Re: Adding comments to help understand psql hidden queries

Greg Sabino Mullane <htamfids(at)gmail(dot)com> writes:
> Going back through all the feedback and comments, plus having some time to
> think things through, I am including a new patch, v7, that greatly
> simplifies things, and only makes changes inside of describe.c. In the
> spirit of not letting the perfect be the enemy of the good, I'm not
> worrying at all about the number of stars, or the width, and simply adding
> a small consistent description at the top of each query. I also realized
> that having these queries show up in someone's server log could be quite
> confusing, so I had them output as part of the query itself. In other
> words, they show up in both psql -E and the server logs.

I like this proposal quite a lot. It seems like about the right level
of development/maintenance effort, and I think it provides a useful
increment of usability.

I have one slightly-orthogonal suggestion. I think that we should
make the header for generated queries be different from that used
for user queries in psql's logfile mode. Right now those are both
/******** QUERY *********/
I propose instead printing
/**** INTERNAL QUERY ****/
if it's a generated query.

I went through the 0007 patch and made some editorial changes
beyond that addition. I fixed a couple more internal queries
to have headers (now each PSQLexec in describe.c has one).
I also changed some of Greg's proposed headers in hopes of
improving uniformity. Notably, I didn't like that some of the
headers said "table" and some said "relation". I made them all
say "table", although you could certainly argue for the opposite.
This is all in-the-eye-of-the-beholder, so feel free to push back
on any changes you don't like.

For ease of review, v8-0001 attached is identical to Greg's 0007,
and v8-0002 is my proposed changes on top of that.

regards, tom lane