Handling Unsupported Objects With Hooks

pgferry reports certain objects rather than guessing how to recreate them — and hooks are the normal way to finish that work yourself.

Objects you should expect to handle yourself

• views
• materialized views
• routines or procedures
• source triggers
• generated-column follow-up DDL
• custom validation SQL

Which hook phase to use

Need	Hook phase
create extensions or helper functions before data load	before_data
run ANALYZE or post-load data cleanup before validation/constraints	after_data
clean invalid child rows before foreign keys	before_fk
recreate views, functions, or final validation queries	after_all

Practical workflow

1. Run pgferry plan migration.toml --output-dir hooks.
2. Use the generated hook skeletons as the starting point.
3. Treat the generated files as sensitive if they include commented source SQL definitions from the source catalog.
4. Keep application-specific SQL in hook files rather than scattering it through runbooks.
5. Rehearse the exact hook set before the final cutover run.

Example

CREATE OR REPLACE VIEW {{schema}}.active_customers AS
SELECT customer_id, email
FROM {{schema}}.customer
WHERE active = true;

Use this in an after_all hook because the view depends on the finished schema and post-load objects.