Re: Triggers on columns

Tom Lane wrote:
> Itagaki Takahiro <itagaki(dot)takahiro(at)oss(dot)ntt(dot)co(dot)jp> writes:
>> Sure, and I found there might be difference between "UPDATE" and
>> "UPDATE OF {all-columns}" triggers. UPDATE trigger is always fired
>> when a row is updated even if none of the columns are actually
>> modified, but UPDATE OF {all-columns} trigger is fired only when
>> at least one of the columns is modified.
>
> I'm betraying the fact that I haven't read the patch, but ...
> exactly how, and when, are you determining whether a column has
> been "modified"? I can't count the number of times somebody
> has proposed simplistic and incorrect solutions to that.
> Usually they forget about BEFORE triggers changing the row.

It uses heap_tuple_attr_equals() to check whether a certain
column is modified, or not.

Itagaki-san, isn't it more suitable to check rte->modifiedCols
than heap_tuple_attr_equals()? Although, this information is
not delivered to executor...

What is the correct behavior when UPDATE statement set a new
value but it was identical to the original value?
In this case, heap_tuple_attr_equals() cannot detect the column
is used as a target of the UPDATE.

Thanks,
--
OSS Platform Development Division, NEC
KaiGai Kohei <kaigai(at)ak(dot)jp(dot)nec(dot)com>