InferenceResolver

The InferenceResolver class provides inference options for mapping Iceberg fields to Deephaven columns when loading an Iceberg table. It is useful when the structure of the table is not known in advance, and the mapping should be inferred from the Iceberg schema.

Syntax

InferenceResolver(
    infer_partitioning_columns: bool = False,
    fail_on_unsupported_types: bool = False,
    schema_provider: Optional[SchemaProvider] = None
)

Parameters

Parameter	Type	Description
infer_partitioning_columns	bool	Whether partitioning columns should be inferred based on the latest Iceberg `PartitionSpec`. Defaults to `False`. Warning: inferring partition columns is only recommended when the latest Iceberg `PartitionSpec` contains identity transforms that are not expected to change.
fail_on_unsupported_types	bool	Whether inference should fail if any Iceberg fields cannot be mapped to Deephaven columns. Defaults to `False`.
schema_provider	Optional[SchemaProvider]	The Iceberg schema to use for inference. If not specified, the current schema from the Iceberg table is used.

Methods

None.

Constructors

An InferenceResolver is constructed directly from the class.

Examples

Use InferenceResolver as the resolver argument when loading an Iceberg table if you want Deephaven to infer the mapping from the Iceberg schema.

The following example creates an InferenceResolver using the default settings, which does not infer partitioning columns and does not fail on unsupported types:

from deephaven.experimental import iceberg

resolver = iceberg.InferenceResolver()

The following example creates an InferenceResolver to fail if any Iceberg fields cannot be mapped to Deephaven columns:

from deephaven.experimental import iceberg

resolver = iceberg.InferenceResolver(
    infer_partitioning_columns=False, fail_on_unsupported_types=True
)

The resolver can be passed as an input argument to the IcebergCatalogAdapter load_table method.