Skip to Content.
Sympa Menu

shibboleth-dev - RE: Testing ODBC store

Subject: Shibboleth Developers

List archive

RE: Testing ODBC store


Chronological Thread 
  • From: "Scott Cantor" <>
  • To: <>
  • Subject: RE: Testing ODBC store
  • Date: Tue, 15 Jan 2008 12:41:26 -0500
  • Organization: The Ohio State University

> "Clob" is the Oracle data type for big charachter strings.

Ok, so basically the "text" type in T-SQL.

> What I've undestood is that using big strings directly inside an INSERT
> statement executed directly is not considered to be a best practice.

No, it's not, but doing it the right way depends on a great deal of support
from the ODBC library.

> I'm currently working in order to check if it works on Oracle.

Well, the problem is that I thought you couldn't just bind a variable
directly for a long string type. You have to feed the data in a loop. That's
what didn't work for me. If it's possible to actually just bind the string
directly, I might not have tried that for some reason.

> > I think that the simplicity and reliability of the straight insert for
> > drivers that support it demands that anything for Oracle be conditional.
> > Probably a conditional macro set by configure.
>
> I think that the prepare/bindparam/execute stuff is general enough to be
> used in the general ODBC code.

On Windows, yes. On Unix, I don't think so.

> The problems arise because the datatypes
> (and consequently the ParameterType argument of SQLBindParam) differ
> from db to db, expecially those related to big text fields and to
> timestamps/dates. But maybe it's not too difficult to generalize it or
> to make it ODBC driver-dependant..

ODBC has a generic way to do this, but I'm pretty sure it's the SQLPutData
loop.

My feeling is that *if* the straight insert works, it's better to use it. So
I believe the complex way should be conditionally compiled or turned on only
for databases that demand it.

-- Scott





Archive powered by MHonArc 2.6.16.

Top of Page