Обсуждение: pgsql: doc: restrictions on alter database moving default tablespace
doc: restrictions on alter database moving default tablespace Mention tablespace must be empty and no one connected to the database. Report by Josh Berkus Branch ------ master Details ------- http://git.postgresql.org/pg/commitdiff/97d554871c3d186db831b35c0eabe6d496e4f54b Modified Files -------------- doc/src/sgml/ref/alter_database.sgml | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
On Sat, Oct 18, 2014 at 02:23:46PM +0000, Bruce Momjian wrote: > doc: restrictions on alter database moving default tablespace > > Mention tablespace must be empty and no one connected to the database. The database must not contain any object explicitly assigned to its future default tablespace, but that tablespace need not be empty. Objects of other databases in the tablespace pose no problem.
On Tue, Oct 21, 2014 at 01:38:43AM -0400, Noah Misch wrote:
> On Sat, Oct 18, 2014 at 02:23:46PM +0000, Bruce Momjian wrote:
> > doc: restrictions on alter database moving default tablespace
> >
> > Mention tablespace must be empty and no one connected to the database.
>
> The database must not contain any object explicitly assigned to its future
> default tablespace, but that tablespace need not be empty. Objects of other
> databases in the tablespace pose no problem.
I am back to look at this comment, and I don't understand what change is
being requested. Are you saying it is just the commit message that
isn't clear?
The new text is:
The new default tablespace for this database must be empty, and no one
can be connected to the database.
That makes the distinction that only the tablespace for the existing
database has to be empty, not the entire tablespace. Also, how would
there be an entry in the new-default per-database tablespace without it
being explicitly assigned, as that was not the previous default.
--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://enterprisedb.com
+ Everyone has their own god. +
On Fri, Dec 12, 2014 at 04:14:50PM -0500, Bruce Momjian wrote: > On Tue, Oct 21, 2014 at 01:38:43AM -0400, Noah Misch wrote: > > On Sat, Oct 18, 2014 at 02:23:46PM +0000, Bruce Momjian wrote: > > > doc: restrictions on alter database moving default tablespace > > > > > > Mention tablespace must be empty and no one connected to the database. > > > > The database must not contain any object explicitly assigned to its future > > default tablespace, but that tablespace need not be empty. Objects of other > > databases in the tablespace pose no problem. > > I am back to look at this comment, and I don't understand what change is > being requested. Are you saying it is just the commit message that > isn't clear? > > The new text is: > > The new default tablespace for this database must be empty, and no one > can be connected to the database. > > That makes the distinction that only the tablespace for the existing > database has to be empty, not the entire tablespace. No, the commit message accurately describes the change. Your reply helps me understand your conception of the software's behavior, but I still read your committed documentation text differently. For "the new default tablespace for this database" to qualify as "empty", it must contain no object from any database. A tablespace is a cross-database, unitary object. Each tablespace is either empty from every perspective or not empty from any perspective. Here are a couple of alternatives for the documentation text: The new default tablespace must not already contain any of the database's objects, and no one can be connected to the database. The database must not contain any object explicitly located in the new default tablespace, and no one can be connected to the database.
Noah Misch-2 wrote > On Fri, Dec 12, 2014 at 04:14:50PM -0500, Bruce Momjian wrote: >> On Tue, Oct 21, 2014 at 01:38:43AM -0400, Noah Misch wrote: >> > On Sat, Oct 18, 2014 at 02:23:46PM +0000, Bruce Momjian wrote: >> > > doc: restrictions on alter database moving default tablespace >> > > >> > > Mention tablespace must be empty and no one connected to the >> database. >> > >> > The database must not contain any object explicitly assigned to its >> future >> > default tablespace, but that tablespace need not be empty. Objects of >> other >> > databases in the tablespace pose no problem. >> >> I am back to look at this comment, and I don't understand what change is >> being requested. Are you saying it is just the commit message that >> isn't clear? >> >> The new text is: >> >> The new default tablespace for this database must be empty, and no one >> can be connected to the database. >> >> That makes the distinction that only the tablespace for the existing >> database has to be empty, not the entire tablespace. > > No, the commit message accurately describes the change. Your reply helps > me > understand your conception of the software's behavior, but I still read > your > committed documentation text differently. For "the new default tablespace > for > this database" to qualify as "empty", it must contain no object from any > database. A tablespace is a cross-database, unitary object. Each > tablespace > is either empty from every perspective or not empty from any perspective. > Here are a couple of alternatives for the documentation text: > > The new default tablespace must not already contain any of the > database's > objects, and no one can be connected to the database. > > The database must not contain any object explicitly located in the new > default tablespace, and no one can be connected to the database. Drive-by here but wouldn't saying (or adding) "cluster" instead of "database" solve the confusion? http://www.postgresql.org/docs/9.3/static/creating-cluster.html A link to that section may be worth considering... David J. -- View this message in context: http://postgresql.nabble.com/pgsql-doc-restrictions-on-alter-database-moving-default-tablespace-tp5823550p5830832.html Sent from the PostgreSQL - committers mailing list archive at Nabble.com.
