My bad. I was retyping code into the email that I was looking at via VNC on
the linux box which is why it got a bit mangled. I was finally able to
transfer it over (had to restart Eclipse). I yanked extra stuff that
doesn't matter (logging, etc). As far as I can tell I have brought
everything up to the absolute latest version either via Fedora 6's Update
mechanism or by building from source. My database has 1 table named items
which has (itemname varchar, quantity int, productid int). I originally had
the code without the semicolon but added it while trying to track this down.
I just tried it with "select" lowercase but that didn't help.
Thanks for the pointer about exceptions to references, I hadn't thought
about it that way.
try {
if ( conn == NULL ) {
conn = new connection("dbname=testdb user=postgres");
work Xaction(*conn, "DemoTransaction");
result R = Xaction.exec("select * from items");
// Load the dispensers from the database
for (result::size_type i = 0; i != R.size(); ++i) {
std::string itemname = String::F("%s", R[i]["itemname"].c_str());
int quantity, productId;
R[i]["quantity"].to(quantity);
R[i]["productid"].to(productId);
mProductList.Load(productId, itemname, quantity);
}
}
} catch(pqxx::sql_error sqlE) {
return false;
} catch (pqxx::broken_connection bcE) {
return false;
} catch (...) {
return false;
}
-----Original Message-----
From: Jeroen T. Vermeulen [mailto:[EMAIL PROTECTED]
Sent: Thursday, November 16, 2006 11:53 PM
To: Kevin Lambert
Cc: [email protected]
Subject: Re: [libpqxx-general] pqxx::sql_error on "SELECT * FROM ITEMS"
On Fri, November 17, 2006 01:38, Kevin Lambert wrote:
> I have a Fedora Linux box that I just upgraded from FC3 to FC6. This
> code worked fine under FC3 (it would show me all of the times in the
> "items" database. The code is at the end of the email. Under FC6 I
> am receiving a pqxx::sql_error message which has my select statement
> in it. If I do that select statement from psql it works fine. Any
> ideas?
Well, lots of variables there (postgres version, libpq version, libpqxx
version, possible changes in setup, and so on) so we may have to cull those
down a bit. Let's look at the code first, though:
> Try {
Ahem. :)
> conn = new connection("dbname=testdb user=postgres");
>
> work Xaction(*conn, "DemoTransaction");
It's at least possible that you're running into a bug with handling of
upper-case characters in names. It shouldn't be a problem with
transactions, but you could try using an all-lower-case name just to be sure
that this isn't related to the problem.
> result R = Xaction.exec("SELECT * from items;");
This is where it hurts, right?
The semicolon isn't really right there. It *shouldn't* cause any problems,
but who knows, it may. The semicolon is used as a separator between
queries, so it's almost like you're issuing two queries at once
here: first a SELECT and then an empty one.
> for (result::size_type i=0; i != R.size(); ++I ) {
> String itemname = R[i]["itemname"]'
It would help if you could post code that really does reproduce the problem.
This line wouldn't even compile because of the extra quote.
> Int quantity, productId;
> R[i]["quantity"].to(quantity);
> R[i]["productid"].to(productId);
There are no "String" and "Int" types in C++, so are these typedefs for
std::string and int?
> }
> } catch(pqxx::sql_error sqlE) {
> }
This really doesn't have anything to do with the problem, but always try to
catch *references* to exception objects:
} catch (pqxx::sql_error &sqlE) {
If you leave out the reference, an exception object of a type derived from
the one you're catching may get "sliced." In this case, that means that if
the exception is an object of some child class of sql_error, what you will
get is a new object of type sql_error. Some information can be lost, which
is bad, but things get even worse when virtual functions are involved.
Jeroen
_______________________________________________
Libpqxx-general mailing list
[email protected]
http://gborg.postgresql.org/mailman/listinfo/libpqxx-general
