Re: Call for review: Database API for the GNADE Project

[Michael Erdmann <michael.erdmann@snafu.de>]

Kenneth Almquist schrieb:

> Michael Erdmann <michael.erdmann@snafu.de> wrote:
>
> >    X : Connection.Object;
> >    Connect(X);
> >    declare
> >       q : Query.Object( X );
> >    begin
> >       Statement( q, "select * from employees where name = ?" );
> >       Bind( q, 1, Name );
> >       Execute( q );
> >       .....
> >    end ;
>
> I wonder if replacing the various Bind procedures with generic
> packages would be better:
>
>     declare
>         Statement : constant SQL_Statement
>              := Prepare(Connection, "select * from employees where empno = ?");
>         package Empno is new Set_Statement_Integer(Statement, 1);
>     begin
>         Empno.Set(Name);
>         Result := Execute_Query(Statement);
>         ...

I agree on this.  The problem might be in the future that i have to handle the
null indicator in a save way which means there will be for each data type hidden
variables. Then the best aproach of doing it will be a generic package which
takes the Ada 95 type and provides a save encapsulation of these hidden
parameter. Such a package will have to provide the following procdures/funtions:

Value   - Returns  or set the value
Is_Null -  Item is null

     x : integer

     package integer_vars is new package( type => integer );
     use package;

     .....
     Integer_Vars.Bind( x, 1, statement );


> Set_Statement_Integer is a generic package containing a single procedure
> named Set.  We use a generic package rather than a generic procedure
> so that we can perform validity checking when the package is instantiated,
> rather than performing a check every time a new value is assigned to
> empno.
>
> I've changed Execute to Execute_Query because when we execute a query
> we get back a set of results, whereas when we execute other types of
> SQL statements we get back a status code.
>
> >    result := Execute( statement );
> >    while not End_of( result ) loop
> >       row  := Fetch( result );
> >       name := Get( row, 1 );
> >       boss := Get( row, 2 );
> >          .......
> >    loop;
>
> Is the row value valid after a subsequent fetch operation?  If not,
> we could eliminate the row variable and get the fields of the current
> row directly from the result variable.
>
> Again, generics might be useful to get the row numbers out of the code:
>
>     declare
>         Row : SQL_Row;
>         function Get_Name is new Get_Row_String(1);
>         function Has_Boss is new Get_Row_Is_Null(2);
>         function Get_Boss is new Get_Row_Integer(2);
>     begin
>         ...
>         Row := Fetch(Result);
>         Name := Get_Name(Row);
>         if Has_Boss(Row) then
>             Boss := Get_Boss(Row);
>         else
>             Put_Line("The CEO does not have a boss");
>         end if;
>         ...

I have to admit, this realy looks better!

>
>
> > Mode:  The fetch mode controls the behavior of the cursor in the
> > result set.  Modes are:  Forward, Backward.

> As far as I know, this is not supported by standard SQL, and I can't
> think of a reason why anyone would need this.

There are somtimes applications which like to browse through
result sets. A lot of native data base bindings and ODBC support
this.

> > If the server drops the connection the procedure Disconnection_Event is
> > called.  In order to implement specific behavior, the procedure may be
> > overloaded.
>
> If I understand correctly, Server_Error will be raised after this
> procedure is called.  It seems like this procedure is unnecessary,
> since anything done by this procedure could instead be done by an
> exception handler.

Yes, this was just an option i wanted to provide. But you are right you can
do the same thing capturing the exception. I think i wil remove it because
i made the experience that with such procedure it is quite difficult to
communicate data into them unless you are using some kind of
global data areas!

>
>
> I hope these random thoughts are helpful.
>                                 Kenneth Almquist

Yes, I realy apreciate your effort. I will do the update today! Please keep
an eye on http://gnade.sourceforge.net/ to seen what happens.

Michael