Tom Lane wrote:
> Andrew Dunstan <andrew@dunslane.net> writes:
>
>> Something someone said on IRC just now triggered a little memory ... I
>> think we should provide an option to have pg_dump work in CSV mode
>> rather than text mode. This probably doesn't have much importance in the
>> case of text dumps, but in custom or tar dumps where you might want to
>> get at individual data members, having an option for CSVs that you want
>> to load into some other product might be nice.
>>
>
> This is silly. You'd just COPY the particular table you want, not use
> pg_dump. pg_dump's already got an unreasonably large number of options
> without adding ones that have essentially zero use. Also, I think there
> are sufficient grounds to worry about whether a CSV dump would always
> reload correctly --- we already know that that's a poorly thought out
> "standard".
>
>
>
Well, if you have dozens or hundreds of tables it might well be more
convenient.
As for not reloading - I went to some trouble to make sure that we could
reload what we dumped, exactly, unless the force options are used. I
might have made a bug in that, but it isn't dependent on the particular
CSV format used.
Naturally you won't have a use for it, but I suspect others might (in
which case they had better speak up ;-) )
I suppose the alternative would be to write a little tool in perl or
whatever to do the same thing for you. Maybe a good pgfoundry project.
cheers
andrew