Between 9.6.5 and 10, the handling of parenthesized single-column UPDATE statements changed. In 9.6.5, they were treated identically to unparenthesized single-column UPDATES. In 10, they are treated as multiple-column updates. This results in this being valid in Postgres 9.6.5, but an error in Postgres 10:
CREATE TABLE test (id INT PRIMARY KEY, data INT);
In 10 and the current master, this produces the error:
errmsg("source for a multiple-column UPDATE item must be a sub-SELECT or ROW() expression")
This is a small patch to the grammar that restores the previous behavior by adding a rule to the set_clause rule and modifying the final rule of the set_clause rule to only match lists of more then one element. I'm not sure if there are more elegant or preferred ways to address this.
Compiled and tested on Ubuntu 17.04 Linux 4.10.0-33-generic x86_64.
Regression test added under the update test to cover the parenthesized single-column case.
I see no reason this would affect performance.
Thanks,
-rob
