Horton

Re: [GENERAL] pg_dump -s dumps data?!

Tom Lane <tgl@sss.pgh.pa.us>

View thread
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andrew Dunstan <andrew@dunslane.net>
Cc: depesz@depesz.com, Adrian Klaver <adrian.klaver@gmail.com>, Dimitri Fontaine <dimitri@2ndQuadrant.fr>, pgsql-hackers@postgresql.org
Date: 2012-02-07T19:56:07Z
Lists: pgsql-hackers

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Add --section option to pg_dump and pg_restore.

  2. Add --exclude-table-data option to pg_dump.

Andrew Dunstan <andrew@dunslane.net> writes:
> On 01/31/2012 11:10 PM, Andrew Dunstan wrote:
>> Here's a possible patch for the exclude-table-data problem along the 
>> lines you suggest.

> Should I apply this?

I'm not happy with this yet.  My core complaint is that pg_dump used to
consider that creation of a TableDataInfo object for a table happens
if and only if we're going to dump the table's data.  And the comments
(eg in pg_dump.h) still say that.  But the previous patch left us in a
halfway zone where sometimes we'd create a TableDataInfo object and then
choose not to dump the data, and this patch doesn't get us out of that.
I think we should either revert to the previous definition, or go over
to a design wherein we always create TableDataInfo objects for all
tables (but probably still excluding data-less relations such as views)
and the whether-to-dump decision is expressed only by setting or not
setting the object's dump flag.

I worked a little bit on a patch to do the latter but found that it was
more invasive than I'd hoped.  Given the lack of any immediate payoff
I think it'd probably make more sense to do the former.  We could still
centralize the decision making into makeTableDataInfo a bit more than
now, but it should take the form of not creating the object at all,
rather than creating it and then clearing its dump flag.

			regards, tom lane