On Mon, Feb 7, 2011 at 10:08 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
>> Which way did we more commonly do it before you applied this patch?
>
> We don't have a standard for this, and an undocumented patch applied
> without any discussion doesn't create one. It's hopeless to imagine
> that you'll ever achieve any uniformity that way. It won't last long
> if you do, since you're outnumbered by committers who won't be following
> whatever you think the convention is.
>
> I'm not even sure why you're trying --- I don't think it even makes
> sense to try to have a standard about this. I can easily imagine that
> integer constants might read better with <literal> in some contexts
> and better without in others.
*reads patch more carefully*
Here are my verdicts:
advanced.sgml: good
array.sgml: good
backup.sgml: unsure
catalogs.sgml: bad
client-auth.sgml: bad
config.sgml: bad
func.sgml: bad
high-availability.sgml: bad
libpq.sgml: bad
runtime.sgml: bad
spi.sgml: unsure
tsearch2.sgml: good
So I guess I'm back agreeing with you. Basically, it seems like we
ought to use <literal> if it's being used as a value that the user
might want to supply (e.g. "if you set this parameter to 0, then no
statements will be logged). It shouldn't use <literal> if it's just
being used as a number (e.g. "this query will return all airplanes
with a height of less than 30,000 feet"). The cases I'm unsure about
are the ones where we're talking about a return value (e.g. in the
event of an error, this function will return -1).
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company