Re: COPY ... FROM stdin WITH FORMAT csv

On Tue, Mar 21, 2017 at 1:45 PM, Adrian Klaver <adrian(dot)klaver(at)aklaver(dot)com>
wrote:

> On 03/21/2017 12:11 PM, Alexander Farber wrote:
>
>> Thank you - this has worked:
>>
>> COPY words_reviews (uid, author, nice, review, updated) FROM stdin WITH
>> (FORMAT csv);
>> 1,2,1,'1 is nice by 2','2017-03-01'
>> 1,3,1,'1 is nice by 3','2017-03-02'
>> 1,4,1,'1 is nice by 4','2017-03-03'
>> 2,1,1,'2 is nice by 1','2017-03-01'
>> 2,3,1,'2 is nice by 3','2017-03-02'
>> 2,4,0,'2 is not nice by 4','2017-03-03'
>> \.
>>
>> but I am confused about the comments that I should use \copy and not
>> just COPY and also that I could leave out WITH and brackets.
>>
>
> The difference between COPY and \copy is explained here:
>
> https://www.postgresql.org/docs/9.6/static/app-psql.html
>
> \copy ....
>
> Basically COPY runs as the server user and so the files it uses have to be
> accessible by the user the server runs as. \copy is a psql meta command
> that runs as local user so it can access files the local user can see and
> has privileges on.

​I take it, then, if one chooses not to use pre-existing files and instead
inline the content as shown here, the choice between \copy and COPY becomes
a matter of style and not correctness. As a matter of style using \copy
makes it clear that everything that is needed to make the command work
exists locally.

The advantage to using "COPY FROM|TO stdin|stdout" is that the copy command
itself could (not tested) be spread out over multiple lines - which
especially for COPY TO can be advantageous. The entire \copy meta-command
cannot be split up.

David J.