Re: [gnome-db] gda_data_model_describe_column

[Rodrigo Moya <rodrigo gnome-db org>]

On Wed, 2004-11-10 at 10:24 +0100, Vivien Malerba wrote:
> On Tue, 9 Nov 2004 21:48:26 -0700, Neil Zanella <nzanella gmail com> wrote:
> > What's the point of this method: ?
> > 
> > GdaFieldAttributes* gda_data_model_describe_column
> >                                             (GdaDataModel *model,
> >                                              gint col);
> > 
> > Or rahter, what's the point of returning this?
> > 
> > struct GdaFieldAttributes {
> > 
> >         gint defined_size;
> >         gchar *name;
> >         gchar *table;
> >         gchar *caption;
> >         gint scale;
> >         GdaValueType gda_type;
> >         gboolean allow_null;
> >         gboolean primary_key;
> >         gboolean unique_key;
> >         gchar *references;
> >         gboolean auto_increment;
> >         glong auto_increment_start;
> >         glong auto_increment_step;
> >         gint position;
> >         GdaValue *default_value;
> > };
> > 
> > Can we change this struct. What does "gboolean unique_key;" mean? Does it mean
> > the column by itself is a unique key? Make that boolean a pointer to a structure
> > containing all unique key structures in which the given column participates. It
> > would make more sense? A table can have more than one unique key
> > composed of more than one field. Then again, struct GdaFieldAttributes
> > is undocumented so it's hard to make sense of it right now. Anyways,
> > please change it, cause it doesn't seem right. Do this:
> > 
> > struct GdaFieldAttributes {
> > 
> >         gint defined_size;
> >         gchar *name;
> >         gchar *table;
> >         gchar *caption;
> >         gint scale;
> >         GdaValueType gda_type;
> >         gboolean allow_null;
> >         GdaPrimaryKey *primary_key;
> >         GdaUniqueKeys **unique_keys;
> >         GdaReferences **references;
> >         gboolean auto_increment;
> >         glong auto_increment_start;
> >         glong auto_increment_step;
> >         gint position;
> >         GdaValue *default_value;
> > };
> > 
> 
> Personnaly I believe that there is too much information in that
> structure: we are mixing database schema information with resultset
> information; the database schema information can be retreived using
> the specific schema information already provided by libgda.
> 
well, we are returning all information that is known about the fields in
the recordset (or when we know that, since in some cases we don't).

> I and would limit down to the same kind of information for the JDBC
> ResultSetMetaData interface (see
> http://java.sun.com/j2se/1.4.2/docs/api/java/sql/ResultSetMetaData.html),
> it would then become:
> 
> struct GdaFieldAttributes {
>        // name and position
>        gchar *name;
>        gchar *label;
>        gint position;
> 
>        // data type
>        GdaValueType gda_type;
>        gchar *dbms_type;                   <-- Real DBMS type name for
> the column
>        gint precision;
>        gint scale;
> 
>        // optionnal
>        gboolean allow_null;
>        gboolean auto_increment;
>        gchar *table_name;              <-- need to define exactely
> when it is filled and what it means; I doubt its utility.
>
probably this is only needed internally in the data model
implementation, not for data model users.

> };
> 
> Maybe renaming that to GdaDataColumnAttributes would make its purpose
> clear and without ambiguities.
> 
maybe, yes. But removing that information might make it difficult for
data model implementations to be self-contained, for allowing updates,
removals, additions. Not sure.
-- 
Rodrigo Moya <rodrigo gnome-db org>