On 6/24/17 11:51, Tom Lane wrote:
> Ah, I was about to suggest the same thing, but I was coming at it from
> the standpoint of not requiring buffers several times larger than
> necessary, which could in itself cause avoidable palloc failures.
>
> I was going to suggest a small variant actually: run the conversion
> function an extra time only if the string is long enough to make the
> space consumption interesting, say
I had thought about something like that, too, but my concern is that we
then have double the code paths to test. I have run some performance
tests and I couldn't detect any differences between the variants. So
unless someone has any other insights, I think I'll go with the proposed
patch by tomorrow.
--
Peter Eisentraut http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services