Re: pg_dump and --inserts / --column-inserts

Thomas Kellerer <spam_eater(at)gmx(dot)net> writes:
> Tom Lane wrote on 17.07.2010 16:36:
>> Well, nobody's offered any actual *numbers* here.

> I measured the runtime as seen from the JDBC client and as reported by explain analyze (the last line reading "Total runtime:")

The "runtime" from explain analyze really should not be measurably
different, since it doesn't include parse time or data transmission
time, and you ought to get the same execution plan with or without the
column names. I'd dismiss those numbers as being within experimental
error, except it seems odd that they all differ in the same direction.
The overall times seen from the client seem plausible though;
particularly since we can see an increase in the percentage overhead
as the number of columns increases, which is what you'd expect if
you were accurately measuring a column-name-lookup overhead.

regards, tom lane