[LucaCappelletti94]

Since sqlparser #2618 parses PostgreSQL's DO, a document holding a DO block no longer fails to parse. The statement fell into the catch-all arm of apply_statements, so ParserDB accepted the document and modelled it as if the block had never run. A block that creates a role, or runs CREATE TABLE through EXECUTE, left the model without that role or table and raised no error, and any comparison against a live catalog read the gap as drift.

What a DO block creates usually depends on the database it runs in, through IF NOT EXISTS checks on pg_roles, EXCEPTION WHEN duplicate_object handlers or dynamic EXECUTE. The model cannot know that state, so apply_statements now refuses every DO block with Error::UnsupportedDoBlock, which carries the statement as written. Blocks that only touch data are refused as well, which keeps the rule simple and matches what callers saw before sqlparser parsed DO.

Both lock files move to sqlparser ebb8b04a, the first revision with Statement::Do.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!