Re: pg_dump multi VALUES INSERT - Mailing list pgsql-hackers

On Fri, 18 Jan 2019 at 01:15, Surafel Temesgen <surafel3000@gmail.com> wrote:
> The attache patch use your method mostly

I disagree with the "mostly" part.  As far as I can see, you took the
idea and then made a series of changes to completely break it.  For
bonus points, you put back my comment change to make it incorrect
again.

Here's what I got after applying your latest patch:

$ pg_dump --table=t --inserts --rows-per-insert=4 postgres

[...]
INSERT INTO public.t VALUES (1);
)
INSERT INTO public.t VALUES (, ( 2);
)
INSERT INTO public.t VALUES (, ( 3);
)
INSERT INTO public.t VALUES (, ( 4);
);
INSERT INTO public.t VALUES (5);
)
INSERT INTO public.t VALUES (, ( 6);
)
INSERT INTO public.t VALUES (, ( 7);
)
INSERT INTO public.t VALUES (, ( 8);
);
INSERT INTO public.t VALUES (9);
)
;

I didn't test, but I'm pretty sure that's not valid INSERT syntax.

I'd suggest taking my changes and doing the plumbing work to tie the
rows_per_statement into the command line arg instead of how I left it
hardcoded as 3.

>> +        When using <option>--inserts</option>, this allows the maximum number
>> +        of rows per <command>INSERT</command> statement to be specified.
>> +        This setting defaults to 1.
>>
> i change it too except "This setting defaults to 1"  because it doesn't have default value.
> 1 row per statement means --inserts option .

If it does not default to 1 then what happens when the option is not
specified