Do stored procedures run in database transaction in Postgres?

Harshal Dhumal picture Harshal Dhumal · Jan 31, 2015 · Viewed 9.4k times · Source

If a stored procedure fails in middle, are changes at that point from the beginning of SP rolled back implicitly or do we have to write any explicit code to make sure that SP runs in a database transaction only?

Answer

Erwin Brandstetter picture Erwin Brandstetter · Jan 31, 2015

Strictly speaking, Postgres did not have stored procedures as defined in the ISO/IEC standard before version 11. The term is often used incorrectly to refer to functions, which provide much of the same functionality (and more) as other RDBMS provide with "stored procedures". The main difference being transaction handling.

True stored procedures are finally introduced with Postgres 11:

Functions are atomic in Postgres and automatically run inside their own transaction unless called within an outer transaction. They always run inside a single transaction and succeed or fail completely. Consequently, one cannot begin or commit transactions within the function. And commands like VACUUM, CREATE DATABASE, or CREATE INDEX CONCURRENTLY which do not run in a transaction context are not allowed.

The manual on PL/pgSQL:

Functions and trigger procedures are always executed within a transaction established by an outer query — they cannot start or commit that transaction, since there would be no context for them to execute in. However, a block containing an EXCEPTION clause effectively forms a subtransaction that can be rolled back without affecting the outer transaction.

Error handling:

By default, any error occurring in a PL/pgSQL function aborts execution of the function, and indeed of the surrounding transaction as well. You can trap errors and recover from them by using a BEGIN block with an EXCEPTION clause.

There are exceptions, including but not limited to:

  • data written to log files
  • changes made to a sequence

    Important: Some PostgreSQL data types and functions have special rules regarding transactional behavior. In particular, changes made to a sequence (and therefore the counter of a column declared using serial) are immediately visible to all other transactions and are not rolled back if the transaction that made the changes aborts.

  • prepared statements

  • dblink calls (or similar)