Noah Misch-2 wrote > On Fri, Dec 12, 2014 at 04:14:50PM -0500, Bruce Momjian wrote: >> On Tue, Oct 21, 2014 at 01:38:43AM -0400, Noah Misch wrote: >> > On Sat, Oct 18, 2014 at 02:23:46PM +0000, Bruce Momjian wrote: >> > > doc: restrictions on alter database moving default tablespace >> > > >> > > Mention tablespace must be empty and no one connected to the >> database. >> > >> > The database must not contain any object explicitly assigned to its >> future >> > default tablespace, but that tablespace need not be empty. Objects of >> other >> > databases in the tablespace pose no problem. >> >> I am back to look at this comment, and I don't understand what change is >> being requested. Are you saying it is just the commit message that >> isn't clear? >> >> The new text is: >> >> The new default tablespace for this database must be empty, and no one >> can be connected to the database. >> >> That makes the distinction that only the tablespace for the existing >> database has to be empty, not the entire tablespace. > > No, the commit message accurately describes the change. Your reply helps > me > understand your conception of the software's behavior, but I still read > your > committed documentation text differently. For "the new default tablespace > for > this database" to qualify as "empty", it must contain no object from any > database. A tablespace is a cross-database, unitary object. Each > tablespace > is either empty from every perspective or not empty from any perspective. > Here are a couple of alternatives for the documentation text: > > The new default tablespace must not already contain any of the > database's > objects, and no one can be connected to the database. +1; though maybe... The newly assigned default tablespace must not contain any of the specified database's objects but otherwise need not be empty. Additionally, there must not be anyone connected to the database. This is a bit more verbose, and somewhat implied since newly created databases have their default set, by definition, to one hat likely already contains objects of other databases in the cluster, but does clarify the specific confusion raised here. Re-reading I think the implicit version is likely sufficiently clear though. Side note: This implies you cannot change the default tablespace of the only database in a cluster...probably not worth mentioning in the docs though. David J. -- View this message in context: http://postgresql.nabble.com/pgsql-doc-restrictions-on-alter-database-moving-default-tablespace-tp5823550p5830836.html Sent from the PostgreSQL - committers mailing list archive at Nabble.com.
On Mon, Dec 15, 2014 at 09:27:38PM -0700, David G Johnston wrote: > Noah Misch-2 wrote > > Here are a couple of alternatives for the documentation text: > > > > The new default tablespace must not already contain any of the > > database's > > objects, and no one can be connected to the database. > > +1; though maybe... > > The newly assigned default tablespace must not contain any of the specified > database's objects but otherwise need not be empty. Additionally, there > must not be anyone connected to the database. > > This is a bit more verbose, and somewhat implied since newly created > databases have their default set, by definition, to one hat likely already > contains objects of other databases in the cluster, but does clarify the > specific confusion raised here. Re-reading I think the implicit version is > likely sufficiently clear though. That text, too, works for me. > Side note: This implies you cannot change the default tablespace of the only > database in a cluster...probably not worth mentioning in the docs though. Agreed. By the way, standard practice on PostgreSQL community mailing lists is to use "reply to all". More often than not, without that To:/Cc: directly to me, I would never become aware of your reply.
On Fri, Dec 19, 2014 at 02:09:10AM -0500, Noah Misch wrote: > On Mon, Dec 15, 2014 at 09:27:38PM -0700, David G Johnston wrote: > > Noah Misch-2 wrote > > > Here are a couple of alternatives for the documentation text: > > > > > > The new default tablespace must not already contain any of the > > > database's > > > objects, and no one can be connected to the database. > > > > +1; though maybe... > > > > The newly assigned default tablespace must not contain any of the specified > > database's objects but otherwise need not be empty. Additionally, there > > must not be anyone connected to the database. > > > > This is a bit more verbose, and somewhat implied since newly created > > databases have their default set, by definition, to one hat likely already > > contains objects of other databases in the cluster, but does clarify the > > specific confusion raised here. Re-reading I think the implicit version is > > likely sufficiently clear though. > > That text, too, works for me. I went with cleaner wording that just rearranges the text to avoid the confusion. Attached patch applied. -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://enterprisedb.com + Everyone has their own god. +