Thread: Bug #805: pg_dump examines all tables even with -t "table_name" speficied

Vitaliy Fuks (vitaliyfuks@yahoo.com) reports a bug with a severity of 3
The lower the number the more severe it is.

Short Description
pg_dump examines all tables even with -t "table_name" speficied

Long Description
Running pg_dump -t table_name -vv shows pg_dump examining every table in the database.  With many tables present (in
hundreds)this can make the process of dumping just one table take a considerable amount of time. 

With pg_dump -t specified it should go straight for the requested table and then figure what other dependent objects
needto be examined. 

Happens on pg_dump 7.2.3 under Linux (RedHat RPM) and i386 Solaris.

Are there some technical limitations forcing pg_dump to read everything?  If not it seems that the logic behind is just
notoptimal. 


Sample Code


No file was uploaded with this report

pgsql-bugs@postgresql.org writes:
> pg_dump examines all tables even with -t "table_name" speficied

7.3 is a little better.  It's not that easy to determine in advance what
information is needed, though; consider inheritance for example.

            regards, tom lane

Hello,

Could there potentially be an option in addition to -t "table_name", to
either:

a) Examine all relationships and dump all dependant objects
b) Dump only table named "table_name"

One can argue that when asking for "a" table one really only wants that
table information, and nothing else.  Some people of course can argue
that just a table by itself may not be usable by itself in some
situations - so that's why the option described above could cater both
situations.

I wish I could just give you a patch but there are probably people who
know pg_dump internals who could produce such patch in 10x less amount
of time.

--Vitaliy

--- Tom Lane <tgl@sss.pgh.pa.us> wrote:
> pgsql-bugs@postgresql.org writes:
> > pg_dump examines all tables even with -t "table_name" speficied
>
> 7.3 is a little better.  It's not that easy to determine in advance
> what
> information is needed, though; consider inheritance for example.
>
>             regards, tom lane


__________________________________________________
Do you Yahoo!?
HotJobs - Search new jobs daily now
http://hotjobs.yahoo.com/

We are thinking of adding a per-schema dump option in 7.4.  Would that
help?


---------------------------------------------------------------------------

Vitaliy Fuks wrote:
> Hello,
>
> Could there potentially be an option in addition to -t "table_name", to
> either:
>
> a) Examine all relationships and dump all dependant objects
> b) Dump only table named "table_name"
>
> One can argue that when asking for "a" table one really only wants that
> table information, and nothing else.  Some people of course can argue
> that just a table by itself may not be usable by itself in some
> situations - so that's why the option described above could cater both
> situations.
>
> I wish I could just give you a patch but there are probably people who
> know pg_dump internals who could produce such patch in 10x less amount
> of time.
>
> --Vitaliy
>
> --- Tom Lane <tgl@sss.pgh.pa.us> wrote:
> > pgsql-bugs@postgresql.org writes:
> > > pg_dump examines all tables even with -t "table_name" speficied
> >
> > 7.3 is a little better.  It's not that easy to determine in advance
> > what
> > information is needed, though; consider inheritance for example.
> >
> >             regards, tom lane
>
>
> __________________________________________________
> Do you Yahoo!?
> HotJobs - Search new jobs daily now
> http://hotjobs.yahoo.com/
>
> ---------------------------(end of broadcast)---------------------------
> TIP 4: Don't 'kill -9' the postmaster
>

--
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073