apache/beam · error · ImportError

Bigquery dependencies are not installed.

Error message

Bigquery dependencies are not installed.

What it means

perf_analysis_utils.get_existing_issues_data queries BigQuery for previously filed GitHub issues about a test+metric regression. The google-cloud-bigquery import may be None when dependencies are missing, in which case it raises ImportError before creating the client.

Solutions

  1. pip install 'apache-beam[gcp]' to pull in google-cloud-bigquery
  2. Or pip install google-cloud-bigquery directly
  3. Run the analysis script inside the Beam load-test container image that includes GCP extras

Example fix

// before
pip install apache-beam==2.xx.0
// after
pip install 'apache-beam[gcp]'
Defensive patterns

Strategy: validation

Validate before calling

import importlib.util
if importlib.util.find_spec('google.cloud.bigquery') is None:
    raise SystemExit("BigQuery missing: pip install 'apache-beam[gcp]'")

Type guard

def has_bigquery() -> bool:
    import importlib.util
    return importlib.util.find_spec('google.cloud.bigquery') is not None

Try / catch

try:
    issues = get_existing_issues_data(test_name, metric_name)
except ImportError as e:
    if 'Bigquery dependencies are not installed' in str(e):
        logging.error("Install: pip install 'apache-beam[gcp]'")
        issues = None
    else:
        raise

Prevention

When it happens

Trigger: Calling run_change_point_analysis (which calls get_existing_issues_data) in a Python environment without google-cloud-bigquery installed (Beam installed without the [gcp] extra).

Common situations: Local dev runs of the change-point analysis without GCP deps; lightweight CI images; post-processing load-test results on a machine with a bare apache-beam install.

Understand the failure class

Background: "not installed", "pip install", "required for": how missing-dependency errors surface across open-source libraries — this error's family across 34 libraries.

Related errors

AI-assisted analysis of apache/beam@12126d8942 (2026-09-13). Data as JSON: /api/errors/70654f96b6df19f8. Report an issue: GitHub.

Appendix: source

Thrown at sdks/python/apache_beam/testing/analyzers/perf_analysis_utils.py:126

def is_change_point_in_valid_window(
    num_runs_in_change_point_window: int, latest_change_point_run: int) -> bool:
  return num_runs_in_change_point_window > latest_change_point_run


def get_existing_issues_data(table_name: str) -> Optional[pd.DataFrame]:
  """
  Finds the most recent GitHub issue created for the test_name.
  If no table found with name=test_name, return (None, None)
  else return latest created issue_number along with
  """
  query = f"""
  SELECT * FROM {constants._BQ_PROJECT_NAME}.{constants._BQ_DATASET}.{table_name}
  ORDER BY {constants._ISSUE_CREATION_TIMESTAMP_LABEL} DESC
  LIMIT 10
  """
  try:
    if bigquery is None:
      raise ImportError('Bigquery dependencies are not installed.')
    client = bigquery.Client()
    query_job = client.query(query=query)
    existing_issue_data = query_job.result().to_dataframe()
  except exceptions.NotFound:
    # If no table found, that means this is first performance regression
    # on the current test+metric.
    return None
  return existing_issue_data


def is_sibling_change_point(
    previous_change_point_timestamps: list[pd.Timestamp],
    change_point_index: int,
    timestamps: list[pd.Timestamp],
    min_runs_between_change_points: int,
    test_id: str,
) -> bool:
  """

View on GitHub (pinned to 12126d8942)