Re: A PGsendQuery API design for your review

On Tue, 2 Feb 2016 10:15:40 -0800
Christophe Pettus <xof@thebuild.com> wrote:

>
> On Feb 2, 2016, at 10:04 AM, Karl O. Pinc <kop@meme.com> wrote:
>
> > OHow?  In order to submit multiple statements to libpq
> > and get back results for all of them PGsendQuery() must
> > be called.
>
> The same way the other clients do it; split the text into queries and
> send them over.

That requires my application contain an SQL parser.  This seems
onerous.  The whole point is that I'm getting unfiltered SQL directly
from a user.  Multiple statements.  "Splitting the text into queries"
is non-trivial.

I can call libpq directly.  First PQsendQuery(), then repeatedly
call PQgetResult() (optionally calling PQgetSingleRowMode().
But I like the psycopg2 API.  Seems like it ought to be able
to make these calls for me.

By the by.  The other clients probably don't split the text
into queries.  psql does, but to do it it uses the SQL
parser internal to pg.  phpPgAdmin does not.  And I don't
know what pgAdmin does.  My point in mentioning these interfaces
is that they all allow the user to submit arbitrary sql.
And the only one that does it "right" is psql, and it's
got "special access" to a SQL parser.

psql is the only client
I know of that delivers query results when multiple queries
are supplied (on stdin, say).  I want to be able to write
a similar application.

(Thanks for taking your time to look at this by the way.)

(And, oops.  The "buffered" argument must be to the
submitquery() method in my design.)

Regards,

Karl <kop@meme.com>
Free Software:  "You don't pay back, you pay forward."
                 -- Robert A. Heinlein