Schema policy.

Schema policy controls how much drift is acceptable before a target write. It should become stricter as data moves from raw capture to curated consumption.

Policies

Policy	Behavior	Typical use
permissive	Allows broad schema movement.	Raw landing and early exploration.
additive_only	Allows new columns, blocks removals and unsafe type changes.	Silver curated tables.
strict	Requires source and target schema alignment.	Gold consumption models.

Examples

YAML

schema_policy: additive_only
allow_type_widening: true

delta_properties:
  delta.enableChangeDataFeed: "true"
  delta.autoOptimize.optimizeWrite: "true"

Python

result = ingest(
    ...,
    schema_policy="additive_only",
    allow_type_widening=True,
    delta_properties={
        "delta.enableChangeDataFeed": "true",
        "delta.autoOptimize.optimizeWrite": "true",
    },
)

Change handling

New columns

Allowed by permissive and additive_only; recorded in schema-change evidence.

Removed columns

Blocked by curated policies. ContractForge does not automatically drop columns.

Type changes

Blocked unless the change is explicitly considered safe widening and allow_type_widening is enabled.

Control columns

Framework technical columns are ignored when comparing source business schema.

Audit

SELECT target_table, change_type, column_name, previous_type, new_type, changed_at_utc
FROM ops.ctrl_ingestion_schema_changes
ORDER BY changed_at_utc DESC;