Re: inconsistency and inefficiency in setup_conversion()

On 12/1/18, Dmitry Dolgov <9erthalion6(at)gmail(dot)com> wrote:
> I see that the author keeps patch updated, but I'm a bit worried because of
> lack of full review since probably May. I'm moving it to the next CF, let's
> see
> if there would be more feedback.
>
> P.S. adding Daniel, since he is assigned as a reviewer.

Having heard nothing in a while, I've removed Daniel as a reviewer to
make room for someone else. He is, of course free to re-add himself.
v8 is attached.

Since it's been a few months since last discussion, I'd like to
summarize the purpose of this patch and advocate for its inclusion in
v12:

1. Correctness
In the intro thread [1], I showed that object comments on some
conversions are wrong, and hard to fix given the current setup. This
is a documentation bug of sorts.

2. Clean-up
Currently, utils/mb/conversion_procs/Makefile has an ad-hoc script to
generate the SQL file, which has to be duplicated in the MSVC tooling,
and executed by initdb.c. Storing the conversions in .dat files
removes the need for any of that.

3. Performance
This patch shaves 5-6% off of initdb. Not as much as hoped, but still
a nice bonus.

[1] https://www.postgresql.org/message-id/CAJVSVGWtUqxpfAaxS88vEGvi%2BjKzWZb2EStu5io-UPc4p9rSJg%40mail.gmail.com

--John Naylor