On 05/05/17 16:56, Tom Lane wrote:
> Petr Jelinek <petr.jelinek@2ndquadrant.com> writes:
>> On 05/05/17 06:50, Tom Lane wrote:
>>> Actually, looking around a bit there, it's not even clear why
>>> we should be booby-trapping the value of an unchanged column in
>>> the first place. So I'd say that not only is the code dubious
>>> but the comment is inadequate too.
>
>> Hmm, as far as I can recollect this is just leftover debugging code that
>> was intended to help ensure that we are checking the "changed"
>> everywhere we are supposed to (since I changed handling of these
>> structured quite a bit during development). Should be changed to NULL,
>> that's what we usually do in this type of situation.
>
> So the comment should be something like "if the column is unchanged,
> we should not attempt to access its value beyond this point. To
> help catch any such attempts, set the string to NULL" ?
>
Yes that sounds about right. We don't get any data for unchanged TOAST
columns (that's limitation of logical decoding) so we better not touch them.
-- Petr Jelinek http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Training &
Services
