Re: [PATCH] Docs: Make notes on sequences and rollback more obvious - Mailing list pgsql-hackers

From Craig Ringer
Subject Re: [PATCH] Docs: Make notes on sequences and rollback more obvious
Date
Msg-id 5020CAEE.9050206@ringerc.id.au
Whole thread Raw
In response to Re: [PATCH] Docs: Make notes on sequences and rollback more obvious  (Robert Haas <robertmhaas@gmail.com>)
Responses Re: [PATCH] Docs: Make notes on sequences and rollback more obvious
Re: [PATCH] Docs: Make notes on sequences and rollback more obvious
List pgsql-hackers
On 08/07/2012 02:27 AM, Robert Haas wrote:
> I did not commit the advanced.sgml changes.

That's arguably the most important point to raise this. The most recent 
question came from someone who actually bothered to RTFM and believed 
based on the advanced-transactions page that rollback rolls *everything* 
back.

Some kind of hint that there are execptions is IMO very important. I'm 
not sure what the best form for it to take is.

>  I am not sure I believe
> the assertion that any function or type with special transactional
> behavior will include a documentation mention.

It absolutely should, but I guess that doesn't mean it's guaranteed to.

> It doesn't seem like a
> terribly future-proof assertion at any rate.  With respect to the
> mention of autocommit, I think it would be good to add something
> there, but maybe it should cross-reference our existing documentation
> mentions of autocommit.  Also, it's a bit ambiguous the way it's
> worded whether you get the automatic BEGIN/COMMIT with autocommit=on
> or with autocommit=off; somehow we should try to clarify what we mean
> a little more there.

Yeah. I should've kept that separate, as it was something I noticed in 
passing, rather than central to the changes.

--
Craig Ringer



pgsql-hackers by date:

Previous
From: "Etsuro Fujita"
Date:
Subject: Re: WIP Patch: Use sortedness of CSV foreign tables for query planning
Next
From: Pavel Stehule
Date:
Subject: is prefix pg_ reservated ?