Ron <ronljohnsonjr@gmail.com> writes:
> On 5/29/19 8:26 AM, Tom Lane wrote:
>> This doesn't explicitly talk about triggers, but I think our attitude
>> about the case you're discussing is that the results are unspecified.
>> If a trigger fired in one WITH arm tries to look at the table(s)
>> modified by other WITH arms, it might or might not see those changes.
> Are CTEs still optimization fences?
> https://www.2ndquadrant.com/en/blog/postgresql-ctes-are-optimization-fences/
Yes, if they contain data-modifying statements ... but that's not
relevant to the point at hand, which is that the run-time behavior
is not specified.
regards, tom lane