Re: BUG #15237: I got "ERROR: source for a multiple-column UPDATE item must be a sub-SELECT or ROW() expression"

>>>>> "Justin" == Justin Pryzby <pryzby(at)telsasoft(dot)com> writes:

>> It then apparently went unnoticed until after the release of pg 10,
>> at which point it got retroactively documented (in the release notes
>> and nowhere else), in response to a brief discussion of a user
>> complaint that happened on -general and not -hackers (or even
>> -bugs).

Justin> Actually I noticed and pointed out during beta;
Justin> https://www.postgresql.org/message-id/flat/20170719174507(dot)GA19616%40telsasoft(dot)com#20170719174507(dot)GA19616(at)telsasoft(dot)com

Interesting, I missed that when searching. So that tells us:

1. Breaking SET (col) = (val) wasn't actually intended or noticed in the
original patch;

2. The break was then defended based on an incorrect reading of the spec
(as I pointed out already on -bugs, the syntax _is_ legal in the spec
with only one column, for just the same reasons that VALUES (1) isn't
required to be VALUES row(1); the spec adds the ROW( ) construct
itself in one of the syntax rules, in a way that's easy to overlook)

--
Andrew (irc:RhodiumToad)