PostgREST 12.2 is out! It comes with Observability and API improvements. In this post, we'll see what's new.

Version 12.2 ships with Prometheus-compatible metrics for PostgREST's schema cache and connection pool. These are useful for troubleshooting, for example, when PostgREST's pool is starved for connections.

A full list of supported metrics is available in the PostgREST documentation.

Sometimes it's handy to set a custom timeout per function. You can now do this on 12.2 projects with:

And calling the function with the RPC interface.

When doing set statement_timeouton the function, the statement_timeout will be “hoisted” and applied per transaction.

By default this also works for other settings, namely plan_filter.statement_cost_limit and default_transaction_isolation. The list of hoisted settings can be extended by modifying the db-hoisted-tx-settings configuration.

Before 12.2, this could be done by setting a statement_timeout on the API roles, but this affected all the SQL statements executed by those roles.

In prior versions of PostgREST, users could limit the number of records impacted by mutations (insert/update/delete) to 1 row using vendor media type application/vnd.pgrst.object+json. That supports a common use case but is not flexible enough to support user defined values.

12.2 introduces the max-affected preference to limit the affected rows up to a custom value.

For example:

If the number of affected records exceeds max-affected , an error is returned:

PostgREST v12.2 is already available on the Supabase platform on its latest patch version (v12.2.3) for new projects. Spin up a new project or upgrade your existing project to try it out!