apache/beam · error · ValueError

ZScore.score_one expected univariate input, but got

Error message

ZScore.score_one expected univariate input, but got %s

What it means

ZScore.score_one computes the z-score of a single point and requires a univariate beam.Row with exactly one field; otherwise a ValueError is raised. Scoring shares the univariate constraint enforced at learning time.

Solutions

  1. Map to a single-field Row before score_one.
  2. Keep the learn/score input shapes identical by reusing the same projection transform.
  3. If multiple features must be scored together, instantiate one ZScore per feature.

Example fix

// before
windowed | beam.Map(detector.score_one)
// after
windowed | beam.Map(lambda r: beam.Row(value=r.metric)) | beam.Map(detector.score_one)
Defensive patterns

Strategy: validation

Validate before calling

if len(row.__dict__) != 1:
    raise ValueError(f"ZScore.score_one requires exactly one field, got: {list(row.__dict__)}")

Type guard

def is_univariate(row): return len(getattr(row, '__dict__', {})) == 1

Try / catch

try:
    score = detector.score_one(row)
except ValueError as e:
    score = None
    logger.warning("unscorable row %r: %s", row, e)

Prevention

When it happens

Trigger: Scoring rows with multiple attributes (or zero) through ZScore.score_one, typically when the scored stream retains extra columns like keys or timestamps.

Common situations: Piping the original multi-column PCollection into both learn and score; enrichment steps adding fields between learn and score; accidental reuse of a generic row-mapping transform.

Understand the failure class

Background: "Must be a positive integer", "Invalid value", "Unsupported": the invalid-argument-value error family, when a library rejects the value you pass — this error's family across 35 libraries.

Related errors

AI-assisted analysis of apache/beam@12126d8942 (2026-09-13). Data as JSON: /api/errors/1b5de3ed59d35b79. Report an issue: GitHub.

Appendix: source

Thrown at sdks/python/apache_beam/ml/anomaly/detectors/zscore.py:109

    if len(x.__dict__) != 1:
      raise ValueError(
          "ZScore.learn_one expected univariate input, but got %s", str(x))

    v = next(iter(x))
    self._stdev_tracker.push(v)
    self._sub_stat_tracker.push(v)

  def score_one(self, x: beam.Row) -> Optional[float]:
    """Scores a data point using the Z-Score.

    Args:
      x: A `beam.Row` containing a single numerical value.

    Returns:
      float | None: The Z-Score.
    """
    if len(x.__dict__) != 1:
      raise ValueError(
          "ZScore.score_one expected univariate input, but got %s", str(x))

    v = next(iter(x))
    if v is None or math.isnan(v):
      return None

    sub_stat = self._sub_stat_tracker.get()
    stdev = self._stdev_tracker.get()

    # not enough data points to compute sub_stat or standard deviation
    if math.isnan(stdev) or math.isnan(sub_stat):
      return float('NaN')

    if abs(stdev) < EPSILON:
      return 0.0

    return abs((v - sub_stat) / stdev)

View on GitHub (pinned to 12126d8942)