Re: Problem with trigger makes Detail record be invalid

=?UTF-8?Q?Fabr=C3=ADzio_de_Royes_Mello?= <fabrizio@timbira.com.br> writes:
> 2018-04-19 15:57 GMT-03:00 Tom Lane <tgl@sss.pgh.pa.us>:
>> (I'm not sure that this issue is adequately documented, though.
>> I'd have expected to find something about it in triggers.sgml and/or
>> create_trigger.sgml, but in a quick look neither of them mentions foreign
>> keys.)

> We don't have it properly documented... at the time I answered this
> question on pt-br stackoverflow I noticed the lack o documentation and
> unfortunately I completely forgot to propose a small patch for it.

It strikes me that there are basically two things a trigger could do to
break FK consistency:

1. Turn an FK-commanded update into a no-op by returning NULL.

2. Change the content of the FK-related columns during an FK-commanded
update.

Both of these apply only to BEFORE ROW triggers, of course.

It might not be unreasonable or unduly expensive to throw an error for
case #1.  I don't think I want to get into the expense of checking
for case #2, but covering case #1 would be enough to catch all of the
reports of this type of problem that I can remember.

IIRC, you can also break FK consistency with poorly-thought-out rules,
but given that rules are close-to-deprecated, I'm not very concerned
about sanding down rough edges in that case.

(But if you feel like writing a documentation patch, please do, because
we'd not be likely to back-patch a behavioral change like this even
if we get around to making it.)

            regards, tom lane