2019-03 CF Summary / Review - Tranche #1

Hi,

As last year [1], I'll try to summarize all commitfest items in 2019-03
to see which I think could realistically be put into 12.

Going through all non bugfix CF entries. Here's the summary for the
entries I could stomach today:

RFC: ready for committer
NR: needs review
WOA: waiting on author.

- pgbench - another attempt at tap test for time-related options

NR. This was already around last year. I think it'd be fair to argue
that there's not been a ton of push to get this committed.

- Psql patch to show access methods info

This adds \ commands to display many properties of [index ]access methods.

NR. This patch has gotten a fair bit of committer feedback via
Alvaro. I personally am not particularly convinced this is
functionally that many people are going to use.

- Show size of partitioned table

NR. There seems to have been plenty discussion over details. Feels
like this ought to be committable for v12?

- pgbench - add pseudo-random permutation function

WOA. I'm not clear as to why we'd want to add this to pgbench. To
revive a discussion from last year's thread, I feel like we're adding
more code to pgbench than we can actually usefully use.

- libpq host/hostaddr consistency

NR. I think the patches in this needs a few more committer eyes. It's
possible we just should fix the documentation, or go further and
change the behaviour. Feels like without more senior attention,
this'll not be resolved.

- pg_dump multi VALUES INSERT

NR. There seems to be some tentative agreement, excepting Tom, that
we probably want this feature. There has been plenty review /
improvements.

Seems like it ought to be possible to get this into v12.

- libpq trace log

NR. There seems to be considerable debate about what exactly this
feature should do, and the code hasn't yet seen a lot of review. I
think we ought to just target v13 here, there seems to be some
agreement that there's a problem, just not exactly what the solution
is.

Andres: punt to v13.

- pg_dumpall --exclude-database option

RFC, and author is committer.

- Add sqlstate output mode to VERBOSITY

RFC, there seems to be agreement.

- DECLARE STATEMENT syntax support in ECPG

NR. There seems to be some tentative agreement that this is
desirable. But the patch was only recently (2018-12-16) revived, and
hasn't yet gotten a ton of review. I pinged Michael Meskes, to make
him aware of this work.

Andres: punt to v13.

- A new data type 'bytea' for ECPG host variable

NR: This seems to be much closer to being ready than the
above. Michael has done a few review cycles. Not sure if it'll be
done by 12, but it seems to have a good chance.

- \describe: verbose commands in psql

NR: This seems like a relatively large change, and only submitted
2019-01-24. Per our policy to not add nontrivial work to the last CF,
I think we should not consider this a candidate for v12.

Andres: punt to v13.

- documenting signal handling with readme

WOA: I'm very unclear as to why this is something worth documenting in
this manner. Right now I'm not clear what the audience of this is
supposed to be.

- First SVG graphic

NR: My reading of the discussion makes this look like we'll probably
have graphics in v12's docs. Neat.

- Update INSTALL file

WOA: It seems we've not really come to much of a conclusion what this
ought to contain.

- Make installcheck-world in a clean environment

NR: I don't feel we really have agreement on what we want here. Thus
I'm doubtful this is likely to be resolvable in short order.

Andres: lightly in favor of punting to v13

- Avoid creation of the free space map for small tables

NR: the majority of this patch has been committed, I assume the
remaining pieces will too.

- Adding a TAP test checking data consistency on standby with
minRecoveryPoint

NR: Seems like it ought to be committable, Andrew Gierth did provide
some feedback.

- idle-in-transaction timeout error does not give a hint

NR: Seems trivial enough.

- commontype and commontypearray polymorphic types

NR: To me this seems like a potentially large change, with user
visible impact. As it was only submitted in December, and is clearly
not yet reviewed much, I think we ought to punt on this for v12.

Andres: punt to v13

- EXPLAIN with information about modified GUC values

NR: The patch seems to be getting closer to completion, but I'm not
sure how many actually agree that we want this.

Andres: aim for v12, but we probably should discuss soon whether we
actually want this.

- Include all columns in default names for foreign key constraints.

WOA: This patch has been submitted to the last CF first. I think it's
straddling the line after which we should just refuse that pretty
closely. Not sure.

- Shared-memory based stats collector

WOA: I think this patch has quite some promise and solve quite a few
issues, and it has been worked on for a while. But at the same time
the code isn't that close to being committable.

Andres: unless there's a new version cleaning up review comments PDQ,
I think we're unfortunately have to punt this to v13.

- timeout parameters in libpq

NR: This doesn't yet seem terribly well reviewed and designed.

Andres: +0.5 for punting to v13

- Log a sample of transactions

WOA: Seems like a useful enough feature to me, but there were a few
other senior community members that didn't quite agree. Issues in the
patch seem resolvable in time for v12. I'm wondering if this doesn't
need a more radical approach to avoid the overhead.

- monitoring CREATE INDEX [CONCURRENTLY]

NR: I think there's agreement on the desirability of the feature. But
while the patch has been submitted to CF 2019-01 that seems to
have been somewhat of a placeholder entry. OTOH, it's a committer's
project, so we can give more leeway there.

- pg_stat_statements should notice FOR UPDATE clauses

NR: This seems like to get in, given how sipmle it is. Some quibbles
about the appropriate approach aside.

Ok, my flight's about to land. So that's it for this round.

Greetings,

Andres Freund

[1] https://postgr.es/m/20180301110344.kyp3wejoxp2ipler%40alap3.anarazel.de