numeric_inference

sqllocks_spindle.schema.inference.numeric_inference

Upgrade numeric columns from generic uniform/normal to semantically appropriate distributions.

Examines each column's ColumnSemantic and replaces placeholder distribution parameters with realistic shapes: log-normal for money, geometric for quantities, bounded normal for percentages, and context-dependent normal for measurements.

Classes

NumericDistributionInferrer

Upgrade numeric column generators based on column semantics.

Only modifies columns whose current strategy is "distribution" with placeholder parameters. Columns that already have well-configured strategies (weighted_enum, derived, etc.) are left untouched.