this ValueError with the transform's identity and the underlying exception message. It exists to give context about which transform's windowing config is broken.

Solutions

1. Fix the YAML/JSON syntax indicated by the wrapped exception e in the message.
2. Prefer expressing windowing as a native mapping in the spec to avoid string parsing entirely.
3. Validate the string with yaml.safe_load locally to see the exact parse error.

Example fix

# before
config:
  windowing: "{type: fixed, size: }"
# after
config:
  windowing:
    type: fixed
    size: 10s

import yaml
w = spec.get('config', {}).get('windowing')
if isinstance(w, str):
    try:
        yaml.safe_load(w)
    except yaml.YAMLError as err:
        print(f'invalid windowing YAML: {err}')

try:
    spec = preprocess_windowing(spec)
except ValueError as e:
    if 'Error parsing windowing config string' in str(e):
        # e's cause shows the exact YAML/typing failure; fix and retry or surface
        print(e)
        raise

When it happens

Trigger: preprocess_windowing on a spec whose windowing string is malformed YAML (e.g. 'type: fixed size:' unterminated) or parses to a non-dict (which first raises TypeError('Windowing config string must be a YAML/JSON map.') and is caught here and re-raised).

Common situations: Multi-line JSON strings with indentation errors; forgetting to quote a string containing YAML-special characters; typos in JSON like single quotes around keys in a JSON-style string.

Related errors

• f"Invalid windowing value ' '. Must provide numeric value.
• f'Unknown window type
• "Invalid windowing time unit ' '. Valid time units are .
• Windowing config string must be a YAML/JSON map.
• accumulation_mode must be provided for non-trivial triggers

AI-assisted analysis of apache/beam@12126d8942 (2026-09-13). Data as JSON: /api/errors/4cd97dea8d7298c0. Report an issue: GitHub.

Appendix: source

Thrown at sdks/python/apache_beam/yaml/yaml_transform.py:1118

def preprocess_windowing(spec):
  if spec['type'] == 'WindowInto':
    # This is the transform where it is actually applied.
    if 'windowing' in spec:
      spec['config'] = spec.get('config', {})
      spec['config']['windowing'] = spec.pop('windowing')

    if spec.get('config', {}).get('windowing'):
      windowing_config = spec['config']['windowing']
      if isinstance(windowing_config, str):
        try:
          # PyYAML can load a JSON string - one-line and multi-line.
          # Without this code, multi-line is not supported.
          parsed_config = yaml.safe_load(windowing_config)
          if not isinstance(parsed_config, dict):
            raise TypeError('Windowing config string must be a YAML/JSON map.')
          spec['config']['windowing'] = parsed_config
        except Exception as e:
          raise ValueError(
              f'Error parsing windowing config string at \
                {identify_object(spec)}: {e}') from e
    return spec
  elif 'windowing' not in spec:
    # Nothing to do.
    return spec

  if spec['type'] == 'composite':
    # Apply the windowing to any reads, creates, etc. in this transform
    # TODO(robertwb): Better handle the case where a read is followed by a
    # setting of the timestamps. We should be careful of sliding windows
    # in particular.
    spec = push_windowing_to_roots(spec)

  windowing = spec.pop('windowing')
  if not is_empty(spec['input']):
    # Apply the windowing to all inputs by wrapping it in a transform that
    # first applies windowing and then applies the original transform.

View on GitHub (pinned to 12126d8942)