On Wed, Oct 07, 2015 at 01:01:41PM -0500, Jim Nasby wrote:
> >Still, I'd welcome a native, streaming md5(loid) which is
> >bound to be more optimized by design.
>
> It would be nice if we had an interface to TOAST that allowed for streaming
> (well, really chunking) data to a function. That wouldn't help in this
> particular case, but it would significantly expand the usefulness of a
> streaming version of md5 and all the other hash operators.
Sure enough. A native md5(oid_of_large_object), which
internally uses lo_read(), didn't seem all that much work to
me.
Maybe I'm wrong on that side of things.
Karsten
--
GPG key ID E4071346 @ eu.pool.sks-keyservers.net
E167 67FD A291 2BEA 73BD 4537 78B9 A9F9 E407 1346