Floats/Numbers, and Changes between Oracle 8,9,10 and 11

Question

Hi
I'm trying to get our application to support Oracle 11, it works fine with Oracle 10 and before. I wonder if someone can help me with this.
Suppose you have an SQL statement with a constant like this:
select my_agent_id, 0 team_id
from my_agent
Note: The second column is a constant (0).
In Oracle 8, 9, 10, when this statement is described, Oracle says that the team_id column is an integer type (number). But in Oracle 11, it reports it as a floating-point type.
This causes errors: our datasets are expecting interger types, but Oracle now says these are floats.
Can anyone help? Please don't say 'take the constant out' because that's quite hard, as we have datasets that are in some cases used with different sql statements. Oracle upgrades are usually easy because Oracle are pretty good at version compatibility.
Here is some additional info.
Our app is an OCI app written using MSVC. The code that says the team_id is a float is here (I have removed error checking code):
               luiResult = OCIAttrGet(     (ORAVOID *)lpssParam,
                                             OCI_DTYPE_PARAM,
                                             (ORAVOID *)&cSc,
                                             0,
                                             OCI_ATTR_SCALE,
                                             (OCIError *)opuoTrans->opssError);
               if (luiResult != OCI_SUCCESS) {     
                    return BINDOUTPUT_GET_PREC;
               if (cSc == 0) {
                    *aiType = DBPROV_INT; // Oracle 8,9,10 goes here
               } else {
                    *aiType = DBPROV_FLOAT; // Oracle 11 returns -127 so here
The value of cSc is -127, the code for FLOAT as written in the manual.
Anyway, this seemingly small thing is enough to make the application go bang.
Any help appreciated.

Answer 1

I believe what you're seeing here is expected behavior. Oracle 9.2.0.3 and later fixed the old behavior. Your code should be looking for a non-zero precision and a scale of -127 to determine that the column is a float.
Metalink 255902.1 Number Datatype Returns -127 instead of 0 for Unspecified Scale seems to be directly on point here.
Justin