Re: Patch for pg_dump: Multiple -t options and new -T option

David F. Skoll wrote:
> Hi,
>
> This is a response to several messages:
>
> 1) Copyright notice:  I have no problem having this removed, although it
> would be nice to credit me somewhere in a comment.

We credit in the commit message, and in the release notes so it will
always be seen.

> 2) I put most of the code in a separate file so that if the patch is
> rejected, it's easy for me to maintain a forked copy.  If the patch is
> accepted, obviously it can be integrated into an existing file.

OK.  Makes sense.

> 3) Multiple -n options:  We need to figure out how this would work, and make
> it non-surprising.  Some ideas:
>
>     pg_dump -t t1 -n s2 -t t2 -t t3 -n s4 -t t5
>
> What does that do?  My guess is:
>
>     - Dump table t1 in any schema
>     - Dump tables t2 and t3 in schema s2
>     - Dump table t5 in schema s4
>
> So now the position of the options matters!  That might surprise people,
> because:
>
>     pg_dump -s s1 -t t2
>
> is no longer the same as:
>
>     pg_dump -t t2 -n s1
>
> What about:
>
>     pg_dump -t t1 -n s2
>
> Should that dump table t1 in any schema, and any table in schema s2?
>
> If we can nail down the semantics, I can implement the patch.  The
> code is very simple.

Even though I suggested it, I am afraid this is just too confusing an API.

And I can't think of another one.  :-(


--
  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