Writing a Foreign Data Wrapper: Christophe Pettus - PGCon 2023

Avoid treating foreign data wrappers and relational tables interchangeably.

Understand the importance of error handling and logging in a foreign data wrapper.

Define a type for a foreign data wrapper, even if it’s not a complex one.

Create a separate extension for a foreign data wrapper, not a foreign data wrapper itself.

Write tests for a foreign data wrapper, due to the complexities involved.

Optimize a foreign data wrapper for performance by handling errors and logging effectively.

Choose a suitable data source for a foreign data wrapper, such as a weather feed.

Create a foreign data wrapper by defining the required functions and types.

Understand the importance of cryptography and security in a foreign data wrapper.

Use a foreign data wrapper as a drop-in replacement for a relational table, but with limitations.

Avoid using a foreign data wrapper as a way to bypass security and permissions checks.

Create a foreign data wrapper for a specific data source, such as an Oracle database or a flat file.

Use a foreign data wrapper to access data from a remote database or a file.

Optimize a foreign data wrapper for specific use cases, such as batch processing.

The planner and executor in Postgres rely on the foreign data wrapper to execute queries correctly.

Foreign data wrappers can be optimized for performance and security by implementing error handling and logging.

Writing a foreign data wrapper requires a good understanding of Postgres internals and C programming.

Foreign data wrappers can be used to access data from various sources, including relational databases and files.

Data types used in a foreign data wrapper must be defined and documented.

The foreign data wrapper is not responsible for foreign-specific logic, such as joins and grouping.

A foreign data wrapper must provide a way to handle errors and connection failures.

Writing a foreign data wrapper requires careful attention to error handling and logging.

Foreign data wrappers should be designed to be reentrant and thread-safe.

The planner and executor in Postgres rely on the foreign data wrapper to provide correct query results.