Paul M Foster wrote:
> Scenario: You have to update a record. One or more fields are unchanged
> from the original record being altered. So you have two options: 1)
> Include those fields in your UPDATE statement, even though they are
> unchanged; 2) Omit unchanged fields from the UPDATE statement.
>
> My first inclination is to omit unchanged fields. However, I have the
> idea that PG simply marks the existing record to be dropped, and
> generates a whole new row by copying unspecified fields from the
> original record.
>
> My question is, which is more efficient? Performance-wise, does it
> matter whether unchanged fields are included or omitted on UPDATE
> statements
my first order guess is, sending and having to parse the additional
unchanged fields in your UPDATE statement is more expensive than letting
the engine just copy them from the old tuple to the new.