ROLLBACK_TO_SAVEPOINT(7pg17)

ROLLBACK TO SAVEPOINT 💅 (PostgreSQL 17.11 docs, section 7)

okay so this is the PostgreSQL 17.11 Documentation page for ROLLBACK TO SAVEPOINT, and I’m gonna explain it like we’re in the group chat. Section 7, bestie. 🫶

NAME

ROLLBACK_TO_SAVEPOINT — roll back to a savepoint. That’s the name, that’s the vibe. 💖

SYNOPSIS ✨

Exact spell, no cap:

ROLLBACK [ WORK | TRANSACTION ] TO [ SAVEPOINT ] savepoint_name

DESCRIPTION 🦋

Okay so picture this: you’ve got a savepoint, and you run some commands after it. ROLLBACK TO SAVEPOINT rolls back all commands that were executed after the savepoint was established, then starts a new subtransaction at the same transaction level. The savepoint stays valid and can be rolled back to again later, if needed. Iconic. 💅

Also — and this is the plot twist — ROLLBACK TO SAVEPOINT implicitly destroys all savepoints that were established after the named savepoint. So it’s not just rolling back, it’s also clearing out the newer savepoints. Not me being dramatic, that’s literally the spec. 💀

PARAMETERS 🎀

savepoint_name — the savepoint to roll back to. That’s it, that’s the parameter. 🎀

NOTES 💀

Use RELEASE SAVEPOINT to destroy a savepoint without discarding the effects of commands executed after it was established. Different move, different vibe. ✨

Specifying a savepoint name that has not been established is an error. No cap, it will not vibe-check as valid. 😭

Cursors have somewhat non-transactional behavior with respect to savepoints — and this part is a whole saga:

Any cursor that is opened inside a savepoint will be closed when the savepoint is rolled back.
If a previously opened cursor is affected by a FETCH or MOVE command inside a savepoint that is later rolled back, the cursor remains at the position that FETCH left it pointing to (that is, the cursor motion caused by FETCH is not rolled back).
Closing a cursor is not undone by rolling back, either.
However, other side-effects caused by the cursor’s query (such as side-effects of volatile functions called by the query) are rolled back if they occur during a savepoint that is later rolled back.
A cursor whose execution causes a transaction to abort is put in a cannot-execute state, so while the transaction can be restored using ROLLBACK TO SAVEPOINT, the cursor can no longer be used. That’s the tragic part, bestie. 😭

EXAMPLES 🌸

To undo the effects of the commands executed after my_savepoint was established:

ROLLBACK TO SAVEPOINT my_savepoint;

Cursor positions are not affected by savepoint rollback:

BEGIN;

DECLARE foo CURSOR FOR SELECT 1 UNION SELECT 2;

SAVEPOINT foo;

FETCH 1 FROM foo;
 ?column?
----------
        1

ROLLBACK TO SAVEPOINT foo;

FETCH 1 FROM foo;
 ?column?
----------
        2

COMMIT;

COMPATIBILITY 👑

The SQL standard specifies that the key word SAVEPOINT is mandatory, but PostgreSQL and Oracle allow it to be omitted. SQL allows only WORK, not TRANSACTION, as a noise word after ROLLBACK. Also, SQL has an optional clause AND [ NO ] CHAIN which is not currently supported by PostgreSQL. Otherwise, this command conforms to the SQL standard. Periodt. 👑