Schema Projections

A Schema Projection defines a named exported schema from ordered steps. Each step uses schema, kind, and names.

Registering a projection

Use register_projection() to register a projection, then export_projection() to materialise the projected pandera.DataFrameSchema.

catalog.register_projection(
    "reporting_projection",
    steps=[
        {
            "schema": "sensor_readings",
            "kind": "columns",
            "names": ["site", "sensor_id", "temperature"],
        }
    ],
    description="Columns used by reporting consumers",
)

projected = catalog.export_projection("reporting_projection")
print(list(projected.columns.keys()))
# ['site', 'sensor_id', 'temperature']

Validation behavior

• Unknown step kinds raise ValueError.

• Unknown columns raise ValueError.

• Duplicate projected columns across steps raise ValueError.

• Missing schema references or duplicate projection names raise KeyError.

• kind: group is reserved for future support and currently raises NotImplementedError during registration.